Part 1 · the operating model
The five-element consulting model
Every engagement in this guide runs through the same five elements. AI touches all of them, but it changes their cost, never their ownership. The middle element — AI support — is the only one a machine performs; the other four stay anchored to a consultant who is accountable to a client.
Problem statement
Define the client challenge with precision and empathy.
Methodology
Apply structured frameworks and proven tools.
AI support
Augment human judgment with ethical, auditable models.
Client experience
Co-create solutions through transparent collaboration.
Outcome
Deliver measurable impact, accountability and trust.
Read the five as a sentence: you frame the problem, choose the method, let AI carry the computational load, keep the client inside the work, and remain the name on the outcome. Hold that order and AI never quietly becomes the consultant.
New · orientation
How to read every infographic in this guide
Each methodology carries the same diagram — the Augmentation Line. Learn to read it once and you can read all nineteen. The spine is the method's sequence of stages, left to right.
The teal track is genuinely delegable: pattern-finding, drafting, calculation, first-pass synthesis at a scale no human matches. The amber track is what you are paid for: framing the question, judging what is true, weighing trade-offs, carrying accountability. A gate marks a moment where shipping an unchecked AI output would be negligent — a recommendation to a client, a number in a board pack, a decision that touches people or money. No output crosses a gate without a named human's sign-off.
Below each diagram, a collapsible panel holds the full original training script — sub-steps, AI-fit notes, consultant actions and the worked prompt for every stage — preserved intact, with four learning blocks added: a learning objective, the part that stays human, the watch-outs, and a responsible-prompt note.
New · the meta-skill
Prompt like a consultant: the FRAME pattern
This guide gives you a hundred prompts. This is how to write the hundred-and-first. Every strong consulting prompt has the same five parts — skip one and the model fills the gap with an assumption you didn't choose.
F
Frame the role
Tell the model who it is acting as — “a strategy analyst preparing a board paper”. Role sets register, depth and rigour.
R
Reference the context
Give the situation it needs — sanitised. Sector, scale, the decision at hand. Never the client's name or confidential data.
A
Ask precisely
One specific output, not a topic. “Generate five MECE hypotheses,” not “help me think about revenue.”
M
Mark constraints & format
Boundaries, exclusions, length, structure, the format you'll actually paste into a deck or model.
E
Examine the output
The human step. Check the facts, stress-test the logic, strip the hallucinations, and put your name to it.
Notice that the worked prompts throughout this guide already obey FRAME — they assign a role, supply context, ask for a specific artefact, and pin down the format. As you reuse them, change the context and constraints to fit the engagement, and never skip E. The model is a fluent, confident, occasionally wrong analyst who never tells you when it's guessing. Examining is not optional politeness; it is the job.
The one habit that separates fluent from dangerous
Before you paste any AI output into a client-facing artefact, you must be able to say where every number came from and why every claim is true. If you can't, it isn't ready — it's a draft for you to verify, not a deliverable. AI gives you a faster first draft, never a finished answer.
New · the rule before every prompt
Client data: the traffic-light protocol
Before anything goes into a model, classify it. This is the single discipline that keeps you, the firm and the client safe — and in regulated work (financial services especially) it is not advisory, it is the line between good practice and a reportable breach.
Green · go
- Public information and published research
- Synthetic or clearly hypothetical examples
- Generic frameworks, definitions, structures
- Your own reasoning, abstracted from any client
Prompt freely. Still verify the output.
Amber · abstract first
- Internal, non-sensitive working material
- Sector context that could narrow to one client
- Figures that are commercially shaping but not secret
Strip identifiers, round numbers, generalise — then use an approved tool only.
Red · never paste
- Anything client-confidential or under NDA
- Personal data (GDPR) — names, records, PII
- Regulated, price-sensitive or privileged material
- Credentials, source code under restriction
Does not enter a model. No exceptions, no “just this once”.
When in doubt, treat it as the next level up. The test is simple: if this prompt leaked, would the client be harmed and would the firm have to disclose it? If yes, it was Red. Use only firm-approved, contractually covered tools; assume anything else is training on your input. And document, in plain terms, when and how AI was used in an engagement — transparency with the client is part of the deal, not a courtesy.
New · knowing where you stand
The competency ladder
AI fluency is a skill that develops in stages. Know which rung you're on for a given methodology — and don't operate above it unsupervised.
Aware
Watches & checks
Knows what the method is for and where AI fits. Can spot an obviously wrong output. Works under review.
Assisted
Runs the prompts
Uses the supplied prompts correctly, classifies data, verifies outputs, and flags anything uncertain to a senior.
Fluent
Adapts & designs
Writes new prompts with FRAME, judges quality independently, and knows the limits of each tool for each task.
Lead
Owns the method
Sets the approach on an engagement, coaches juniors, makes the gate calls, and stands behind the outcome with the client.
A practical assessment for each methodology: can you (1) state when not to use it, (2) run its prompts and catch a wrong answer, (3) name the one thing that stays human, and (4) explain to a client why AI was used and how you assured the result? Four yeses is Fluent. The fourth alone is what makes you trusted.
Design Thinking
Use when the problem is human-centred and ill-defined — you need to understand real user needs before committing to a solution.
Learning objective
Run a full empathise-to-test loop where AI accelerates research and prototyping, while you protect the framing of the problem and the integrity of what users actually told you.
Stays human
Choosing which problem is worth solving, and sitting with real users. AI can summarise a thousand transcripts; it cannot feel the room or be accountable for the reframe.
Watch-outs
- AI personas drift into stereotypes — validate against real people.
- Sentiment scores look precise but flatten nuance; read the quotes.
- Don't let a tidy AI synthesis end divergence too early.
Responsible prompt note
Interview transcripts are Amber-to-Red: remove names and identifying detail before analysis. Never let AI invent quotes — every quote it returns must be traceable to a real transcript line.
Full training script — 5 stages, sub-steps, AI fit, consultant actions & prompts preserved from source›
1.1Empathise
Sub-steps: conduct user interviews and observations · gather qualitative and quantitative data · map stakeholder ecosystems · identify pain points and emotional drivers.
AI fit
- AI-powered research: NLP analysis of thousands of reviews, posts and support tickets to identify patterns
- Sentiment analysis: process interview transcripts at scale
- Conversation intelligence: analyse service calls for unmet needs
- Persona generation: synthesise research into detailed personas
Consultant action
Conduct 15–20 in-depth stakeholder interviews using semi-structured questionnaires while observing actual user behaviours in context. Synthesise findings into empathy maps that capture what users say, think, feel and do.
Responsible prompt — Empathise
"Analyze these 500 customer support transcripts and identify the top 10 unmet needs expressed by customers. For each need, provide: (1) the frequency of mention, (2) emotional sentiment score, (3) specific quotes that illustrate the need, and (4) categorize by customer segment. Format as a table with actionable insights."
1.2Define
Sub-steps: synthesise research findings · create problem statements · identify patterns and themes · frame the right problem to solve.
AI fit
- Pattern recognition: cluster research data into themes
- Problem-statement generator: draft multiple variations
- Insight extraction: find non-obvious connections in qualitative data
Consultant action
Facilitate a synthesis workshop with cross-functional stakeholders to cluster insights and co-create problem statements. Validate problem framing against business objectives and user needs before proceeding.
Responsible prompt — Define
"Based on these research findings [paste findings], generate 5 alternative problem statements using the format: 'How might we [action] for [user] so that [outcome]?' Rank them by: (1) alignment with business goals, (2) user impact potential, and (3) feasibility. Explain your reasoning for the top recommendation."
1.3Ideate
Sub-steps: brainstorm solutions · challenge assumptions · explore diverse perspectives · generate a wide range of ideas.
AI fit
- Idea generation: produce 100+ concept variations in minutes
- Analogous inspiration: find solutions from unrelated industries
- Constraint-based ideation: generate within specific parameters
- Bias detection: identify groupthink or narrow thinking
Consultant action
Run divergent-thinking sessions using SCAMPER, brainwriting and “worst possible idea” to overcome cognitive fixedness. Ensure psychological safety so all participants contribute freely.
Responsible prompt — Ideate
"Generate 30 innovative solutions for [specific problem] that meet these constraints: budget under $50K, implementation within 3 months, no new hires required. Include: (1) 10 incremental improvements, (2) 10 adjacent innovations, and (3) 10 transformational ideas. For each, note the key assumption that must be true for it to work."
1.4Prototype
Sub-steps: build tangible representations · create low-fidelity models · develop testable artefacts · iterate quickly.
AI fit
- Rapid prototyping: code generators for digital prototypes
- Visual design: mockups and storyboards
- Wireframing: convert sketches to digital wireframes
- Content generation: populate prototypes with realistic copy
Consultant action
Create minimum viable prototypes that test the riskiest assumptions first, starting with paper sketches before investing in high-fidelity versions. Document design decisions and a prototype testing plan.
Responsible prompt — Prototype
"Create a detailed description of a low-fidelity prototype for [service/product concept] that I can build in 2 hours using only paper, sticky notes, and markers. Include: (1) the 3 key user flows to test, (2) specific questions each screen should answer, (3) what materials I need, and (4) step-by-step assembly instructions. Also suggest what NOT to include to keep it truly low-fi."
1.5Test
Sub-steps: gather user feedback · observe interactions · measure effectiveness · refine solutions.
AI fit
- Automated usability testing: analyse session recordings
- A/B testing at scale: optimise multiple variables at once
- Feedback analysis: process open-ended feedback from hundreds of users
- Predictive analytics: predict which iterations will perform best
Consultant action
Conduct moderated usability tests with 5–8 representative users per iteration, using think-aloud protocols to capture cognitive processes. Synthesise findings into a prioritised iteration backlog within 24 hours.
Responsible prompt — Test
"Analyze these 47 user testing session notes and identify: (1) the top 5 usability issues ranked by severity and frequency, (2) patterns in user confusion or errors, (3) unexpected positive reactions or workarounds users discovered, and (4) specific design changes recommended for the next iteration. Present as an action-oriented report with quotes as evidence."
Lean Startup
Use when you're operating under high uncertainty and need to learn fast and cheaply before committing capital — new products, ventures or business models.
Learning objective
Design and run Build–Measure–Learn loops where AI ranks assumptions and crunches results, while you decide what to test, what counts as success, and whether to pivot.
Stays human
The pivot-or-persevere call. AI can show what the data says; the decision to change direction carries judgment, conviction and consequences a model cannot own.
Watch-outs
- AI happily reports significance on tiny samples — check the n.
- “Validated” by a model is not validated by reality.
- Vanity metrics survive if you let AI pick metrics unchecked.
Responsible prompt note
Experiment data is often Amber. Abstract it. Treat AI's statistical verdicts as a draft analysis to be checked by someone who understands the test design — confidence claims are easy to fabricate convincingly.
Full training script — 3 stages, sub-steps, AI fit, consultant actions & prompts preserved from source›
2.1Build (MVP development)
Sub-steps: identify riskiest assumptions · design minimum viable product · develop core features only · create measurable experiments.
AI fit
- Assumption prioritisation: rank assumptions by risk and testability
- MVP scope definition: analyse what's truly “minimum” from market data
- Rapid development: coding assistants to build MVPs faster
- Feature prioritisation: predict which features drive adoption
Consultant action
Facilitate an assumption-mapping session to identify and rank all business-model assumptions by risk level. Define MVP success metrics and a build timeline that maximises learning per dollar spent.
Responsible prompt — Build
"Here is our product concept: [describe concept]. List the 15 riskiest assumptions we're making, categorized by: (1) desirability assumptions, (2) viability assumptions, and (3) feasibility assumptions. For each assumption, recommend: (a) the fastest experiment to test it, (b) what metric would validate or invalidate it, and (c) the minimum sample size needed. Rank by which assumption, if false, would kill the business."
2.2Measure
Sub-steps: define actionable metrics · set up analytics infrastructure · collect real user data · establish baselines.
AI fit
- Metric selection: recommend leading vs lagging indicators
- Automated analytics: dashboards that surface insights automatically
- Cohort analysis: segment users and track behaviour patterns
- Anomaly detection: flag unusual patterns in real time
Consultant action
Establish a metrics framework distinguishing vanity metrics from actionable metrics, focusing on cohort-based analysis. Implement tracking and capture baselines before the experiment begins.
Responsible prompt — Measure
"For a [type of business] testing [specific hypothesis], recommend the 7 most important metrics to track, explaining for each: (1) why it matters for this specific test, (2) how to calculate it, (3) what baseline we should expect, and (4) what threshold would indicate success vs. failure. Also identify 3 vanity metrics we should ignore and explain why they're misleading in this context."
2.3Learn
Sub-steps: analyse experimental results · validate or invalidate hypotheses · extract actionable insights · make pivot/persevere decisions.
AI fit
- Statistical analysis: determine significance automatically
- Insight generation: translate data into plain-language insights
- Pivot recommendations: suggest pivot strategies from results
- Competitive benchmarking: compare results against industry standards
Consultant action
Lead a learning review that separates facts from interpretations and makes a clear pivot/persevere recommendation. Document learnings and update the business-model canvas accordingly.
Responsible prompt — Learn
"Analyze these experiment results: [paste data]. Determine: (1) whether the hypothesis was validated, invalidated, or inconclusive with statistical confidence levels, (2) the 3 most important insights regardless of hypothesis outcome, (3) whether we should pivot, persevere, or stop with specific reasoning, and (4) if pivoting, recommend 3 specific pivot types (zoom-in, zoom-out, customer segment, etc.) with rationale for each. Write this as a 1-page decision memo for leadership."
Jobs to be Done
Use when you need to understand why customers “hire” a product — focusing on the outcome they want, not the features you ship.
Learning objective
Move from features to outcomes: use AI to mine behaviour and quantify importance-vs-satisfaction gaps, while you write the job statements and decide which underserved outcome to chase.
Stays human
Hearing the real “job” behind a switch. The emotional and social dimensions of why someone changed surface in conversation, not in a cluster diagram.
Watch-outs
- AI conflates the job with the current solution — keep them separate.
- Opportunity scores are only as good as the survey design behind them.
- Named-tool claims (Productboard, Pendo, Crayon) date fast — verify currency before relying on them.
Responsible prompt note
Switch-interview transcripts are personal data — anonymise. Don't let AI fabricate outcome statements the data doesn't support; every job statement should trace to evidence.
Full training script — 4 stages, sub-steps, AI fit, consultant actions & prompts preserved from source›
3.1Identify the job
Sub-steps: observe customer behaviours · understand circumstances and context · define the functional job · identify emotional and social jobs.
AI fit
- Behavioural pattern mining: analyse journey data for job patterns
- Job-statement formulation: craft precise statements from raw data
- Context analysis: extract situational factors from interviews
- Job-hierarchy mapping: organise jobs into hierarchies
Consultant action
Conduct “switch interviews” with customers who recently adopted or rejected a solution, focusing on the moment of decision. Map the job timeline from first thought through final outcome, capturing all circumstances and anxieties.
Responsible prompt — Identify the job
"Based on these 20 customer interviews [paste transcripts], write 3 complete job statements using the format: 'When [situation], I want to [motivation], so I can [expected outcome].' For each job, identify: (1) the functional job, (2) the emotional job, (3) the social job, and (4) the consuming tasks vs. the job itself. Include specific quotes that reveal the job clearly."
3.2Uncover needs
Sub-steps: identify desired outcomes · prioritise underserved needs · map pain points and gains · quantify importance and satisfaction.
AI fit
- Outcome extraction: identify desired outcomes from research
- Gap analysis: calculate importance-vs-satisfaction at scale
- Need clustering: group similar needs across segments
- Predictive modelling: predict which unmet needs drive switching
Consultant action
Run outcome-based surveys rating the importance of 50–100 potential outcomes and current satisfaction. Analyse to identify overserved, underserved and appropriately served outcomes.
Responsible prompt — Uncover needs
"Analyze this outcome survey data [paste data with importance and satisfaction scores for 60 outcomes]. Create an opportunity map identifying: (1) the top 10 underserved outcomes (high importance, low satisfaction) with opportunity scores, (2) the 5 overserved outcomes where we could reduce features, and (3) the 3 outcomes that segment customers into different groups. Calculate the opportunity score for each underserved outcome using the formula: Importance + (Importance - Satisfaction)."
3.3Create solutions
Sub-steps: generate solution concepts · match solutions to needs · design value propositions · test solution-job fit.
AI fit
- Solution ideation: create solutions for specific outcome statements
- Value-prop design: craft compelling value propositions
- Solution scoring: predict which solutions best address needs
Consultant action
Facilitate ideation focused on the top 5 underserved outcomes, ensuring each concept directly addresses a specific job outcome. Create value-proposition canvases for the top 3 and test them with target customers.
Responsible prompt — Create solutions
"Generate 15 solution concepts that specifically address these 3 underserved job outcomes: [list outcomes with importance-satisfaction gaps]. For each concept, provide: (1) a one-sentence description, (2) which specific outcome it addresses and how, (3) the mechanism by which it creates value, (4) potential trade-offs or new problems it might create, and (5) a value proposition statement in the format: 'For [customer] who [job situation], our [solution] is a [category] that [benefit]. Unlike [competition], we [differentiator].'"
3.4Measure outcomes
Sub-steps: track job completion rates · measure satisfaction · monitor switching behaviour · assess competitive displacement.
AI fit
- Outcome tracking: monitor job-completion metrics across touchpoints
- Churn prediction: predict switching from unmet jobs
- Competitive intelligence: track how rivals satisfy the same job
Consultant action
Establish a job-performance dashboard tracking how well customers complete the job across time, cost, reliability and emotional satisfaction. Run quarterly reviews to detect shifts in job importance or competitive performance.
Responsible prompt — Measure outcomes
"Create a job performance measurement framework for customers trying to [specific job]. Include: (1) 5 leading indicators that predict successful job completion, (2) 3 lagging indicators that confirm job success, (3) specific data sources and collection methods for each metric, (4) benchmark targets for each metric, and (5) alert thresholds that signal customers are struggling. Also recommend how often to measure each metric and how to visualize trends over time."
Blue Ocean Strategy
Use when competition is brutal and margins are thin — you want to create uncontested market space rather than out-fight rivals on the same factors.
Learning objective
Build a strategy canvas and ERRC grid where AI maps the competitive field and proposes moves, while you make the value-innovation bet and sequence price, cost and adoption.
Stays human
The act of value innovation — deciding what to eliminate and what to create. That courage to diverge from the industry is a leadership judgment, not an optimisation.
Watch-outs
- AI infers competitor positions from public data that may be wrong or stale.
- An AI canvas can look divergent yet be commercially naïve — pressure-test the economics.
- “Create” ideas need real buyer-utility testing, not model confidence.
Responsible prompt note
Competitor and pricing analysis can stray into Amber. Keep inputs public-source. Treat AI competitive claims as hypotheses to verify, not facts to publish.
Full training script — 5 stages, sub-steps, AI fit, consultant actions & prompts preserved from source›
4.1Analyse current state
Sub-steps: map industry boundaries · identify competitive factors · understand buyer utility · assess price corridors.
AI fit
- Industry mapping: analyse structure from vast data sources
- Competitive-factor extraction: identify what rivals compete on
- Market-boundary analysis: find adjacencies and convergence points
- Price analysis: analyse pricing strategies across the industry
Consultant action
Conduct competitive analysis covering direct competitors, substitutes and alternatives customers use. Map the strategic canvas across key factors and identify the industry's dominant logic.
Responsible prompt — Analyse current state
"Analyze the [specific industry] competitive landscape and create a strategy canvas. Identify: (1) the 8-12 key competitive factors the industry competes on, (2) how the top 5 competitors perform on each factor (high/medium/low), (3) the average industry offering level for each factor, and (4) which factors have been raised over time versus which have been ignored. Present as a table and describe the industry's dominant strategic logic in one paragraph."
4.2Reconstruct market boundaries
Sub-steps: look across alternative industries · explore strategic groups · examine buyer groups · scan complementary offerings · challenge functional/emotional orientation · anticipate trends.
AI fit
- Cross-industry analysis: identify patterns from unrelated industries
- Trend forecasting: detect weak signals and emerging trends
- Complementary mapping: map ecosystems and adjacencies
- Buyer-segment discovery: identify underserved buyer groups
Consultant action
Facilitate “six paths” workshops, systematically examining each path. Bring in non-customer and adjacent-industry perspectives to challenge assumptions and reveal opportunities.
Responsible prompt — Reconstruct boundaries
"Apply the six paths framework to [specific industry/market]. For each path, provide 3 specific insights: Path 1 (Look across alternative industries): What industries solve similar problems differently? Path 2 (Strategic groups): What would happen if we moved from [current strategic group] to [different group]? Path 3 (Buyer groups): How would the value proposition change if we focused on [different buyer role]? Path 4 (Complementary offerings): What happens before, during, and after our product is used? Path 5 (Functional/emotional): How could we shift from [current orientation] to the opposite? Path 6 (Trends): What irreversible trends will change the industry? Be specific and provocative."
4.3Apply ERRC grid
Sub-steps: eliminate (what to remove) · reduce (what to minimise) · raise (what to enhance) · create (what to invent).
AI fit
- ERRC brainstorming: suggest elimination/reduction opportunities
- Value-innovation ideas: propose what to create that's never been offered
- Impact prediction: predict the business impact of each ERRC decision
Consultant action
Lead ERRC workshops with cross-functional teams, challenging every industry norm. Force specific actions in each quadrant and calculate the cost savings from eliminate/reduce to fund raise/create.
Responsible prompt — ERRC grid
"Create an ERRC grid for [company/product] in [industry]. ELIMINATE: List 10 factors the industry takes for granted that we could completely eliminate, with estimated cost savings for each. REDUCE: List 8 factors we could reduce well below industry standard, explaining why over-serving exists. RAISE: List 8 factors we should raise well above industry standard to create new demand. CREATE: List 10 factors never offered in the industry that we could create, describing the new value each would generate. Ensure the cost savings from eliminate/reduce fund the raise/create investments."
4.4Build strategy canvas
Sub-steps: plot current state · design future state · visualise value curve · test divergence from competition.
AI fit
- Automated canvas creation: generate from competitive data
- Value-curve optimisation: suggest optimal positioning
- Scenario modelling: simulate different canvas configurations
Consultant action
Create a visual strategy canvas showing current vs future value curves with focus, divergence and a compelling tagline. Test with customers to validate the new value proposition resonates.
Responsible prompt — Strategy canvas
"Design a strategy canvas for [company] showing the transformation from current state to blue ocean future state. The canvas should have 12 competitive factors on the x-axis and offering level on the y-axis. Plot: (1) the industry's current value curve, (2) our current value curve, and (3) our future blue ocean value curve. The future curve must show: focus (fewer factors emphasized), divergence (different shape from competition), and a compelling tagline. Describe what makes the new curve a blue ocean and calculate the expected cost savings versus value created."
4.5Sequence the strategy
Sub-steps: test buyer utility · set strategic price · determine cost targets · address adoption hurdles.
AI fit
- Utility testing: simulate buyer response to value propositions
- Price optimisation: determine optimal price points
- Cost modelling: identify cost drivers and reductions
- Adoption-barrier analysis: predict and address resistance
Consultant action
Validate the strategy through utility testing, set a strategic price that attracts the mass of buyers, and work back to a target cost for profitable delivery. Identify adoption hurdles and plan to overcome them before launch.
Responsible prompt — Sequence strategy
"Sequence the blue ocean strategy for [product/service] by addressing: (1) Buyer Utility: Does it offer exceptional utility across the 6 stages of the buyer experience cycle (purchase, delivery, use, supplements, maintenance, disposal)? Rate each stage and identify gaps. (2) Strategic Price: What price will attract the mass of target buyers while signaling quality? Analyze price corridors of substitutes. (3) Target Cost: What cost must we achieve to profit at the strategic price? Work backward from price. (4) Adoption: What are the 5 biggest adoption hurdles (internal and external) and how will we overcome each? Create an action plan with owners and timelines."
TRIZ — Theory of Inventive Problem Solving
Use when you're stuck on a genuine technical or design contradiction — improving one thing keeps breaking another — and want systematic, not lucky, invention.
Learning objective
Frame a contradiction, map it to inventive principles with AI's help, and evaluate concepts against the ideal final result — keeping the engineering judgment about feasibility yours.
Stays human
Defining the real contradiction and judging physical feasibility. A model can recite the 40 principles; it cannot know whether your materials, budget or physics will bear the solution.
Watch-outs
- AI matches keywords to principles, not the underlying physics — sanity-check.
- It may assert a contradiction is resolved when it's merely moved.
- Patent or domain analogies can be misremembered — verify the source exists.
Responsible prompt note
Proprietary technical detail is Red — abstract the problem to its parameters before prompting. Confirm any cited patent or precedent independently; do not repeat fabricated references to a client.
Full training script — 4 stages, sub-steps, AI fit, consultant actions & prompts preserved from source›
5.1Define the problem
Sub-steps: identify the specific problem · map system components · understand contradictions · define the ideal final result (IFR).
AI fit
- Problem formulation: articulate technical contradictions clearly
- System mapping: diagram components and interactions
- Contradiction identification: detect hidden contradictions
- IFR visualisation: depict the ideal final result
Consultant action
Facilitate problem-definition sessions that move beyond symptoms to the root technical or physical contradiction. Map the system and super-system, then define the IFR where the problem solves itself.
Responsible prompt — Define the problem
"Help me define this TRIZ problem: [describe problem situation]. Provide: (1) The specific problem statement in one sentence, (2) A system map showing all components and their interactions, (3) The technical contradiction stated as: 'Improving [feature A] causes [feature B] to worsen,' (4) Any physical contradiction where one element must have opposing properties, and (5) The Ideal Final Result (IFR) stated as: 'The system itself eliminates the harmful action without increasing complexity or cost.' Be specific and use TRIZ terminology."
5.2Analyse contradictions
Sub-steps: identify technical contradictions · map to the TRIZ contradiction matrix · identify physical contradictions · apply separation principles.
AI fit
- Contradiction-matrix lookup: map problems to the 40 inventive principles
- Pattern matching: find similar contradictions solved in patents
- Separation-principle selection: recommend which applies
Consultant action
Determine whether the problem is a technical contradiction (parameter trade-off) or physical contradiction (opposing requirements). Map to the TRIZ matrix to find the most relevant inventive principles and separation strategies.
Responsible prompt — Analyse contradictions
"Analyze this contradiction: [describe contradiction]. Determine: (1) Is this a technical contradiction or physical contradiction? (2) For technical: Which improving parameter and which worsening parameter from the 39 TRIZ parameters? (3) What are the top 4 inventive principles from the contradiction matrix? (4) For physical: Which separation principle applies (separation in time, space, scale, or condition)? (5) Provide 2 examples from patents or nature where this contradiction was resolved. Explain the logic of each principle for this specific case."
5.3Apply innovative principles
Sub-steps: select from the 40 inventive principles · adapt principles to context · generate solution concepts · combine multiple principles.
AI fit
- Principle generation: apply all 40 principles to a problem
- Cross-domain adaptation: adapt principles from other industries
- Solution synthesis: create detailed solution concepts
Consultant action
Systematically apply the top 3–5 principles, forcing concrete concepts even where they seem impractical. Combine principles where synergistic and adapt solutions from other industries.
Responsible prompt — Apply principles
"Apply these TRIZ inventive principles to solve [specific problem]: Principle #7 (Nesting), Principle #10 (Prior Action), Principle #35 (Parameter Changes). For each principle: (1) Explain what the principle means, (2) Provide 2 examples from other industries, (3) Generate 3 specific solution concepts for our problem, (4) Describe how each concept resolves the contradiction, and (5) Note potential implementation challenges. Then suggest how to combine principles #7 and #10 for a more powerful solution."
5.4Evaluate solutions
Sub-steps: test against the IFR · assess feasibility · check for new contradictions · refine and iterate.
AI fit
- Solution scoring: evaluate against multiple criteria
- Feasibility analysis: assess technical and economic viability
- Contradiction prediction: identify potential new contradictions
Consultant action
Evaluate each concept against the IFR, identifying new contradictions introduced. Run feasibility analysis across technical, economic and organisational dimensions before recommending for prototyping.
Responsible prompt — Evaluate solutions
"Evaluate these 5 TRIZ solution concepts [list concepts] for [problem]. For each concept, assess: (1) IFR Proximity: How close to ideal final result on 1-10 scale, (2) Contradiction Resolution: Does it fully resolve the original contradiction or just move it?, (3) New Problems: What new contradictions or issues does it create?, (4) Feasibility: Technical (1-10), Economic (1-10), Organizational (1-10), (5) Implementation Timeline: Quick win (<3 months), Medium (3-12 months), or Long-term (12+ months). Rank the solutions and recommend which to prototype first with justification."
Hypothesis-Driven Problem Solving
Use when you face a complex business problem on a deadline and need a structured, MECE path from question to recommendation — the default consulting engine.
Learning objective
Drive a hypothesis tree from problem to recommendation, using AI to draft issue trees, generate hypotheses and crunch analysis, while you own the framing, the disconfirming tests and the synthesis.
Stays human
The recommendation. AI builds the scaffolding; the judgment that this is the answer, and that you'd stake your name on it to the client, is yours alone.
Watch-outs
- AI confirms the hypothesis you hint at — ask it to disprove, not support.
- “MECE” from a model is often neither — check for overlaps and gaps yourself.
- It will assert statistical significance it cannot justify; demand the working.
Responsible prompt note
Client data behind the analysis is Amber-to-Red — abstract it. Any figure AI produces that lands in a deck must be reproducible from a source you can name. This is a gated method: the recommendation needs partner sign-off.
Full training script — 6 stages, sub-steps, AI fit, consultant actions & prompts preserved from source›
6.1Define the problem
Sub-steps: clarify the question · set scope and boundaries · identify success criteria · align stakeholders.
AI fit
- Problem scoping: analyse statements for clarity and completeness
- Stakeholder analysis: extract concerns from meeting notes
- Success-criteria definition: suggest measurable outcomes
Consultant action
Facilitate a problem-definition workshop to create a specific, actionable problem statement that distinguishes symptoms from root causes. Document scope, exclusions and SMART success criteria.
Responsible prompt — Define the problem
"Help me refine this problem statement: '[initial vague problem statement].' Transform it into a consulting-grade problem statement that includes: (1) The specific decision that needs to be made, (2) The criteria for success with measurable targets, (3) The scope boundaries (what's in and out), (4) The constraints (time, budget, resources), (5) The key stakeholders and their interests. Format as a one-page problem definition document that could be signed off by the steering committee."
6.2Structure the problem (issue tree)
Sub-steps: break down into MECE components · create a logic tree · identify key drivers · prioritise branches.
AI fit
- MECE validation: check for overlaps and gaps
- Tree generation: create issue trees from problem statements
- Driver identification: identify highest-impact factors
- Prioritisation: rank branches by importance and solvability
Consultant action
Create a hypothesis-driven issue tree breaking the problem into MECE components, typically 3–5 levels deep. Prioritise branches by potential impact and probability of being root cause, focusing effort on the highest-value areas.
Responsible prompt — Structure the problem
"Create a MECE issue tree for this problem: '[problem statement].' Structure it as: Level 1: 3-5 major hypotheses that could explain the problem, Level 2: For each hypothesis, 3-4 sub-issues that would need to be true, Level 3: Specific analyses or data needed to test each sub-issue. Ensure the tree is truly MECE with no overlaps or gaps. Then prioritize the branches using an impact vs. testability matrix, recommending which 3 branches to investigate first and why. Present as a visual tree structure with clear logic flow."
6.3Formulate hypotheses
Sub-steps: develop initial hypotheses · state hypotheses clearly · identify assumptions · determine what would prove/disprove.
AI fit
- Hypothesis generation: propose multiple hypotheses from available data
- Assumption mapping: surface hidden assumptions
- Falsifiability testing: suggest what evidence would disprove each
Consultant action
Develop 3–5 mutually exclusive hypotheses, each a clear, testable proposition. Document the key assumptions and define criteria that would validate or invalidate each one.
Responsible prompt — Formulate hypotheses
"Generate 5 distinct hypotheses to explain [problem situation]. For each hypothesis, provide: (1) Clear statement in 'If...then...' format, (2) The 3-5 key assumptions that must be true for this hypothesis to hold, (3) What specific evidence would prove this hypothesis (must be observable and measurable), (4) What evidence would disprove it (falsifiability criteria), (5) Estimated probability this is the root cause (1-100%), and (6) The cost/time to test this hypothesis. Rank by likelihood and recommend a testing sequence."
6.4Design analysis plan
Sub-steps: identify required data · select analytical methods · assign workstreams · set timeline.
AI fit
- Data-requirements mapping: specify exactly what data is needed
- Method selection: recommend best analytical approaches
- Work breakdown: create detailed project plans
- Resource optimisation: allocate team members to workstreams
Consultant action
Create a detailed plan mapping each hypothesis to data, methods and owners. Establish a timeline with milestones, scheduling quick wins early and deeper analysis later.
Responsible prompt — Design analysis plan
"Create an analysis plan to test these 5 hypotheses [list hypotheses]. For each hypothesis, specify: (1) Data needed (internal systems, external sources, primary research), (2) Analytical method (quantitative analysis, qualitative research, benchmarking, modeling), (3) Owner (which team member), (4) Timeline (start/end dates), (5) Dependencies (what must be completed first), (6) Deliverable format (dashboard, report, presentation). Create a Gantt chart showing the critical path and identify which analyses can run in parallel. Flag any data availability risks and suggest workarounds."
6.5Conduct analysis
Sub-steps: gather data · perform quantitative/qualitative analysis · test hypotheses · synthesise findings.
AI fit
- Data collection: web scraping and aggregation
- Statistical analysis: run complex analyses automatically
- Pattern detection: find non-obvious patterns
- Synthesis: turn analysis into insights
Consultant action
Execute the plan, maintaining rigorous data-quality standards and documenting assumptions. Run weekly synthesis sessions to integrate findings, updating hypotheses on emerging evidence and pivoting analysis as needed.
Responsible prompt — Conduct analysis
"Analyze this dataset [paste or describe data] to test the hypothesis that '[specific hypothesis].' Perform: (1) Descriptive statistics showing distributions, trends, and outliers, (2) Correlation analysis between key variables, (3) Statistical significance testing where appropriate (t-tests, chi-square, ANOVA), (4) Segmentation analysis to identify patterns across different groups, (5) Visualization recommendations that would best communicate the findings. State clearly whether the hypothesis is supported, refuted, or inconclusive with confidence levels. Identify any surprising patterns that warrant further investigation."
6.6Iterate and refine
Sub-steps: validate or reject hypotheses · refine based on findings · develop recommendations · build implementation plan.
AI fit
- Recommendation generation: formulate actionable recommendations
- Impact modelling: predict outcomes of different recommendations
- Risk assessment: identify implementation risks
Consultant action
Lead validation sessions testing findings against the original criteria, rejecting or refining hypotheses on evidence. Develop specific, actionable recommendations with roadmaps spanning quick wins, medium-term initiatives and transformations.
Responsible prompt — Iterate and refine
"Based on these analysis findings [summarize key findings], develop recommendations for [client situation]. For each recommendation, provide: (1) Specific action (what exactly to do), (2) Rationale (why this, citing evidence from analysis), (3) Expected impact (quantified where possible: revenue, cost, time, quality), (4) Implementation requirements (resources, budget, timeline, capabilities needed), (5) Risks and mitigation strategies, (6) Success metrics and how to measure them. Prioritize recommendations into: Quick Wins (0-3 months), Strategic Initiatives (3-12 months), and Transformations (12+ months). Create an executive summary that tells a compelling story from problem to solution."
Cynefin Framework
Use when you need to match your decision approach to the kind of problem you actually have — before applying a method built for the wrong domain.
Learning objective
Diagnose which domain a situation occupies and apply the matching response, using AI to structure the assessment and design safe-to-fail probes — while you make the categorisation call.
Stays human
The domain judgment, especially the dangerous slide from “clear” to chaotic. Mis-categorising is the cardinal Cynefin error and no model can be accountable for it.
Watch-outs
- AI defaults to “complicated” (its comfort zone) — push it to justify complex/chaotic.
- Best-practice answers are dangerous in complex contexts; don't accept them by default.
- A confident classification is still just a hypothesis.
Responsible prompt note
Situation descriptions can carry sensitive context — sanitise. Use AI to widen your thinking about domain, not to make the final call; in chaotic situations, act on human judgment first.
Full training script — 4 stages, sub-steps, AI fit, consultant actions & prompts preserved from source›
7.1Assess the situation
Sub-steps: gather contextual information · identify problem characteristics · understand causality · map constraints.
AI fit
- Situation analysis: process situation reports, extract characteristics
- Causality mapping: identify cause-effect relationships
- Constraint identification: surface explicit and implicit constraints
Consultant action
Conduct a situational assessment gathering multiple stakeholder perspectives. Identify knowns, unknowns and the nature of cause-effect relationships to prepare for domain categorisation.
Responsible prompt — Assess the situation
"Analyze this situation: [describe organizational challenge]. Extract: (1) What is known with certainty, (2) What is unknown but knowable, (3) What is unknowable in advance, (4) The nature of cause-effect relationships (clear, discoverable, only retrospectively coherent, or incoherent), (5) Explicit constraints (rules, resources, time), (6) Implicit constraints (culture, politics, capabilities), (7) The level of urgency and consequences of wrong decisions. Organize this assessment to prepare for Cynefin domain categorization."
7.2Categorise the domain
Sub-steps: determine if Clear (obvious) · Complicated · Complex · Chaotic · or Confused.
AI fit
- Domain classification: categorise situations against Cynefin patterns
- Pattern recognition: detect which domain characteristics are present
- Multi-domain detection: identify when problems span domains
Consultant action
Facilitate a categorisation workshop using diagnostic questions to find the primary domain. Recognise that different aspects may sit in different domains and map them accordingly.
Responsible prompt — Categorise the domain
"Categorize this situation into the appropriate Cynefin domain: [describe situation]. Apply these diagnostic questions: (1) Are best practices clear and universally accepted? (CLEAR), (2) Are there multiple right answers requiring expert analysis? (COMPLICATED), (3) Can we only understand cause-effect in retrospect and need to probe-sense-respond? (COMPLEX), (4) Is there no time for analysis and we must act-sense-respond? (CHAOTIC), (5) Are we unsure which domain applies? (CONFUSED). Provide your domain assessment with evidence for each characteristic. If the situation spans multiple domains, map which aspects belong where and identify the primary domain for decision-making."
7.3Apply the appropriate response
Sub-steps: Clear → sense-categorise-respond · Complicated → sense-analyse-respond · Complex → probe-sense-respond · Chaotic → act-sense-respond.
AI fit
- Clear: expert systems for best-practice recommendations
- Complicated: diagnostic tools and good-practice analysis
- Complex: design safe-to-fail experiments, detect emerging patterns
- Chaotic: rapid situation assessment and triage
Consultant action
Apply the response pattern matching the domain — best practice for clear, expert analysis for complicated, safe-to-fail experiments for complex, immediate stabilising action for chaotic — and avoid importing approaches from the wrong domain.
Responsible prompt — Apply response (Complex)
"Design an intervention strategy for this [COMPLEX domain] situation: [describe]. Following the probe-sense-respond pattern, recommend: (1) 5 safe-to-fail experiments we can run in parallel (each should be low-cost, reversible, and provide learning), (2) What signals to monitor to sense what's emerging (leading indicators, not just lagging metrics), (3) How to amplify positive patterns that emerge, (4) How to dampen negative patterns, (5) When to shift from probing to responding more decisively. Ensure experiments are truly safe-to-fail (not just small-scale failures) and describe what success looks like for each."
7.4Monitor for domain shifts
Sub-steps: track changing conditions · detect domain transitions · adapt approach accordingly · prevent complacency.
AI fit
- Real-time monitoring: continuously assess situation characteristics
- Transition detection: identify early signals of domain shifts
- Adaptive recommendations: suggest when to change approach
Consultant action
Establish monitoring to detect when situations shift between domains — complex becoming complicated as patterns emerge, or clear becoming complex under disruption — and reassess regularly.
Responsible prompt — Monitor for shifts
"Create a domain shift monitoring system for this situation: [describe]. Specify: (1) The current domain and why, (2) The most likely domain shifts (e.g., complex → complicated as we learn, or clear → complex due to disruption), (3) Early warning signals for each potential shift (what to monitor), (4) Trigger points that indicate a shift has occurred, (5) How the response approach must change for each potential new domain, (6) Who is responsible for monitoring and calling the shift. Create a one-page dashboard showing these signals and thresholds for quick assessment."
Scenario Planning
Use when the future is genuinely uncertain and a single-point forecast would be false confidence — you need strategy robust across several plausible worlds.
Learning objective
Build a set of divergent, internally consistent scenarios and stress-test strategy against each — using AI to scan forces and draft narratives, while you choose the axes and read the strategic implications.
Stays human
Selecting the two critical uncertainties that define the scenario space. That framing choice determines everything downstream and is an act of strategic judgment.
Watch-outs
- AI tends to write best-case/worst-case, not genuinely divergent worlds — insist on plausibility, not extremity.
- Trend extrapolation hides discontinuities; the interesting scenarios break the trend.
- Vivid narratives can smuggle in the model's biases — interrogate the assumptions.
Responsible prompt note
PESTLE scans pull public data — verify before citing. Keep the focal question abstract enough that it doesn't reveal confidential strategy. Scenarios inform a strategic decision, so the implications cross a sign-off gate.
Full training script — 6 stages, sub-steps, AI fit, consultant actions & prompts preserved from source›
8.1Define focal question
Sub-steps: clarify the strategic decision · set time horizon · define scope · identify key stakeholders.
AI fit
- Question formulation: craft precise, strategic focal questions
- Horizon analysis: recommend time horizons from industry dynamics
- Stakeholder mapping: identify all relevant stakeholders
Consultant action
Facilitate sessions with senior leadership to craft a clear focal question — specific enough to guide scenarios, broad enough to explore multiple futures — with a horizon set to the industry's rate of change.
Responsible prompt — Focal question
"Help craft a focal question for scenario planning for [organization] facing [situation]. The question should: (1) Be strategic rather than tactical, (2) Address a critical uncertainty that matters to the organization's future, (3) Be open-ended (not yes/no), (4) Have a time horizon of [X years] appropriate for the industry, (5) Be actionable (the organization can influence outcomes). Provide 5 alternative formulations of the focal question, critique each for clarity and strategic importance, and recommend the best one with justification. Also list the 10 key stakeholders who should be involved in developing and using these scenarios."
8.2Identify key factors
Sub-steps: list environmental forces · identify predetermined elements · map critical uncertainties · assess driving forces.
AI fit
- Environmental scanning: scan vast sources for trends and forces
- PESTLE analysis: extract Political, Economic, Social, Tech, Legal, Environmental factors
- Uncertainty identification: separate truly uncertain from predetermined
Consultant action
Conduct PESTLE scanning to identify forces affecting the focal question, distinguishing predetermined elements from critical uncertainties (high impact, high uncertainty).
Responsible prompt — Identify key factors
"Conduct a PESTLE analysis for [industry/organization] to identify forces affecting [focal question]. For each PESTLE category, list: (1) 5-7 key trends or forces, (2) Whether each is predetermined (relatively certain) or a critical uncertainty, (3) The potential impact on the focal question (high/medium/low), (4) The time horizon for each force to materialize. Then identify the 20 most important factors overall and rank them by: (a) Impact on the focal question, and (b) Degree of uncertainty. Highlight the top 10 critical uncertainties (high impact, high uncertainty) that will form the basis for scenario axes."
8.3Rank by impact and uncertainty
Sub-steps: assess importance of each factor · evaluate degree of uncertainty · plot on impact/uncertainty matrix · select critical uncertainties.
AI fit
- Impact scoring: predict which factors have highest impact
- Uncertainty quantification: calculate uncertainty from data
- Automated plotting: create 2×2 matrices automatically
Consultant action
Facilitate scoring sessions rating each factor on impact and uncertainty. Plot all factors on a 2×2 and select the two most important, most uncertain and most orthogonal as scenario axes.
Responsible prompt — Rank factors
"Help me rank these 20 factors [list factors] by impact and uncertainty for scenario planning. For each factor, provide: (1) Impact score (1-10) with justification of how it affects [focal question], (2) Uncertainty score (1-10) with explanation of what makes it unpredictable, (3) Time horizon (when this factor will be resolved or materialize), (4) Interdependencies with other factors (which factors influence or are influenced by this one). Create a 2x2 impact/uncertainty matrix plotting all 20 factors. Recommend the top 2 factors to use as scenario axes, explaining why they are: (a) Most important, (b) Most uncertain, and (c) Most independent of each other."
8.4Develop scenario logics
Sub-steps: select 2–3 critical uncertainties · create scenario axes · define scenario quadrants · develop narrative logic.
AI fit
- Axis selection: recommend uncertainties that create most divergent scenarios
- Scenario generation: create detailed scenario narratives
- Consistency check: validate scenario logic
Consultant action
Define the two axes from the selected uncertainties, creating four distinct quadrants. Develop each scenario's internal logic — plausible, internally consistent and meaningfully different — avoiding simple best/worst-case thinking.
Responsible prompt — Scenario logics
"Develop 4 scenario logics using these 2 axes: Axis 1: [uncertainty 1, e.g., 'Level of government regulation: High to Low'], Axis 2: [uncertainty 2, e.g., 'Rate of technological disruption: Fast to Slow']. For each quadrant, provide: (1) A descriptive name that captures the essence (not just 'Quadrant 1'), (2) The core logic of how these forces interact in this scenario, (3) 3-4 key characteristics that define this world, (4) Why this scenario is plausible (not just possible), (5) How this scenario differs fundamentally from the other three. Ensure each scenario is challenging, relevant, and avoids being obviously good or bad. Write a 2-paragraph narrative description of each scenario's world."
8.5Flesh out scenarios
Sub-steps: develop detailed narratives · identify implications · create scenario indicators · name scenarios memorably.
AI fit
- Narrative writing: write rich, detailed scenario stories
- Implication analysis: generate strategic implications
- Indicator identification: suggest early-warning signals
- Creative naming: generate memorable scenario names
Consultant action
Write rich narratives that bring each future to life across economy, society, technology and competition. Identify strategic implications and a set of indicators that would signal which scenario is emerging.
Responsible prompt — Flesh out scenarios
"Flesh out this scenario: [scenario name and basic logic]. Create: (1) A vivid 3-page narrative set in [target year] written in present tense as if you're living in that future, describing the political, economic, social, technological, legal, and environmental landscape, (2) How [specific industry] operates in this world, (3) Who wins and who loses, (4) What [organization] would need to do to thrive, (5) 10 strategic implications for [organization] specific to this scenario, (6) 15 early warning indicators that this scenario is emerging (mix of leading and lagging indicators), (7) A memorable, evocative name for the scenario. Make the narrative specific and concrete, not abstract."
8.6Test strategies
Sub-steps: stress-test current strategy · evaluate options across scenarios · identify robust strategies · develop contingency plans.
AI fit
- Strategy simulation: model how strategies perform in each scenario
- Robustness analysis: identify strategies that work across scenarios
- Risk quantification: calculate downside risks
Consultant action
Facilitate leadership in testing strategy against each scenario — “would this work in this world?” Identify robust strategies, scenario-specific options and no-regret moves, and build contingency plans with trigger points.
Responsible prompt — Test strategies
"Test this strategy [describe current or proposed strategy] against these 4 scenarios [list scenarios]. For each scenario, assess: (1) Would this strategy succeed, partially work, or fail in this world? Why? (2) What aspects of the strategy are strong in this scenario? (3) What aspects would fail or need modification? (4) What new opportunities does this scenario create that the strategy doesn't capture? (5) What threats does this scenario pose that the strategy doesn't address? Then across all scenarios, identify: (a) Robust elements that work in 3-4 scenarios, (b) Vulnerable elements that only work in 1 scenario, (c) No-regret moves that should be done regardless of which scenario emerges, (d) Contingency plans needed for scenario-specific risks with trigger points. Recommend strategic adaptations based on this analysis."
Wardley Mapping
Use when you need to see how a value chain is evolving — which components are commodities, which are still genesis — so strategy follows the landscape instead of fighting it.
Learning objective
Build a map anchored on user need, position every component on the evolution axis, and read strategic moves off the terrain. AI accelerates component discovery and evolution scoring; you own the positioning calls and the bets.
Stays human
Where a component truly sits on the evolution axis is a judgement, not a fact a model can settle. Mis-place it and the whole map lies to you. Anchoring on the real user need — not the one that flatters your offer — is also yours.
Watch-outs
- AI will assert evolution stages with false confidence — demand the evidence behind each.
- A tidy map is not a correct one; pressure-test the value chain with people who run it.
- Doctrine and climatic patterns need a human who has read Wardley, not a paraphrase.
Responsible prompt note
Component lists drawn from internal architecture or supplier contracts are frequently Amber. Describe capabilities generically rather than pasting system names, vendors or commercial terms.
Full training script — 6 stages, sub-steps, AI fit, consultant actions & prompts preserved from source›
9.1Define purpose
Sub-steps: identify mission/goal · clarify scope · define success · set boundaries.
AI fit
- Mission clarity: help articulate clear, actionable missions
- Scope definition: identify appropriate boundaries
Consultant action
Work with leadership to define a clear, user-centric mission the map will serve. Set scope boundaries so the map is neither too narrow (missing context) nor too broad (unmanageable), and define what success looks like.
Responsible prompt — Define purpose
"Help define the purpose for a Wardley map for [organization]. Craft: (1) A clear mission statement in the format 'Meet [user need] for [user] better than [alternatives],' (2) The scope boundaries (what's in scope, what's explicitly out of scope), (3) Success criteria for the mapping exercise (what decisions will this inform?), (4) The key stakeholders who need to be involved, (5) The time horizon for strategic planning. Ensure the mission is user-focused (not product-focused) and specific enough to guide mapping decisions. Provide 3 alternative mission statements and recommend the best one with justification."
9.2Identify user needs
Sub-steps: map user journeys · list user needs · prioritise needs · validate with data.
AI fit
- Journey mapping: analyse user data and map journeys
- Need extraction: identify needs from user research
- Prioritisation: rank needs by importance
Consultant action
Conduct user research and journey mapping to surface all the needs users are trying to meet. Prioritise by importance and frequency so the map focuses on what matters to users, not what the organisation wants to sell.
Responsible prompt — Identify user needs
"Identify user needs for [mission]. Create: (1) A user journey map showing all steps from [starting point] to [final outcome], including emotional highs and lows, (2) A comprehensive list of 30-50 user needs across the journey (not solutions, but what users are trying to accomplish), (3) For each need: importance to user (1-10), frequency (daily/weekly/monthly/occasionally), and current satisfaction (1-10), (4) Prioritization of needs using an importance-satisfaction gap analysis to identify the biggest opportunities, (5) Validation sources (data, research, assumptions) for each need. Present as a table sorted by opportunity size (importance × (importance - satisfaction))."
9.3Map value chain
Sub-steps: list all components · identify dependencies · map relationships · create initial map.
AI fit
- Component discovery: identify components from documentation
- Dependency mapping: map complex dependencies
- Visualisation: generate Wardley map diagrams
Consultant action
Identify every component — capabilities, activities, data, knowledge, infrastructure — needed to meet user needs, arranged from user-visible at the top to foundational at the bottom, with dependencies mapped.
Responsible prompt — Map value chain
"Create a value chain map for [mission/user need]. List all components required, organized as: User-facing layer (what users directly interact with), Interaction layer (how users engage with the system), Process layer (activities and workflows), Data layer (information and knowledge), Infrastructure layer (technology and physical resources). For each component, specify: (1) Name and description, (2) Which user need(s) it supports, (3) What components it depends on, (4) What components depend on it, (5) Whether it's built in-house, bought, or outsourced. Create a visual representation showing the value chain flow from user needs at the top through all components to foundational elements at the bottom. Ensure you capture 40-60 components for a complete map."
9.4Assess evolution stage
Sub-steps: evaluate each component's maturity · place on evolution axis (Genesis → Custom → Product → Commodity) · identify evolution patterns · predict future evolution.
AI fit
- Stage classification: classify components by evolution stage
- Pattern recognition: identify evolution patterns across components
- Evolution prediction: predict when components evolve to the next stage
Consultant action
Assess each component's stage on the horizontal axis — Genesis, Custom-built, Product (+rental) or Commodity (+utility) — looking for patterns and predicting future evolution. This positioning is the crux of the whole map.
Responsible prompt — Assess evolution stage
"Assess the evolution stage for each component on this Wardley map [list components]. For each component, determine if it's in: Genesis (newly emerging, high uncertainty, low certainty), Custom-built (early market, rapidly changing, competitive differentiation), Product (+rental) (standardized, defined market, volume operations), or Commodity (+utility) (industrialized, high volume, low margin, focus on cost). Provide evidence for each assessment: (1) Market maturity indicators, (2) Level of standardization, (3) Number of suppliers, (4) Rate of change, (5) Focus of competition (innovation vs. cost). Then identify: (a) Components likely to evolve in the next 12-24 months and why, (b) Components that should evolve but aren't (barriers), (c) Patterns in evolution across the map, (d) Strategic implications of the evolution landscape."
9.5Identify strategic moves
Sub-steps: apply strategic patterns · identify opportunities · assess risks · prioritise actions.
AI fit
- Pattern matching: suggest applicable patterns from Wardley's catalogue
- Opportunity identification: spot strategic opportunities
- Action prioritisation: rank moves by impact
Consultant action
Apply Wardley's strategic patterns — focus on user needs, reduce inertia, manage evolution, exploit the landscape — to identify specific moves, then prioritise by impact, feasibility and fit with the organisation's capabilities.
Responsible prompt — Identify strategic moves
"Identify strategic moves for this Wardley map [describe map]. Apply these strategic patterns: (1) Focus: Are we organized around user needs or internal structure? (2) Reduce Inertia: Where is organizational inertia blocking necessary change? (3) Manage Evolution: Which components should we accelerate or slow in their evolution? (4) Exploit Landscape: What opportunities exist based on component positions? For each pattern, recommend 3-5 specific actions with: (a) What to do, (b) Why it matters (strategic rationale), (c) Expected impact (high/medium/low), (d) Effort required (high/medium/low), (e) Timeline, (f) Risks and mitigations. Prioritize all actions using an impact/effort matrix and recommend the top 5 to execute first with justification."
9.6Monitor and adapt
Sub-steps: track component evolution · update maps regularly · adjust strategy · learn from outcomes.
AI fit
- Automated monitoring: track market evolution signals
- Map updates: suggest updates based on new data
- Strategy adjustment: recommend strategic pivots
Consultant action
Establish a cadence (typically quarterly, or on major change) to update the map — tracking component evolution, market shifts and competitor moves — with feedback loops that learn from strategic actions and adjust both map and strategy.
Responsible prompt — Monitor and adapt
"Create a Wardley map monitoring and adaptation system. Specify: (1) Update frequency (quarterly, semi-annually, or trigger-based), (2) Key signals to monitor for each component (market developments, technology changes, competitor moves, customer behavior shifts), (3) Data sources for each signal (news, market research, customer feedback, competitive intelligence), (4) Who is responsible for monitoring each area, (5) Trigger points that warrant immediate map updates (not waiting for scheduled review), (6) Process for updating the map and communicating changes, (7) How to track whether strategic moves are working (metrics and review cadence), (8) Learning loops to capture insights from outcomes and adjust strategy. Create a one-page dashboard template showing the monitoring framework."
Root Cause Analysis — 5 Whys & Fishbone
Use when something has gone wrong and you must reach the cause, not the symptom, before spending money on a fix — incidents, recurring defects, missed targets, process failures.
Learning objective
Drive from a symptom to a verified root cause using the 5 Whys and an Ishikawa (fishbone) diagram, then design countermeasures that stop recurrence rather than masking it. AI widens the set of candidate causes and drafts the diagram; you test causality against evidence.
Stays human
Causation. AI can propose a plausible chain, but only evidence and domain knowledge confirm that fixing X actually removes the problem. The discipline of stopping at the true root — not a convenient one that blames a junior or an absent vendor — is a human responsibility.
Watch-outs
- Five Whys can manufacture a tidy single cause where reality has several — don't force a single line.
- AI-generated chains sound causal even when they're just correlation; demand the evidence at each step.
- Blame-shaped root causes (“human error”) usually mean you stopped one Why too early.
Responsible prompt note
Incident detail is often Amber or Red — names, customers, dates, safety data. Anonymise before prompting. Never let an AI write the official cause statement for a regulated or safety incident; it drafts, a qualified human owns it.
Full method — 5 stages, sub-steps, AI fit, consultant actions & prompts added resource›
R.1Define the effect
Sub-steps: state the problem as an observable effect · quantify it (magnitude, frequency, when it started) · set the boundary of what's in scope · agree what “solved” looks like.
AI fit
- Problem framing: tighten a vague complaint into a measurable effect statement
- Data triage: summarise logs, tickets or defect records into a timeline
- Scope check: flag where the statement smuggles in an assumed cause
Consultant action
Write a single, specific effect statement — what is happening, to what, how much, since when — that contains no presumed cause. Confirm it with the people closest to the work before going further.
Responsible prompt — Define the effect
"Turn this rough problem description into a rigorous root-cause problem statement: [anonymised description]. Produce: (1) a one-sentence effect statement with no assumed cause, written as 'what is happening, to what/whom, how much, since when', (2) the metrics that quantify it and the baseline before it appeared, (3) what is explicitly in and out of scope, (4) a clear definition of 'resolved'. Flag any place where my original wording already assumes a cause."
R.2Trace the causal chain (5 Whys)
Sub-steps: ask “why” iteratively · follow each branch to a controllable cause · stop at root, not at blame · capture the logic explicitly.
AI fit
- Chain drafting: propose candidate why-chains for each branch
- Branch detection: spot where one effect has multiple parents
- Stop-test: challenge whether a step is really a root or a symptom
Consultant action
Run the Whys live with the team, testing each link with “is that actually true, and would removing it remove the effect?” Keep going past the first human-error answer to the system that allowed it.
Responsible prompt — 5 Whys
"For this verified effect statement [statement], generate up to 3 distinct 5-Whys chains that could each explain it. For every step, state: (1) the claimed cause, (2) what evidence would confirm or refute it, (3) whether it is controllable by us. Mark any step that is correlation rather than demonstrated causation. Do NOT collapse to a single chain — show the alternatives so we can test them. Flag where a chain stops at 'human error' and suggest the system-level Why beneath it."
R.3Build the fishbone (Ishikawa)
Sub-steps: place the effect at the head · branch causes by category (e.g. People, Process, Plant/Equipment, Materials, Measurement, Environment) · brainstorm within each · cluster and de-duplicate.
AI fit
- Category prompts: generate candidate causes under each 6M heading
- Blind-spot scan: surface categories the team is ignoring
- Clustering: group overlapping causes across branches
Consultant action
Facilitate a cross-functional group so each category is populated by people who actually own it. Use AI's list to provoke, not to replace, the room's knowledge — then prune to the causes worth testing.
Responsible prompt — Fishbone
"Build an Ishikawa (fishbone) cause inventory for this effect: [statement]. Use the categories People, Process, Equipment, Materials, Measurement, Environment. Under each, list 4-6 candidate causes phrased as testable statements, and note which we could verify with data we plausibly hold. Then identify: (a) the 2 categories most likely to contain the root given the symptom pattern, (b) any category that looks under-explored, (c) causes that appear under more than one category (systemic signals). Present as a structured list ready to transfer onto a diagram."
R.4Verify the root cause
Sub-steps: form a falsifiable hypothesis · gather data · test (does removing it remove the effect?) · rule competing causes in or out.
AI fit
- Test design: suggest the cheapest discriminating test per hypothesis
- Data analysis: check whether the data pattern fits the proposed cause
- Competing causes: list what each rival cause would predict differently
Consultant action
Insist on evidence before declaring a root cause. Where you can, alter the suspected cause and watch the effect respond. Document what would have changed your mind — that is how you know it is rigour, not confirmation bias.
Responsible prompt — Verify
"I have these candidate root causes [list]. For each, give me: (1) a falsifiable hypothesis, (2) the single most discriminating test to confirm or kill it, (3) what result would prove it, (4) what result would disprove it, (5) what a competing cause would predict instead. Then design a short verification plan that tests them in the order that resolves the most uncertainty fastest. State clearly what evidence would make us reject our current favourite cause."
R.5Counter & prevent recurrence
Sub-steps: design countermeasures at the root · distinguish containment from permanent fix · add a control that prevents return · confirm the fix held.
AI fit
- Countermeasure options: generate fixes at root, not symptom
- Control design: propose poka-yoke / monitoring to prevent recurrence
- Side-effect scan: predict second-order consequences of each fix
Consultant action
Choose countermeasures that address the verified root, separate quick containment from the durable fix, and build in a control plus a date to confirm recurrence has stopped. Own the recommendation to the client.
Responsible prompt — Countermeasures
"For this verified root cause [cause], propose countermeasures. Separate them into: (1) immediate containment (stops the bleeding now), (2) permanent corrective action (removes the root), (3) preventive control (stops it returning — e.g. error-proofing, a check, a monitored metric). For each, give expected effect, effort, owner, and any second-order risk it introduces. Recommend a sequence and define the metric and review date that will confirm the problem has not recurred."
Pyramid Principle & Storylining
Use when you have the analysis and now must communicate it — a recommendation, a board paper, a deck — so a busy executive grasps the answer in thirty seconds and the logic holds underneath.
Learning objective
Structure any communication answer-first using Minto's pyramid: governing thought on top, MECE supporting arguments beneath, evidence at the base — introduced by a Situation–Complication–Question setup. AI drafts and stress-tests the structure; you own the answer and whether the logic actually holds.
Stays human
The governing thought — the single answer you are willing to defend to the client — is yours. AI can arrange arguments beautifully around a thesis that is wrong. Deciding what you are recommending, and standing behind it, cannot be delegated.
Watch-outs
- AI produces fluent structure that masks a hollow argument — confirm each group genuinely supports the thesis.
- “MECE-looking” is not MECE; check for real overlaps and gaps.
- A model will invent a confident SCQA even when the underlying answer is unsettled — don't let polish outrun evidence.
Responsible prompt note
Storylining prompts often carry the substance of the engagement — client name, findings, numbers. Much of that is Amber. Abstract the specifics, or keep the prompt to structure only and insert the real content yourself.
Full method — 5 stages, sub-steps, AI fit, consultant actions & prompts added resource›
P.1Lead with the answer
Sub-steps: state the single governing thought · make it a recommendation, not a topic · ensure it answers the reader's actual question · sense-check you'd defend it.
AI fit
- Thesis drafting: turn a pile of findings into 3 candidate governing thoughts
- Sharpening: rewrite a topic-shaped heading as an answer-shaped claim
- Reader test: check the thesis answers the question that was asked
Consultant action
Decide the one thing you are recommending and write it as a sentence with a verb and a consequence. If you cannot, you are not ready to write — go back to the analysis.
Responsible prompt — Governing thought
"Here are my key findings [abstracted findings] and the decision the reader must make [decision]. Propose 3 candidate governing thoughts — each a single declarative recommendation (verb + so-what), not a topic. For each, note what it commits us to and what evidence it would need. Then tell me which most directly answers the reader's question and where each is weakest. Do not pad; if the findings don't support a clear answer yet, say so."
P.2Set up with SCQA
Sub-steps: Situation (agreed context) · Complication (what changed / the tension) · Question (the one it raises) · Answer (your governing thought).
AI fit
- Intro drafting: write a tight Situation–Complication–Question opener
- Tension test: check the complication genuinely provokes the question
- Variants: offer alternative framings for different audiences
Consultant action
Choose a Situation the reader will nod at, a Complication they feel, and let the Question land so your Answer is the obvious next word. Tune the emphasis to the room.
Responsible prompt — SCQA
"Draft a Situation-Complication-Question introduction that leads to this answer: [governing thought], for this audience [audience]. The Situation must be something the reader already accepts; the Complication must create real tension; the Question must be the one the Complication forces. Keep it to a short paragraph. Give me 2 versions — one that emphasises risk, one that emphasises opportunity — and note which suits a sceptical board."
P.3Group arguments MECE
Sub-steps: identify the 3–4 arguments that prove the thesis · ensure they are mutually exclusive · ensure collectively exhaustive · make each a complete claim.
AI fit
- Grouping: cluster supporting points into candidate argument lines
- MECE check: flag overlaps between groups and gaps in coverage
- Headline writing: phrase each group as an assertion, not a label
Consultant action
Reduce to the few arguments that, if true, force the thesis. Test the set: do they overlap? Is anything missing that a smart reader would raise? Each becomes a section headline that asserts, never just names.
Responsible prompt — MECE grouping
"My governing thought is [thesis]. Here are the supporting points I have [list]. Organise them into 3-4 argument groups that are mutually exclusive and collectively exhaustive in supporting the thesis. For each group: (1) write a one-line assertion (a claim with a so-what, not a topic label), (2) list the evidence beneath it, (3) rate how strongly it supports the thesis. Then explicitly flag: any overlap between groups, any gap a critical reader would attack, and any group that is actually a restatement of the thesis rather than support for it."
P.4Order the logic
Sub-steps: choose deductive (chain) or inductive (parallel reasons) · order by importance or by argument flow · ensure each level answers the “why?” of the one above · keep one logic type per group.
AI fit
- Structure check: identify whether a group is deductive or inductive
- Sequencing: propose the most persuasive order of arguments
- Consistency: flag where logic types are mixed within a group
Consultant action
Decide whether the story marches as a deductive chain or stacks as parallel reasons, and order for the reader you have. Verify every supporting point answers the question its parent raises.
Responsible prompt — Order the logic
"Here is my pyramid: thesis [thesis], with these argument groups and their support [paste structure]. For each group, tell me whether it reads as deductive (a logical chain) or inductive (a set of parallel reasons of the same kind), and whether that is the right choice. Recommend the order of the groups for maximum persuasion with [audience], explaining the reasoning. Flag any point that does not actually answer the 'why?' raised by the line above it, and any group mixing two logic types."
P.5Pressure-test & write
Sub-steps: red-team the thesis · check each assertion is evidenced · draft headlines that tell the story alone · confirm the deck reads top-down.
AI fit
- Red-team: attack the thesis as a hostile reader would
- Headline storyline: check the headlines alone convey the argument
- Evidence audit: flag assertions with no support beneath them
Consultant action
Have the structure challenged, then write. Read the section headlines in sequence — they should tell the whole story without the body. Fix any assertion you cannot evidence, and own the final narrative before it reaches the client.
Responsible prompt — Pressure-test
"Act as a sceptical executive reading only the headlines of this storyline [paste thesis + group assertions in order]. Tell me: (1) does the headline sequence alone convey a complete, convincing argument? Where does it break? (2) what is the strongest objection to the thesis, and does the structure answer it? (3) which assertions have no visible evidence beneath them? (4) what would you, as the decision-maker, still need before saying yes? Be hard on it — I would rather hear the objection now than in the room."
Systems Thinking
Use when a problem keeps coming back after you “fix” it — the cause is structure and feedback, not any single part, and you need to see the whole.
Learning objective
Draw the system — boundary, elements, interconnections, feedback loops — find its archetypes and leverage points, and design interventions that shift structure rather than symptoms. AI maps and simulates; you choose the boundary and the bet.
Stays human
Where you draw the boundary decides what solutions become visible — a values-laden choice no model should make for you. So is the call on which leverage point is worth the political cost of pulling.
Watch-outs
- AI will happily draw loops that look causal but are unevidenced — validate with people in the system.
- High-leverage points (paradigms, goals) are exactly the ones AI under-weights; don't stop at parameters.
- Second-order effects are where interventions fail — make AI surface them, then judge them yourself.
Responsible prompt note
System maps often encode org politics, named individuals and confidential structures — frequently Amber. Describe roles and flows generically; keep real names and sensitive dynamics out of the prompt.
Full training script — 6 stages, sub-steps, AI fit, consultant actions & prompts preserved from source›
10.1Define the system
Sub-steps: identify system boundaries · map stakeholders · define purpose · set scope.
AI fit
- Boundary analysis: help define appropriate system boundaries
- Stakeholder identification: identify all system stakeholders
Consultant action
Facilitate workshops to define what's inside and outside the boundary — recognising boundary choices determine what patterns and solutions become visible — and clarify the system's purpose from multiple perspectives.
Responsible prompt — Define the system
"Define the system for addressing [problem situation]. Specify: (1) The system boundary - what's included inside the system vs. what's in the external environment, (2) Justification for why this boundary is appropriate (not too narrow, not too broad), (3) All stakeholders: those who affect the system, those affected by the system, and those who make decisions about the system, (4) The system's purpose from at least 3 different stakeholder perspectives (these may conflict), (5) Key external forces that influence the system but are outside our control, (6) What happens if we expand or contract the boundary? Provide a visual representation showing the boundary, stakeholders, and external forces."
10.2Identify elements
Sub-steps: list system components · map actors · identify resources · document constraints.
AI fit
- Component extraction: identify elements from system documentation
- Resource mapping: catalogue system resources
Consultant action
Create a comprehensive inventory of elements — physical, people, information flows, financial resources, institutional structures — documenting the state of each and how it changes over time.
Responsible prompt — Identify elements
"Identify all elements in this system: [describe system]. Categorize elements as: (1) Physical elements (facilities, equipment, materials), (2) Human elements (roles, teams, skills), (3) Information elements (data, knowledge, communication channels), (4) Financial elements (revenue streams, costs, assets), (5) Institutional elements (rules, policies, culture). For each element, specify: (a) Current state, (b) Key properties and attributes, (c) How it can change, (d) Its role in the system, (e) Interactions with other elements. Create a comprehensive list of 30-50 elements. Then identify which elements are stocks (accumulations) vs. flows (rates of change)."
10.3Map interconnections
Sub-steps: identify relationships · map flows (information, material, energy) · document feedback loops · create causal loop diagrams.
AI fit
- Relationship mapping: extract relationships from data
- Flow analysis: map information and material flows
- Feedback loop detection: identify reinforcing and balancing loops
- Diagram generation: create causal loop diagrams
Consultant action
Map how elements connect and influence each other — feedback loops (reinforcing and balancing), delays, non-linear relationships — building causal loop diagrams that reveal structure and explain behaviour.
Responsible prompt — Map interconnections
"Map the interconnections in this system [describe elements]. Create causal loop diagrams showing: (1) All significant cause-effect relationships between elements, (2) Feedback loops - identify at least 5 reinforcing loops (R1, R2, etc. that amplify change) and 5 balancing loops (B1, B2, etc. that stabilize or resist change), (3) Delays in the system (where cause and effect are separated in time), (4) Non-linear relationships (where small changes can have large effects or vice versa), (5) Information flows, material flows, and decision points. For each feedback loop, explain: (a) The loop's purpose or effect, (b) Whether it's desirable or problematic, (c) Where leverage points might exist. Use standard CLD notation (+ for same direction, - for opposite direction). Create both a detailed map and a simplified version showing only the most important loops."
10.4Identify archetypes
Sub-steps: recognise system patterns · map to system archetypes · understand dynamics · predict behaviour.
AI fit
- Pattern recognition: match situations to system archetypes
- Behaviour prediction: simulate system behaviour over time
Consultant action
Analyse the causal loop diagrams to identify archetypes — Fixes that Fail, Shifting the Burden, Tragedy of the Commons, Escalation, Success to the Successful, Limits to Growth — and explain how each creates the observed problem.
Responsible prompt — Identify archetypes
"Identify system archetypes present in this causal loop diagram [describe or paste CLD]. Look for these common archetypes: (1) Fixes that Fail - quick fixes that worsen the problem long-term, (2) Shifting the Burden - treating symptoms instead of root causes, (3) Tragedy of the Commons - shared resource depletion, (4) Escalation - competitive one-upmanship, (5) Success to the Successful - advantage accumulates to winners, (6) Limits to Growth - growth hits constraints, (7) Accidental Adversaries - potential allies undermine each other, (8) Growth and Underinvestment - failure to invest limits growth. For each archetype identified: (a) Explain how it manifests in this specific system, (b) Show which feedback loops create the archetype, (c) Describe the characteristic behavior pattern over time, (d) Predict what will happen if nothing changes, (e) Identify the leverage points to break the archetype's hold. Provide specific examples from the system."
10.5Find leverage points
Sub-steps: identify intervention points · assess leverage potential · evaluate feasibility · prioritise actions.
AI fit
- Leverage analysis: identify high-leverage intervention points
- Impact modelling: predict effects of interventions
Consultant action
Apply Donella Meadows' framework from low-leverage (parameters) to high-leverage (paradigms), prioritising interventions that address structure rather than symptoms — the highest-leverage points that are also feasible to influence.
Responsible prompt — Find leverage points
"Identify leverage points for intervening in this system [describe system and archetypes]. Apply Meadows' 12 leverage points framework, from lowest to highest leverage: (12) Numbers/parameters, (11) Buffers, (10) Stock-and-flow structures, (9) Delays, (8) Balancing feedback loops, (7) Reinforcing feedback loops, (6) Information flows, (5) Rules of the system, (4) Power to add/change system structure, (3) Goals of the system, (2) Mindset/paradigm, (1) Power to transcend paradigms. For each leverage point level, identify: (a) Specific intervention opportunities in this system, (b) Expected impact (high/medium/low), (c) Feasibility (easy/moderate/difficult), (d) Time to see results, (e) Risks or unintended consequences. Prioritize the top 5 leverage points to act on, focusing on high leverage + feasible interventions. Explain why these offer the best return on effort."
10.6Design interventions
Sub-steps: develop intervention strategies · test in simulation · plan implementation · monitor effects.
AI fit
- Intervention design: generate intervention options
- Simulation testing: test interventions in digital twins
- Unintended-consequence prediction: forecast second-order effects
Consultant action
Design interventions for the chosen leverage points, weighing intended effects and unintended consequences, planning for system resistance and delays, and establishing monitoring of both first- and second-order effects.
Responsible prompt — Design interventions
"Design interventions for this system targeting these leverage points [list leverage points]. For each intervention, specify: (1) The specific action to take, (2) The leverage point it targets and why, (3) The intended effects (first-order outcomes), (4) Potential unintended consequences (second and third-order effects), (5) How to test this intervention safely (pilot, simulation, or phased rollout), (6) Implementation timeline with milestones, (7) Resources required, (8) Potential resistance points and how to address them, (9) Metrics to monitor (both leading and lagging indicators), (10) Criteria for success, scaling, or stopping. Create an implementation roadmap showing the sequence of interventions, explaining why this order (considering dependencies and system readiness). Include a monitoring plan that tracks both intended and unintended effects over time."
Theory of Constraints
Use when throughput is capped and you need to lift it — one bottleneck governs the whole system, and local efficiency drives everywhere else are wasting effort.
Learning objective
Run the five focusing steps — identify, exploit, subordinate, elevate, repeat — so the system's true constraint governs every decision. AI finds bottlenecks and models elevation ROI; you make the subordination and investment calls.
Stays human
Subordination asks teams to be deliberately “less efficient” for the good of the whole — a change-management and political act AI cannot lead. Deciding when to invest capital to elevate is a judgement on risk and timing.
Watch-outs
- AI may flag a local bottleneck, not the system constraint — validate it governs total throughput.
- Elevating before exploiting wastes money; hold the sequence even if a tool suggests a quick capital fix.
- After elevation the constraint moves — a model won't feel the organisational inertia that keeps old policies alive.
Responsible prompt note
Throughput, cost and capacity data are usually Amber and commercially sensitive. Use ratios and relative figures, or abstract the process, rather than pasting raw operational and financial data.
Full training script — 5 stages, sub-steps, AI fit, consultant actions & prompts preserved from source›
11.1Identify the constraint
Sub-steps: map the system · measure throughput · find the bottleneck · validate the constraint.
AI fit
- Bottleneck detection: analyse process data to identify constraints
- Throughput analysis: measure and analyse flow metrics
- Constraint validation: statistically validate the true constraint
Consultant action
Map the full value stream and measure throughput at each stage to find where work accumulates and flow is restricted, validating that it is the system constraint, not a local bottleneck.
Responsible prompt — Identify the constraint
"Identify the constraint in this system: [describe process with throughput data for each step]. Analyze: (1) Map the entire process flow from start to finish with all steps, (2) Measure throughput (units/time) at each step, (3) Identify where inventory/work accumulates (sign of bottleneck), (4) Calculate utilization rates for each resource, (5) Determine which constraint limits the entire system's throughput (not just local optimization). Apply these tests to validate the constraint: (a) If you elevate this constraint, does system throughput increase?, (b) Is this constraint active most of the time (not occasionally)?, (c) Does work pile up before this point? Provide data-driven evidence for the constraint identification. Also identify if this is a physical constraint (capacity), policy constraint (rules), or market constraint (demand)."
11.2Exploit the constraint
Sub-steps: maximise constraint utilisation · eliminate waste at constraint · ensure constraint never idle · optimise constraint output.
AI fit
- Utilisation optimisation: maximise constraint throughput
- Schedule optimisation: create optimal constraint schedules
- Waste detection: identify non-value-add activities at the constraint
Consultant action
Ensure the constraint is never idle — buffers before it, no non-value-added work on it — and optimise its schedule for the most valuable output, accepting that non-constraints run below 100%.
Responsible prompt — Exploit the constraint
"Design exploitation strategies for this constraint: [describe constraint]. Create an action plan to: (1) Ensure the constraint is never idle - what buffer strategy (time buffer, inventory buffer, or capacity buffer) and how large?, (2) Eliminate all non-value-added activities at the constraint - list current activities and which to stop, delegate, or automate, (3) Optimize the constraint's schedule - what prioritization rule (throughput per constraint minute, due date, etc.)?, (4) Quality at the constraint - how to prevent defects from consuming constraint capacity?, (5) Reduce setup/changeover times at the constraint - specific techniques?, (6) Cross-train resources to maximize constraint flexibility. Calculate the expected throughput increase from full exploitation. Create a detailed implementation checklist with owners and timeline."
11.3Subordinate everything else
Sub-steps: align non-constraints to constraint · adjust policies · modify behaviours · synchronise flow.
AI fit
- Synchronisation planning: align all processes to constraint pace
- Policy analysis: identify policies that conflict with constraint optimisation
- Change-impact analysis: predict effects of subordination
Consultant action
Align all non-constraint resources to the constraint's pace, changing policies, metrics and behaviours that reward local optimisation at the expense of system throughput.
Responsible prompt — Subordinate everything else
"Design subordination strategies for this system with constraint at [constraint location]. Specify: (1) How non-constraint resources should be scheduled (to what pace and priority?), (2) What utilization rates are acceptable for non-constraints (typically 70-90%, not 100%), (3) Which policies need to change (e.g., efficiency metrics, batch size rules, local optimization incentives), (4) How to communicate to teams that their utilization will decrease but system performance will improve, (5) What new metrics to implement (throughput, on-time delivery, inventory turns vs. individual efficiency), (6) How to synchronize flow from release through constraint to delivery, (7) Buffer management - where to place buffers and how to monitor them. Create a change management plan addressing resistance from teams asked to be 'less efficient.' Provide specific policy changes with before/after examples."
11.4Elevate the constraint
Sub-steps: invest in constraint capacity · add resources · upgrade technology · increase capability.
AI fit
- Investment analysis: evaluate elevation options and ROI
- Capacity planning: determine optimal capacity additions
- Technology assessment: identify technologies that can elevate the constraint
Consultant action
Evaluate elevation through investment, resources, technology or capability — calculating ROI on increased throughput — and only invest after fully exploiting and subordinating.
Responsible prompt — Elevate the constraint
"Evaluate elevation options for this constraint: [describe constraint and current capacity]. Analyze these options: (1) Add capacity (more equipment/people) - cost, lead time, capacity increase, ROI, (2) Upgrade technology - cost, implementation time, throughput improvement, risks, (3) Outsource part of constraint work - cost per unit, quality implications, capacity freed, (4) Improve capability (training, process improvement) - investment, time to benefit, sustainability, (5) Redesign product/process to reduce constraint demand - feasibility, investment, impact. For each option: (a) Total investment required, (b) Expected throughput increase, (c) Time to implement, (d) ROI calculation (throughput gain × margin / investment), (e) Risks and mitigation, (f) Reversibility. Recommend the optimal elevation strategy (may be combination) with implementation plan. Calculate the financial impact of elevating vs. not elevating."
11.5Prevent inertia
Sub-steps: monitor for new constraints · repeat the process · avoid complacency · continuous improvement.
AI fit
- Constraint monitoring: detect when the constraint shifts
- Continuous improvement: identify ongoing optimisation opportunities
- Alert systems: notify when new constraints emerge
Consultant action
Build monitoring to detect when the constraint shifts (it will after elevation) and stop old policies persisting. Institutionalise the five focusing steps as an ongoing process, not a one-off project.
Responsible prompt — Prevent inertia
"Create a constraint monitoring and continuous improvement system. Specify: (1) Metrics to monitor daily/weekly (throughput, constraint utilization, buffer status, inventory levels), (2) Trigger points that signal the constraint has shifted (e.g., constraint utilization drops below X%, inventory accumulates elsewhere), (3) Who is responsible for monitoring and calling constraint shifts, (4) Review cadence (daily constraint meetings, weekly system reviews, monthly strategy updates), (5) Process for re-applying the 5 focusing steps when constraint shifts, (6) How to prevent inertia - which old policies/metrics must be permanently changed, (7) Training plan to build internal TOC capability, (8) Communication strategy to keep organization aligned as constraint moves. Create a dashboard showing key TOC metrics and alert thresholds. Develop a 90-day roadmap for institutionalizing TOC as the operating system, not just a project."
Appreciative Inquiry
Use when change has stalled and deficit-focus has drained energy — you want transformation built on what already works, with the organisation's own strengths as fuel.
Learning objective
Run the 4-D cycle — Discover, Dream, Design, Destiny — surfacing the positive core and building forward from it. AI gathers and themes stories and drafts propositions; you facilitate the human energy that makes change self-sustaining.
Stays human
Appreciative inquiry works because people feel heard and own the dream. That belonging cannot be generated by a model — the facilitation, trust and collective authorship are irreducibly human.
Watch-outs
- AI-written “provocative propositions” can sound aspirational but hollow — they must be owned by the room, not delivered to it.
- Strength-focus is not denial; don't let positivity suppress real issues that need naming.
- Story analysis at scale can flatten the very specifics that give stories their power.
Responsible prompt note
Interview stories contain personal, identifiable accounts — often Amber. Strip names and identifying detail before theming at scale, and tell participants if their words will pass through an AI tool.
Full training script — 4 stages, sub-steps, AI fit, consultant actions & prompts preserved from source›
12.1Discover — what gives life?
Sub-steps: identify positive core · collect success stories · map strengths · appreciate what works.
AI fit
- Story collection: gather and analyse positive-deviance stories
- Strength identification: extract strengths from interview data
- Pattern recognition: identify what's working best
- Sentiment analysis: find positive patterns in organisational data
Consultant action
Conduct appreciative interviews about peak experiences and high points, analysing the stories to identify the positive core — the unique strengths and life-giving forces behind exceptional performance.
Responsible prompt — Discover
"Design an appreciative inquiry discovery process for [organization]. Create: (1) Interview protocol with questions like: 'Describe a time when this organization was at its best,' 'What gives this organization life and energy?', 'What are our unique strengths that we should never lose?', (2) Sampling strategy to capture diverse perspectives across levels, functions, and tenure, (3) Method to collect and document stories (interviews, focus groups, written submissions), (4) Analysis framework to identify themes, patterns, and the positive core, (5) Process to validate findings with participants. Analyze these [X] stories [paste or describe] to identify: (a) Recurring strengths and capabilities, (b) Conditions that enable peak performance, (c) Values in action, (d) Relationships and processes that work exceptionally well. Synthesize into a 'positive core' statement capturing what gives this organization life."
12.2Dream — what might be?
Sub-steps: envision ideal future · create provocative propositions · imagine possibilities · dream boldly.
AI fit
- Vision generation: create compelling future visions
- Provocative-proposition crafting: write inspiring possibility statements
- Scenario creation: depict ideal future states
Consultant action
Facilitate participants in envisioning the ideal future, grounded in the positive core but stretching beyond current constraints, crafting provocative propositions that are both grounded and aspirational.
Responsible prompt — Dream
"Facilitate the dream phase for [organization] with this positive core: [describe strengths]. Design activities to: (1) Help participants envision the organization 3-5 years in the future at its absolute best, (2) Ground the vision in the positive core while stretching beyond current limitations, (3) Create provocative propositions in the format: 'We are [bold aspiration] by [how we leverage strengths],' (4) Use creative methods (visualization, storytelling, art, theater) to access aspirational thinking, (5) Ensure the dream is collectively owned, not imposed. Generate 5-7 provocative propositions that are: (a) Grounded in reality (based on positive core), (b) Aspirational (stretching current reality), (c) Affirmative (stated as already achieved), (d) Bold (not incremental), (e) Action-oriented (implying what to do). Test each proposition: Does it energize? Is it memorable? Does it point toward action?"
12.3Design — how can it be?
Sub-steps: co-construct ideal organisation · develop design principles · create structures and processes · align systems.
AI fit
- Design collaboration: facilitate co-creation sessions
- Principle development: synthesise design principles from input
- System alignment: check alignment between proposed designs
Consultant action
Facilitate co-creation of the structures, processes, relationships and systems needed to realise the propositions — participatory, strength-based, and aligned toward the dream.
Responsible prompt — Design
"Facilitate the design phase to realize these provocative propositions: [list propositions]. Guide participants to co-create: (1) Organizational structures - how we organize to leverage strengths and achieve the dream, (2) Processes and practices - how work gets done, decisions are made, and learning occurs, (3) Relationships and partnerships - how we relate internally and externally, (4) Systems and infrastructure - technology, physical space, resources needed, (5) Culture and values - how we embody our aspirations daily. For each design element, ensure: (a) It builds on the positive core, (b) It directly supports one or more provocative propositions, (c) It's co-created by those who will implement it, (d) It's specific enough to act on. Create design principles that guide all decisions. Develop prototypes or pilots for key designs. Check alignment: Do all elements reinforce each other toward the dream?"
12.4Destiny — what will be?
Sub-steps: implement designs · build momentum · sustain change · continuous learning.
AI fit
- Implementation planning: create action plans
- Momentum tracking: monitor adoption and energy
- Learning systems: capture and share learnings
Consultant action
Support implementation through voluntary, momentum-building action rather than mandated programmes, with structures for continuous learning and celebration so the inquiry becomes self-sustaining.
Responsible prompt — Destiny
"Design the destiny phase for sustained transformation. Create: (1) Implementation approach - how to launch voluntary, momentum-building initiatives (not top-down mandates), (2) Action learning sets - small groups experimenting with designs and sharing learnings, (3) Momentum indicators - how to track energy, engagement, and adoption (not just metrics), (4) Celebration and recognition - how to amplify successes and positive deviance, (5) Learning infrastructure - how to capture, share, and scale what works, (6) Ongoing AI cycles - how to repeat discovery-dream- design-destiny continuously, (7) Leadership development - how to build internal AI facilitation capability, (8) Sustainability plan - how to maintain momentum beyond initial enthusiasm. Create a 90-day action plan with: (a) Quick wins to build confidence, (b) Pilot projects to test designs, (c) Communication strategy, (d) Resource allocation, (e) Support structures. Design feedback loops that keep the organization learning and evolving."
Complex Adaptive Systems Management
Use when outcomes emerge from many interacting agents and can't be commanded — culture, innovation, networks — and you must shape conditions rather than dictate results.
Learning objective
Manage by shaping the system: understand its dynamics, set enabling constraints, foster connectivity, enable emergence, then learn and adapt. AI maps networks and detects emerging patterns; you decide which patterns to amplify and which to dampen.
Stays human
You cannot control emergence — and neither can AI. The wisdom to set minimal rules, tolerate ambiguity, and resist the urge to over-manage is a leadership stance, not an output.
Watch-outs
- AI pattern-detection can mistake noise for emergence — confirm a pattern is real before amplifying it.
- “Optimising” a complex system often kills the diversity it needs to adapt; beware tidy recommendations.
- Network maps can expose individuals; handle relationship data with care.
Responsible prompt note
Interaction and network data is personal data — frequently Amber or Red under GDPR. Aggregate and anonymise; never feed identifiable relationship maps of named employees into an external tool.
Full training script — 5 stages, sub-steps, AI fit, consultant actions & prompts preserved from source›
13.1Understand system dynamics
Sub-steps: map agents and interactions · identify emergence patterns · assess connectivity · understand adaptation mechanisms.
AI fit
- Agent mapping: identify all agents in the system
- Network analysis: map interaction networks
- Emergence detection: identify emergent patterns
Consultant action
Map the agents, their interactions and the patterns that emerge, assessing connectivity, diversity and adaptation mechanisms to understand how the system self-organises and evolves.
Responsible prompt — Understand dynamics
"Analyze this organization as a complex adaptive system: [describe organization]. Map: (1) Agents - identify all types of agents (individuals, teams, departments, external partners) and their characteristics, (2) Interactions - how agents connect, communicate, and influence each other (formal and informal networks), (3) Emergent patterns - what behaviors, cultures, or outcomes emerge that aren't designed or mandated, (4) Connectivity - density and quality of connections (too sparse = fragmented, too dense = chaotic), (5) Diversity - variety of perspectives, skills, approaches (essential for adaptation), (6) Adaptation mechanisms - how agents learn and adjust behavior based on feedback, (7) Attractors - what patterns or states the system naturally moves toward. Create a network map showing agents and interactions. Identify 3-5 emergent patterns and explain how they arise from local interactions. Assess whether the system has optimal complexity (enough diversity and connectivity to adapt without becoming chaotic)."
13.2Create enabling constraints
Sub-steps: set simple rules · define boundaries · establish feedback mechanisms · allow autonomy.
AI fit
- Rule design: suggest minimal effective rules
- Boundary optimisation: determine optimal constraint levels
- Feedback-system design: design effective feedback loops
Consultant action
Design enabling constraints — minimal rules and boundaries that give enough structure for coordination while allowing maximum autonomy and emergence — with rapid feedback so agents self-correct.
Responsible prompt — Enabling constraints
"Design enabling constraints for this CAS: [describe system and desired outcomes]. Create: (1) Simple rules - 3-7 minimal rules that guide behavior without prescribing actions (e.g., 'Share information openly,' 'Test before scaling,' 'Prioritize customer value'), (2) Boundaries - clear limits within which agents have full autonomy (what's non-negotiable vs. what's flexible), (3) Feedback mechanisms - rapid, transparent feedback loops so agents can adjust behavior (what data, how often, who sees it), (4) Information flow - ensure agents have access to information they need to make local decisions, (5) Resource constraints - how resources are allocated and what agents can access autonomously. For each constraint, ensure it's: (a) Minimal (only what's necessary), (b) Clear (unambiguous), (c) Enabling (creates freedom within boundaries, not restriction), (d) Adaptive (can evolve as system learns). Contrast enabling constraints vs. controlling constraints. Test: Do these constraints allow emergence while preventing chaos?"
13.3Foster connectivity
Sub-steps: enhance information flow · build relationships · create platforms for interaction · encourage diversity.
AI fit
- Network optimisation: suggest connectivity improvements
- Information-flow analysis: identify information bottlenecks
- Diversity measurement: assess and enhance cognitive diversity
Consultant action
Enhance connectivity through better information flow, platforms for interaction and diverse relationships — neither too sparse (no coordination) nor too dense (overload and groupthink).
Responsible prompt — Foster connectivity
"Design connectivity enhancements for this CAS: [describe current network and gaps]. Create strategies to: (1) Enhance information flow - reduce bottlenecks, increase transparency, create feedback channels (specific mechanisms), (2) Build relationships - create opportunities for agents to connect across silos (formal and informal), (3) Create interaction platforms - physical and virtual spaces for emergence (communities of practice, open spaces, digital platforms), (4) Encourage diversity - ensure cognitive, functional, and demographic diversity in interactions (how to attract and include diverse voices), (5) Optimize network density - identify where to add connections (isolated agents) and where to reduce (overloaded hubs), (6) Weak ties - strengthen bridging connections between clusters (essential for innovation). Analyze the current network: Where are the clusters? Who are the hubs? Where are the structural holes? Recommend specific interventions to improve connectivity while maintaining healthy diversity and preventing echo chambers. Create a network health dashboard with metrics."
13.4Enable emergence
Sub-steps: create safe-to-fail experiments · allow self-organisation · detect emerging patterns · amplify positive emergence.
AI fit
- Experiment design: design safe-to-fail probes
- Pattern detection: identify emerging patterns in real time
- Amplification strategies: suggest how to scale positive emergence
Consultant action
Create conditions for positive emergence — safe-to-fail experiments, room to self-organise, sensors to detect patterns — amplifying what aligns with desired outcomes and dampening what doesn't, without trying to control emergence directly.
Responsible prompt — Enable emergence
"Design interventions to enable positive emergence in this CAS: [describe system and desired direction]. Create: (1) Safe-to-fail experiments - 5-10 probes to test new approaches (each must be: low-cost, reversible, diverse, and provide learning regardless of outcome), (2) Self-organization spaces - where agents can form initiatives without permission (what's allowed, what resources available), (3) Pattern detection - how to sense what's emerging (observation methods, metrics, feedback channels, who's responsible), (4) Amplification strategies - how to scale positive emergence that aligns with goals (provide resources, remove barriers, connect to others, tell stories), (5) Dampening strategies - how to reduce negative emergence (not by force, but by changing conditions that enable it), (6) Learning loops - how to capture and share what emerges. For each experiment, specify: hypothesis, actions, resources needed, success indicators, learning questions, how to scale if successful. Create a portfolio of experiments (some incremental, some transformational). Design sensors to detect emergence early."
13.5Learn and adapt
Sub-steps: monitor system behaviour · capture learnings · adjust constraints · evolve approach.
AI fit
- Real-time monitoring: track system dynamics continuously
- Learning extraction: capture insights from system behaviour
- Adaptive recommendations: suggest constraint adjustments
Consultant action
Establish continuous learning that monitors behaviour, captures insight from experiments and emergence, and adapts constraints and approach — with feedback loops at individual, team and organisational levels.
Responsible prompt — Learn and adapt
"Create a learning and adaptation system for this CAS: [describe system]. Design: (1) Monitoring framework - what to track at multiple levels (agent behavior, interaction patterns, emergent outcomes, system health indicators), (2) Learning capture - how to document insights from experiments, successes, failures, and unexpected outcomes (templates, processes, ownership), (3) Sense-making forums - regular gatherings to interpret data, share learnings, and adjust understanding (cadence, participants, format), (4) Adaptation mechanisms - how learnings translate into changes in constraints, strategies, or approaches (decision rights, process), (5) Feedback loops - rapid feedback to agents on experiment results and system status (transparency, accessibility), (6) Meta-learning - how the CAS management approach itself evolves based on what works (reflection, adjustment), (7) Knowledge sharing - how learnings spread across the system (platforms, communities, stories). Create a learning cycle diagram showing how observation, interpretation, action, and adaptation connect. Specify metrics for system learning capacity (not just outcomes). Design quarterly CAS health assessments."
Stakeholder Mapping & Change Adoption
Use when the analysis is right but the change still has to land — you must move real people from awareness to adoption against inertia, politics and fatigue.
Learning objective
Map stakeholders by power and interest, then drive adoption through a recognised model — ADKAR at the individual level, Kotter at the organisational. AI drafts the map, messages and plan; you read the room and carry the trust.
Stays human
Influence runs on relationship and credibility. AI can draft a stakeholder grid, but reading unspoken resistance, holding a difficult conversation, and being trusted to lead change are human to the core.
Watch-outs
- A stakeholder grid is a hypothesis about people — never share the raw “low power / against us” labels; they damage trust if leaked.
- AI-written change comms can read as corporate and hollow; make them human before they go out.
- Adoption is emotional, not logical — don't let a tidy plan convince you resistance is irrational.
Responsible prompt note
Stakeholder assessments are sensitive personal commentary — treat as Amber/Red. Use roles not names, keep influence/attitude judgements out of any external tool, and store the real map securely offline.
Full method — 5 stages, sub-steps, AI fit, consultant actions & prompts added resource›
C.1Identify & map stakeholders
Sub-steps: list everyone affected or influential · assess power and interest · place on the grid (manage closely / keep satisfied / keep informed / monitor) · note current attitude.
AI fit
- Stakeholder discovery: generate a complete list by role and function
- Grid drafting: propose power/interest placements to challenge
- Blind-spot scan: surface affected groups you've overlooked
Consultant action
Build the power–interest grid with people who know the politics, capturing each stakeholder's influence, interest and current stance. Hold this map confidentially — it is a working tool, not a deliverable.
Responsible prompt — Map stakeholders
"For a change initiative described as [abstracted change], help me build a stakeholder map. Using ROLES not names: (1) list the stakeholder groups likely affected by or influential over this change, (2) for each, suggest typical level of power (high/low) and interest (high/low) and the resulting grid quadrant (manage closely / keep satisfied / keep informed / monitor), (3) note what each group most likely cares about and fears, (4) flag any affected group I may have overlooked. Present as a table I can refine with people who know the actual politics."
C.2Diagnose readiness & resistance
Sub-steps: assess change readiness · locate sources of resistance · understand the “what's in it for me” per group · find the real blockers.
AI fit
- Resistance patterns: categorise likely objections by group
- WIIFM drafting: articulate the benefit each group will actually feel
- Survey design: draft readiness-assessment questions
Consultant action
Get beneath stated objections to the real ones — loss of status, workload, history of failed change. Listen for what isn't said; that is usually where adoption succeeds or dies.
Responsible prompt — Diagnose resistance
"For this change [abstracted change] and these stakeholder groups [roles], help me anticipate resistance. For each group: (1) the most likely objections, separated into stated reasons vs. probable underlying reasons (status, workload, fear, past failed change), (2) a clear 'what's in it for me' that this group would genuinely value, (3) what would have to be true for them to actively support it. Then draft 8-10 neutral questions for a change-readiness pulse survey. Remind me which of these judgements I must validate with real conversations rather than assume."
C.3Build the change story (Kotter)
Sub-steps: establish urgency · form a guiding coalition · craft a clear vision · communicate it relentlessly.
AI fit
- Vision drafting: turn the case for change into a memorable narrative
- Message tailoring: adapt the story per stakeholder group
- Channel planning: propose a multi-channel communication cadence
Consultant action
Assemble a credible guiding coalition and tell a true urgency story — not manufactured fear. Choose who delivers each message; the messenger often matters more than the words.
Responsible prompt — Change story
"Help me build the change narrative for [abstracted change] using Kotter's early steps. Produce: (1) an honest urgency case (why now, what happens if we don't — grounded, not scaremongering), (2) the profile of a guiding coalition (which roles and why their sponsorship matters), (3) a one-paragraph change vision that is clear, memorable and free of jargon, (4) tailored key messages for each stakeholder group, (5) a communication cadence across channels for the first 90 days. Keep the tone human and specific; flag anything that sounds like corporate boilerplate so I can rewrite it in my own voice."
C.4Drive individual adoption (ADKAR)
Sub-steps: Awareness · Desire · Knowledge · Ability · Reinforcement — diagnose where each group is stuck and act on that barrier.
AI fit
- Barrier diagnosis: infer which ADKAR step a group is stuck at
- Enablement design: draft training, job aids and reinforcement ideas
- Gap check: flag where we've skipped Desire and jumped to Knowledge
Consultant action
Find the actual ADKAR barrier — most failed change pushes Knowledge at people who lack Desire. Match the intervention to the barrier, and own the reinforcement so the change sticks past go-live.
Responsible prompt — ADKAR
"Using the ADKAR model (Awareness, Desire, Knowledge, Ability, Reinforcement), help me plan adoption for [abstracted change] across these groups [roles]. For each group: (1) estimate which ADKAR element is the likely barrier and why, (2) recommend specific interventions for that element (don't default to training), (3) flag where we risk delivering Knowledge before securing Desire. Then propose reinforcement mechanisms that keep the change in place 3-6 months after launch. Note which estimates I should confirm through direct conversation."
C.5Sustain & embed
Sub-steps: secure short-term wins · remove barriers · anchor in process and culture · measure adoption, not just deployment.
AI fit
- Win identification: suggest visible early wins to publicise
- Metric design: define adoption (not deployment) measures
- Embedding check: list processes/policies that must change to lock it in
Consultant action
Measure whether people actually changed behaviour, not whether the system went live. Remove the structural barriers that quietly pull people back, and anchor the change in how the organisation runs day to day.
Responsible prompt — Sustain & embed
"For [abstracted change], help me sustain adoption. Produce: (1) 3-5 visible short-term wins we could engineer and publicise in the first 90 days, (2) adoption metrics that measure changed behaviour rather than deployment (e.g. usage depth, not licences issued), (3) the processes, policies, incentives or rituals that must change to anchor this in 'how we work', (4) the structural barriers most likely to pull people back to old behaviour and how to remove them, (5) a simple dashboard of leading and lagging adoption indicators. Recommend a review cadence and who owns it after we exit."
Agile / Scrum / Kanban
Use when requirements will change as you learn and value comes from shipping increments — delivery, product build, transformation — rather than a fixed up-front plan.
Learning objective
Run the cadence — backlog, planning, standup, execution, review, retrospective, flow — with AI drafting stories, estimates and analyses while the team owns commitment, collaboration and improvement. Velocity is a measure, not a target.
Stays human
The team's commitment to a sprint goal, and the psychological safety that makes a retrospective honest, are human. AI can suggest a format; it cannot make people trust each other enough to tell the truth.
Watch-outs
- AI estimates from “historical data” it doesn't have — treat story points as the team's, not the model's.
- Over-automating ceremonies hollows them out; the value is the conversation, not the artefact.
- Don't let AI-generated acceptance criteria substitute for talking to the product owner.
Responsible prompt note
Backlogs can contain client requirements, security detail and roadmap — often Amber. Generalise feature descriptions; keep client identifiers and sensitive technical specifics out of the prompt.
Full training script — 7 stages, sub-steps, AI fit, consultant actions & prompts preserved from source›
14.1Product backlog creation
Sub-steps: gather requirements · write user stories · prioritise backlog · estimate effort.
AI fit
- Story generation: write user stories from requirements
- Prioritisation: prioritise by value, risk, dependencies
- Estimation: predict story points from historical data
- Acceptance criteria: generate comprehensive criteria
Consultant action
Facilitate backlog workshops to gather requirements and translate them into user stories with clear acceptance criteria, establishing prioritisation (e.g. WSJF) and facilitating estimation via planning poker or affinity sizing.
Responsible prompt — Backlog creation
"Create a product backlog for [product/initiative]. Generate: (1) 30-50 user stories in the format: 'As a [user type], I want to [action] so that [benefit],' covering all major features and requirements, (2) Acceptance criteria for each story using Given/When/Then format, (3) Story point estimates (using Fibonacci scale) based on complexity, effort, and uncertainty, (4) Prioritization using WSJF (Weighted Shortest Job First) considering: user/business value, time criticality, risk reduction, and job size, (5) Dependencies between stories, (6) Definition of Ready (criteria for when a story can enter sprint) and Definition of Done (criteria for completion). Organize backlog into themes/epics. Identify top 10 highest priority stories for next sprint. Flag stories needing clarification or splitting. Create a backlog refinement roadmap showing when stories need preparation for future sprints."
14.2Sprint planning
Sub-steps: select backlog items · break down tasks · commit to sprint goal · allocate capacity.
AI fit
- Sprint-goal suggestion: propose optimal sprint goals
- Task breakdown: decompose stories into tasks
- Capacity planning: predict team capacity
- Commitment optimisation: suggest realistic commitments
Consultant action
Facilitate planning where the team selects items by priority and capacity, breaks stories into tasks and commits to a sprint goal — using historical velocity and real capacity to keep a sustainable pace.
Responsible prompt — Sprint planning
"Facilitate sprint planning for this team: [describe team capacity, velocity, and top backlog items]. Guide the team to: (1) Define sprint goal - what outcome will achieve (not just list of features), (2) Select stories - which items from backlog fit capacity while achieving sprint goal, (3) Break down tasks - decompose each story into specific tasks (development, testing, documentation, etc.) with ownership, (4) Estimate tasks - hours or ideal days for each task, (5) Check capacity - total task hours vs. available team hours (typically 6 hours/day per person to account for meetings, interruptions), (6) Identify risks - what could prevent achieving sprint goal, (7) Commitment - team's commitment to sprint goal and selected work. Create sprint plan showing: sprint goal, selected stories with tasks and owners, capacity allocation, risks and mitigations, definition of done. Ensure team commits to what's achievable, not optimistic. Flag if velocity is consistently over/under-estimated."
14.3Daily standups
Sub-steps: share progress · identify blockers · coordinate work · adjust plans.
AI fit
- Progress tracking: track progress automatically from tools
- Blocker detection: identify patterns that cause blockers
- Coordination suggestions: suggest who needs to talk to whom
Consultant action
Coach the team on effective 15-minute standups focused on the three questions, driving coordination and problem-solving rather than status reporting, with blockers addressed immediately after.
Responsible prompt — Daily standups
"Optimize daily standup effectiveness for this team: [describe current standup challenges]. Design: (1) Standup format - three questions format or walk-the-board approach, time-boxed to 15 minutes, (2) Participation rules - whole team stands in circle, focus on coordination not status, one person speaks at a time, (3) Blocker management - how blockers are raised, who owns resolution, when follow-up happens (immediately after standup), (4) Visual management - what board or tools show current state, (5) Metrics to track - standup duration, blocker resolution time, team engagement, (6) Anti-patterns to avoid - problem-solving during standup, status reporting to manager, late arrivals, side conversations. Create a standup charter with team agreements. Design a blocker escalation path with time limits (e.g., if not resolved in 24 hours, escalate to Scrum Master, then Product Owner). Suggest improvements based on common issues: running long, lack of focus, recurring blockers, or disengagement."
14.4Sprint execution
Sub-steps: complete work · collaborate · update progress · manage impediments.
AI fit
- Work automation: automate repetitive tasks
- Collaboration facilitation: connect people with needed expertise
- Progress prediction: predict if sprint goals will be met
- Impediment resolution: suggest solutions to common blockers
Consultant action
Support execution by removing impediments, facilitating collaboration and protecting the team from scope creep — monitoring the burndown and enabling mid-sprint adjustments without abandoning the sprint goal.
Responsible prompt — Sprint execution
"Support sprint execution for this team working on [sprint goal]. Provide: (1) Progress tracking - how to update task status, maintain burndown chart, and visualize progress, (2) Collaboration support - when to swarm on difficult stories, how to pair program effectively, when to seek help, (3) Impediment management - process for raising blockers, escalation path, target resolution times, (4) Scope change protocol - how to handle new requests mid-sprint (typically defer to next sprint unless critical), (5) Quality practices - definition of done checklist, code review process, testing requirements, (6) Mid-sprint inspection - when to call team huddle if behind track, how to re-plan without changing goal. Create a sprint execution playbook with: daily rhythms, communication channels, quality gates, escalation procedures. Predict risks to sprint goal based on current progress and recommend proactive actions. Suggest collaboration tools and practices for distributed teams if applicable."
14.5Sprint review
Sub-steps: demonstrate work · gather feedback · adapt backlog · celebrate success.
AI fit
- Demo preparation: help create compelling demos
- Feedback analysis: analyse stakeholder feedback
- Backlog adjustment: suggest backlog changes from feedback
Consultant action
Facilitate reviews where the team demonstrates working software, gathers feedback and discusses what's next — collaborative working sessions, not formal presentations, feeding directly into backlog refinement.
Responsible prompt — Sprint review
"Facilitate an effective sprint review for [product/team]. Design: (1) Demo format - show working software (not slides), focus on sprint goal achievement, keep to 1 hour max, (2) Participation - whole team presents, stakeholders attend and provide feedback, create safe environment for honest discussion, (3) Feedback collection - what did we learn, what works well, what needs adjustment, new opportunities identified, (4) Backlog adaptation - which items to add, remove, or reprioritize based on feedback and learning, (5) Release planning - update release forecast based on velocity and new information, (6) Celebration - acknowledge team achievements and progress. Create sprint review agenda: (1) Sprint goal review, (2) Demo of done items, (3) Discussion of what was not completed, (4) Feedback collection, (5) Backlog refinement, (6) Next steps. Prepare template for capturing feedback and decisions. Coach team on receiving feedback constructively and adapting based on learning, not defending work."
14.6Sprint retrospective
Sub-steps: reflect on process · identify improvements · create action items · commit to changes.
AI fit
- Retrospective facilitation: suggest retrospective formats
- Pattern detection: identify recurring issues
- Improvement suggestions: propose specific improvements
- Action-item tracking: track retrospective commitments
Consultant action
Facilitate retrospectives in varied formats so the team reflects honestly, ensuring outcomes are specific, owned, actionable improvements with follow-up — so the same issues don't recur sprint after sprint.
Responsible prompt — Sprint retrospective
"Design a sprint retrospective for this team: [describe team dynamics and recent challenges]. Create: (1) Retrospective format - choose based on team needs (Start/Stop/Continue for straightforward, Sailboat for vision-focused, 4Ls for emotional check, Timeline for complex sprints), (2) Prime directive reminder - 'Regardless of what we discover, we understand and truly believe that everyone did the best job they could,' (3) Data gathering - collect facts about the sprint (what happened, metrics, events), (4) Generate insights - why did things happen, patterns, root causes, (5) Decide actions - 1-3 specific improvements with owners and deadlines (not vague aspirations), (6) Close retrospective - feedback on the retro itself, appreciation. Design activities for each phase with time boxes (total 1-3 hours depending on sprint length). Create action item tracking system to ensure follow-through. Vary formats to prevent retrospective fatigue. Address common issues: lack of psychological safety, recurring problems, or no action follow-through."
14.7Kanban flow optimisation
Sub-steps: visualise work · limit WIP · manage flow · optimise cycle time.
AI fit
- Flow analysis: analyse workflow patterns
- Bottleneck detection: identify flow constraints
- WIP-limit optimisation: suggest optimal WIP limits
- Cycle-time prediction: predict completion dates
Consultant action
Implement Kanban — visualise work, set WIP limits to prevent overload, manage flow to optimise cycle time — using cumulative flow, lead time and throughput to find bottlenecks and continuously improve.
Responsible prompt — Kanban flow
"Optimize Kanban flow for this team: [describe current workflow and challenges]. Design: (1) Kanban board - columns representing workflow stages (To Do, Analysis, Development, Testing, Done, plus team-specific stages), (2) WIP limits - set limits for each column based on team capacity (typically team size +1 for most columns), (3) Policies - explicit definitions for when work can enter/exit each column, (4) Classes of service - different handling for standard, expedite, fixed date, and intangible work, (5) Metrics to track - cycle time (start to finish), lead time (request to delivery), throughput (items completed per time period), cumulative flow diagram, (6) Replenishment - how and when to pull new work, (7) Delivery rhythm - when to release (continuous or cadence). Analyze current flow data to identify: bottlenecks (where work piles up), blockers (what stops flow), and variability (what causes unpredictability). Recommend WIP limit adjustments and process improvements. Create a flow efficiency dashboard."
Design for Six Sigma — DMADV
Use when you're designing something new — a product, service or process — to meet customer-critical requirements first time, where no existing process exists to improve.
Learning objective
Design to measurable quality using Define–Measure–Analyse–Design–Verify: capture the voice of the customer as CTQs, generate and de-risk concepts, then prove the design holds. DMADV vs DMAIC: use DMAIC to improve an existing process that underperforms; use DMADV to create a new one (or replace a process beyond repair). AI runs the statistics and generates concepts; you own the CTQ targets and the go/no-go.
Stays human
Translating a customer's words into the right CTQ — and the tolerance you'll hold it to — is a judgement with cost and risk attached. The verify-gate decision to release a design carries accountability AI cannot hold.
Watch-outs
- AI capability claims (Cp, Cpk) are only as good as the data behind them — validate the measurement system first.
- An FMEA generated by AI lists failure modes; weighting severity for your context is human work.
- A statistically “verified” design can still fail real users — keep human acceptance testing in the loop.
Responsible prompt note
Design specs, test data and customer requirements are usually Amber and may be IP. Abstract proprietary detail; never paste raw test datasets or full specifications into an external tool.
Full training script — 5 stages, sub-steps, AI fit, consultant actions & prompts preserved from source›
15.1Define
Sub-steps: identify project goals · understand customer needs · define CTQs (Critical to Quality) · scope the project.
AI fit
- Goal definition: help craft SMART goals
- Voice-of-customer analysis: extract CTQs from customer data
- Scope optimisation: define optimal project boundaries
- Risk identification: identify potential project risks
Consultant action
Define the project with clear, strategy-aligned goals, gather voice-of-customer data to identify CTQ requirements, and scope with clear boundaries, deliverables and metrics — captured in a charter with stakeholder alignment.
Responsible prompt — Define
"Define a DFSS (DMADV) project for [new product/process]. Create project charter with: (1) Business case - why this project, strategic alignment, financial impact (cost of poor quality, opportunity size), (2) Problem/opportunity statement - clear description of what we're designing and why, (3) Goal statement - SMART objectives (specific, measurable, achievable, relevant, time-bound), (4) Voice of Customer - gather customer needs through surveys, interviews, focus groups, observations, (5) CTQs (Critical to Quality) - translate customer needs into measurable requirements with targets and tolerances, (6) Project scope - what's in scope, out of scope, boundaries, (7) Stakeholders - who affects/is affected by this project, (8) Team - roles and responsibilities, (9) Timeline - major milestones and gate reviews, (10) Risks - potential obstacles and mitigation plans. Extract CTQs from this customer data [paste data]. Create SIPOC diagram (Suppliers, Inputs, Process, Outputs, Customers) at high level."
15.2Measure
Sub-steps: identify customer needs · translate to specifications · measure capabilities · establish baselines.
AI fit
- QFD (Quality Function Deployment): create the house of quality
- Specification translation: translate needs to technical specs
- Capability analysis: assess current capabilities
- Baseline establishment: set statistical baselines
Consultant action
Translate customer needs into technical specifications using QFD and the House of Quality, assess current capability, establish baselines and identify gaps between requirements and planned performance.
Responsible prompt — Measure
"Execute the Measure phase for [product/process]. Create: (1) QFD House of Quality - matrix showing: customer requirements (WHATs) with importance ratings, technical requirements (HOWs), relationship matrix (strong/medium/weak correlations), correlation matrix (HOWs vs HOWs - positive/negative interactions), competitive assessment, target values for technical requirements, (2) Specifications - translate CTQs into measurable technical specifications with upper/lower limits, (3) Capability assessment - current process/product capability (Cp, Cpk) vs. required capability, (4) Baseline data - collect data on current state or benchmarks from similar products/processes, (5) Gap analysis - where are we vs. where we need to be, (6) Measurement system analysis - ensure measurement tools are accurate and precise (Gage R&R). Prioritize technical requirements based on customer importance and competitive gaps. Identify trade-offs between conflicting requirements. Create data collection plan for validation."
15.3Analyse
Sub-steps: develop design concepts · evaluate alternatives · select best concept · perform risk analysis.
AI fit
- Concept generation: create design alternatives
- Concept evaluation: score concepts against criteria
- Trade-off analysis: analyse design trade-offs
- FMEA: identify potential failure modes
Consultant action
Facilitate development of multiple concepts, evaluate against CTQs and constraints with weighted matrices, select the optimal concept, and run FMEA to identify and mitigate failure modes before detailed design.
Responsible prompt — Analyse
"Execute the Analyze phase for [design challenge]. Create: (1) Design concepts - generate 3-5 distinct design alternatives that meet CTQs (use brainstorming, TRIZ, biomimicry, or analogous inspiration), (2) Evaluation criteria - weighted scoring matrix including: customer value, technical feasibility, cost, time to implement, risk, alignment with CTQs, (3) Concept evaluation - score each concept against criteria, calculate weighted scores, identify strengths/weaknesses, (4) Trade-off analysis - where do concepts excel vs. compromise, Pugh matrix comparing concepts to baseline, (5) FMEA (Failure Modes and Effects Analysis) - for selected concept(s): identify potential failure modes, effects, causes, current controls, severity (1-10), occurrence (1-10), detection (1-10), calculate RPN (Risk Priority Number = S×O×D), prioritize high-RPN items for mitigation, (6) Concept selection - recommend best concept with justification. Generate concept descriptions with sketches or diagrams. Perform tolerance analysis. Create risk mitigation plan for top 10 failure modes."
15.4Design
Sub-steps: develop detailed design · optimise design · validate through simulation · prepare for verification.
AI fit
- Detailed design: AI-assisted CAD and design tools
- Design optimisation: optimise design parameters
- Simulation: AI-powered simulation and digital twins
- Tolerance analysis: optimise tolerances
Consultant action
Develop detailed specifications — drawings, models, process flows, procedures — using optimisation, simulation and tolerance analysis so the design meets CTQs robustly across expected variation, and prepare verification plans.
Responsible prompt — Design
"Execute the Design phase for [product/process]. Create: (1) Detailed design specifications - complete technical drawings, 3D models, process flow diagrams, work instructions, material specifications, (2) Design optimization - use DOE (Design of Experiments) or simulation to optimize design parameters for performance, cost, and robustness, (3) Tolerance design - specify tolerances that ensure CTQs are met considering variation (use statistical tolerance analysis, not just additive), (4) Robustness testing - Taguchi methods or simulation to ensure design performs well across noise factors (environmental variation, wear, manufacturing variation), (5) Simulation and modeling - digital twin, FEA (Finite Element Analysis), CFD (Computational Fluid Dynamics), or process simulation to validate design before physical prototyping, (6) Design for X - optimize for manufacturability (DFM), assembly (DFA), reliability (DFR), cost (DFC), (7) Verification plan - detailed test plan to verify design meets all CTQs and specifications. Create design FMEA update. Generate bills of materials (BOM), process maps, control plans. Document design rationale and decisions."
15.5Verify
Sub-steps: test design · validate performance · pilot production · implement and control.
AI fit
- Test design: design optimal test plans
- Performance validation: analyse test results
- Statistical process control: monitor process stability
- Control-plan development: create control plans
Consultant action
Verify through rigorous testing, pilot runs and validation against all CTQs, implementing SPC and control plans and establishing monitoring so the design performs in the real world before full-scale rollout.
Responsible prompt — Verify
"Execute the Verify phase for [design]. Create: (1) Test plan - comprehensive testing protocol covering: functional tests (does it work?), performance tests (meets specifications?), stress tests (limits and failure points?), reliability tests (long-term performance?), user acceptance tests (meets customer needs?), (2) Pilot production - small-scale production run to validate manufacturing process, identify issues, collect capability data (Cp, Cpk), (3) Validation results - statistical analysis showing design meets all CTQs with evidence (hypothesis testing, confidence intervals), (4) Control plan - how to maintain performance: control charts, inspection frequencies, reaction plans for out-of-control conditions, (5) SPC (Statistical Process Control) - implement control charts for key CTQs, establish control limits, train operators, (6) Process capability - demonstrate process is capable (Cpk ≥ 1.33 or per requirements) and stable, (7) Implementation plan - rollout strategy, training, documentation, change management, (8) Lessons learned - what worked, what to improve, knowledge transfer. Create final project report with ROI calculation. Develop monitoring dashboard for ongoing performance. Plan for continuous improvement."
Open Innovation
Use when the best idea won't come from inside — you need to source, evaluate and integrate external technology, startups or research to move faster than internal R&D alone.
Learning objective
Run the funnel — define needs, scan, evaluate, engage, integrate, manage the portfolio — with AI scouting patents, startups and papers at scale while you make the partnership, IP and integration calls. Reach is automated; judgement is not.
Stays human
Partnership is relationship and trust; deal terms and IP carry legal consequence. AI can shortlist and draft, but choosing whom to partner with and on what terms is a human commitment with real downside.
Watch-outs
- AI-sourced startup/patent lists go stale and miss context — verify funding, status and freedom-to-operate.
- Don't let AI draft binding terms; legal review is non-negotiable.
- “Strategic fit” scores are only as good as the criteria you set — own those.
Responsible prompt note
Innovation needs and partner discussions reveal strategy and may touch confidential or IP-sensitive ground — often Amber/Red. Keep specific technology gaps and partner identities out of external tools; describe the search field generically.
Full training script — 6 stages, sub-steps, AI fit, consultant actions & prompts preserved from source›
16.1Define innovation needs
Sub-steps: identify technology gaps · articulate challenges · define search fields · set objectives.
AI fit
- Gap analysis: identify technology and capability gaps
- Challenge formulation: craft clear innovation challenges
- Search-field definition: identify relevant domains
Consultant action
Work with internal teams to identify specific gaps and challenges suited to external solutions, articulating clear challenge statements and the search fields — technologies, industries, geographies — to explore.
Responsible prompt — Define needs
"Define open innovation needs for [organization]. Create: (1) Technology roadmap - current capabilities vs. future needs, identify gaps where external innovation could accelerate development, (2) Innovation challenges - specific, well-defined problems that external partners could solve (format: 'How to [achieve outcome] while [meeting constraints]'), (3) Search fields - which technologies, industries, scientific domains, or geographies to explore (be specific: e.g., 'battery technology for EVs,' 'AI for predictive maintenance in manufacturing'), (4) Objectives - what we want from open innovation (acquire technology, co-develop, license, invest, acquire companies), (5) Make vs. buy vs. partner analysis - which capabilities to build internally vs. source externally, (6) Success criteria - how we'll measure open innovation success (time to market, cost savings, revenue from new products, etc.). Prioritize challenges by strategic importance and internal capability gap. Create challenge briefs for top 5 opportunities suitable for external posting."
16.2Scan external environment
Sub-steps: search for solutions · monitor trends · identify partners · map innovation ecosystem.
AI fit
- Automated scouting: scan patents, papers, startups continuously
- Trend detection: identify emerging technologies
- Partner identification: find potential innovation partners
- Ecosystem mapping: map innovation landscapes
Consultant action
Scan the external landscape — startups, universities, research institutions, suppliers, competitors, adjacent industries — using AI tools to monitor emerging technologies, patents and papers, mapping the relevant ecosystem.
Responsible prompt — Scan environment
"Scan the external innovation environment for [challenge/technology area]. Create: (1) Technology landscape - map of existing solutions, approaches, and technologies addressing similar challenges, (2) Startup scouting - identify 20-30 startups working on relevant technologies (name, technology, stage, funding, contact), (3) Academic research - key universities, research centers, and professors working on relevant topics (recent publications, patents, ongoing projects), (4) Corporate innovation - companies (including competitors) with relevant capabilities or technologies, (5) Emerging trends - weak signals and emerging technologies that could impact this space in 3-5 years, (6) Innovation ecosystem map - visualize all players (startups, corporates, universities, investors, suppliers) and their relationships, (7) Technology intelligence - patent analysis, publication trends, funding patterns. Use AI tools to scan: patent databases (Google Patents, Lens), academic databases (Google Scholar, IEEE), startup databases (Crunchbase, PitchBook), and news/trends. Create shortlist of top 10 most promising external solutions/partners with rationale."
16.3Evaluate opportunities
Sub-steps: assess fit · evaluate maturity · analyse IP landscape · calculate value potential.
AI fit
- Fit assessment: score external solutions against needs
- Maturity evaluation: assess technology readiness levels
- IP analysis: analyse patent landscapes and freedom to operate
- Value prediction: estimate potential value
Consultant action
Evaluate opportunities against strategic fit, technology maturity (TRL), IP landscape and value potential, conducting due diligence on partners and building business cases for the most promising.
Responsible prompt — Evaluate opportunities
"Evaluate these external innovation opportunities [list startups/technologies/ partners]. For each opportunity, assess: (1) Strategic fit - how well does it address our challenge and align with strategy (1-10 score), (2) Technology maturity - TRL (Technology Readiness Level 1-9), what's proven vs. what's theoretical, development timeline to commercial readiness, (3) IP landscape - patent position (strength, breadth, freedom to operate), licensing requirements, IP risks, (4) Team/partner quality - expertise, track record, cultural fit, stability, (5) Business model - how partnership would work (license, joint development, acquisition, investment), costs, revenue sharing, (6) Value potential - market size, competitive advantage, revenue/cost impact, strategic option value, (7) Risks - technical, commercial, legal, integration risks with mitigation plans. Create scoring matrix to rank opportunities. Perform competitive analysis - how does this compare to alternatives? Develop business case for top 3 opportunities including: investment required, expected ROI, timeline, risks, and recommendation. Conduct reference checks and technical due diligence."
16.4Engage partners
Sub-steps: initiate contact · negotiate terms · structure deals · build relationships.
AI fit
- Partner outreach: draft personalised outreach
- Deal structuring: suggest optimal partnership models
- Contract analysis: review and suggest contract terms
- Relationship management: track and nurture partnerships
Consultant action
Facilitate engagement — initiate contact, negotiate terms, structure deals (licensing, JV, co-development, acquisition) and build collaborative relationships — ensuring agreements protect IP while enabling collaboration.
Responsible prompt — Engage partners
"Design partner engagement strategy for [selected partner/opportunity]. Create: (1) Outreach approach - how to initiate contact (warm introduction, direct approach, conference, intermediary), personalized value proposition for the partner, (2) Engagement model - type of partnership (licensing, co-development, joint venture, strategic investment, acquisition) with pros/cons of each, (3) Deal structure - key terms (exclusivity, territory, field of use, payment structure - upfront, milestones, royalties, equity), IP ownership and licensing terms, governance structure, (4) Negotiation strategy - our priorities vs. flexibility, their likely priorities, potential deal breakers, BATNA (Best Alternative to Negotiated Agreement), (5) Due diligence checklist - technical, legal, financial, commercial, cultural due diligence items, (6) Contract framework - key clauses (IP, confidentiality, termination, dispute resolution, liability), (7) Relationship management - governance structure, communication cadence, escalation paths, success metrics. Draft initial outreach email and term sheet. Create partnership playbook for ongoing collaboration."
16.5Integrate solutions
Sub-steps: adapt to context · combine with internal capabilities · implement · scale.
AI fit
- Integration planning: plan integration pathways
- Capability mapping: identify how to combine external and internal
- Implementation support: guide implementation
- Scaling optimisation: optimise scaling strategies
Consultant action
Support integration of external innovation with internal capabilities — adapting to context, combining knowledge, managing implementation — and create scaling strategies from pilot to enterprise-wide deployment.
Responsible prompt — Integrate solutions
"Plan integration of [external solution/technology] into [organization]. Create: (1) Integration assessment - gaps between external solution and internal needs (customization required), compatibility with existing systems/processes, cultural fit, (2) Capability combination - how to merge external technology with internal capabilities for maximum value (what each side contributes), knowledge transfer plan, (3) Adaptation plan - modifications needed for our context (regulatory, market, technical, cultural), who will do the adaptation, timeline, (4) Implementation roadmap - pilot phase (scope, success criteria, timeline), scale-up phase (staged rollout plan), full deployment, (5) Change management - stakeholder analysis, communication plan, training needs, resistance management, (6) Resource requirements - budget, people, facilities, technology infrastructure, (7) Risk management - integration risks (technical, organizational, cultural) and mitigation strategies, (8) Scaling strategy - how to move from pilot to enterprise-wide (standardization, replication, localization). Create integration team structure with roles. Develop 90-day integration plan with milestones. Design metrics to track integration success."
16.6Manage portfolio
Sub-steps: track performance · balance portfolio · optimise mix · learn and adapt.
AI fit
- Portfolio analytics: track open innovation portfolio performance
- Balance optimisation: optimise internal vs external mix
- Performance prediction: predict which partnerships will succeed
Consultant action
Establish portfolio management to track multiple initiatives, balance risk/reward and time horizons, optimise the internal-vs-external mix, and create learning loops to keep improving the open-innovation approach.
Responsible prompt — Manage portfolio
"Design open innovation portfolio management system for [organization]. Create: (1) Portfolio dashboard - track all open innovation initiatives (status, investment, milestones, risks, outcomes), (2) Performance metrics - input metrics (number of partnerships, investment), process metrics (time to deal, time to integration), output metrics (revenue from partnerships, cost savings, new products), outcome metrics (market share, competitive advantage), (3) Portfolio balance - ensure mix across: risk levels (low/medium/high), time horizons (short/medium/long-term), innovation types (incremental/radical), partnership types (licensing/co-development/acquisition), (4) Resource allocation - how to allocate budget and people across portfolio (strategic priority, potential value, probability of success), (5) Stage-gate process - decision points for continue/kill/scale decisions with criteria, (6) Learning system - capture lessons from successes and failures, share across organization, continuously improve open innovation capabilities, (7) Internal vs. external balance - optimal mix of internal R&D vs. external innovation based on strategy and capabilities. Create portfolio review cadence (monthly operational, quarterly strategic). Design kill criteria to stop underperforming initiatives. Benchmark portfolio performance against industry peers."
Conclusion · integrating AI responsibly
Six commitments that keep you the consultant
Every diagram in this guide is one instance of a single idea: AI carries the computational load, and a named human carries the judgment and the accountability. These six commitments are how that idea survives contact with a real engagement, a real deadline, and a real client.
Augmentation, not replacement
AI handles pattern-finding, drafting, calculation and first-pass synthesis at scale. You provide the strategic wisdom, ethical judgment and stakeholder empathy. If a task needs none of the latter, automate it; if it needs any, you stay in the loop.
Transparency
Clients should understand when and how AI was used in their engagement. Disclose it plainly, the way you would disclose any method. Trust is the asset the profession runs on; a hidden tool is a borrowed risk against it.
Validation
No AI output reaches a client unchecked. Verify every fact, source and figure; stress-test the logic; strip the confident inventions. The gate on each diagram marks where this is not optional — a number in a board pack or a recommendation that moves money or people.
Bias awareness
Models inherit the skew of their training data and the framing of your prompt. Actively probe for who or what is missing, whose perspective is over-weighted, and where a plausible answer would quietly disadvantage a group. Name the bias before it ships.
Data privacy
Client data put into an AI tool must be protected, governed and lawful. Sanitise before you prompt, respect the Green/Amber/Red protocol, and never let convenience override confidentiality, IP or the terms you agreed with the client.
Continuous learning
Both the consultant and the tooling must improve from outcomes. Capture what worked, retire what didn't, and revisit which rung of the competency ladder you genuinely stand on as the tools — and the risks — keep moving.
Master the traditional methodology and the responsible-AI discipline together and you deliver more, faster, without surrendering the trust and accountability that define the work. The machine makes you quicker. The judgment is still yours to give — and still the reason a client called you.
The whole guide in one line
Let the model compute.·You decide.