Clean Data, Kept Clean, for the CFO and CPO← the function · storycltd.co.uk

FOR THE CFO AND THE CPO · SUPPLIER DATA & BILLS OF MATERIAL

“AI on dirty data will fail.” They have it backwards.

Your data is never clean. It is a moving baseline: suppliers change banks, merge, rename and go dormant; engineering changes the bill of material every week; prices move with every quote. A one-off cleanse is out of date by the next month-end.

So don't wait for clean data before you bring in AI. Make keeping it clean the first job you give AI. It checks every record, every day, and fixes what it can. When a change moves money, risk or the product, a named human signs it.

Watch the AI steward work →Book a 45-minute session
The argument in the room
“Clean the data first, then bring in AI.”
Data is never finished. AI is what keeps it clean.
Supplier master quality over 12 months, illustrative. Red is decay after a one-off cleanse. Teal is a daily AI steward.

A cleanse project leaves a snapshot. A steward keeps working: it catches each new duplicate, bank change, unit-of-measure error and price move as it arrives, and it records why it acted.

What dirty data costs you, in your language

Same data, two chairs. Pick yours. The demo below will flag the decisions that are yours to sign.

Cash out the door

Duplicate and fraudulent payments

Two records for one supplier means two chances to pay one invoice. A bank-detail change that nobody called back to verify is the classic mandate fraud. Master data is your first payment control.

Working capital

Terms that drift from the contract

A duplicate record set up at 30 days on a 60-day contract gives away a month of cash on every invoice. Nobody decided that. The data decided it.

The P&L and the balance sheet

Standard cost you can't defend

A unit-of-measure error in one BOM line inflates the cost roll-up. That flows into inventory valuation, COGS and margin by product. Your auditor will ask where the number came from.

Close and controls

Evidence, not effort

Every change the AI makes is logged with its source and rule. Every change above the Line carries a named signature. That is a control you can show, not a cleanse you once paid for.

Spend visibility

One supplier, three names

“Acme Fasteners Ltd”, “ACME FASTENERS LIMITED” and a typo at onboarding. Spend splits across all three, so your leverage in negotiation is understated before you walk in.

Category strategy

“Misc” is not a category

Unclassified spend is spend you can't source strategically. The steward classifies to UNSPSC as records arrive, so the spend cube is right on the day you need it.

Supply risk

Sole sources and end-of-life parts

A sole-source supplier whose credit score halves. A capacitor that goes end-of-life with six months' notice. The BOM knows. Someone has to be told in time to act.

Should-cost

The BOM is your negotiation brief

Duplicate part numbers split volume. Price creep hides in quotes. A clean, current BOM tells you exactly what a product should cost and where the next saving is.

Watch the AI steward keep a moving baseline clean

A small supplier master and one bill of material, dirty the way real ones are. Press Let the AI steward work. Then move the calendar on a week and watch new dirt arrive. It always does. Decisions marked Your signature are for the chair you picked above.

0data health
0Week on the calendar
0Fixed by AI, with evidence
0Signed by a named human
0Analyst hours returned (illustrative)

Issue foundWaiting on a signatureJust changedMerged or blocked

Evidence log every change, its source, who decided

    Sample data, invented for this demo. It runs entirely in your browser; nothing is sent anywhere. The checks here are rules so the demo behaves the same every time. In an engagement, the same Line and the same evidence log wrap your own model and your own ERP. The model changes. The Line stays where you drew it.

    The Line for master data: where AI acts, where a named human signs

    This is what makes AI the implementer and not the risk. The rule is simple: if a change can be reversed and is backed by evidence, AI makes it. If it moves money, risk or the product, AI prepares the case and a named human signs.

    Data eventAI acts: fixes it and logs the evidenceA named human signs: AI prepares the case
    Duplicate supplierMerges when tax ID, bank and terms all agree. Keeps records apart when names match but tax IDs differ.Any merge where bank or payment terms conflict. CFO: which terms are contracted?
    Bank-detail changeHolds payments to the new details. Drafts the call-back script to the number already on file.Always. Release only after an out-of-band check. CFO / AP.
    Missing or inconsistent fieldsEnriches tax ID and legal name from the company registry. Fixes country from tax prefix. Normalises postcodes.Only when the registry has no match.
    ClassificationClassifies spend to UNSPSC and replaces “Misc”.Category owner reviews the monthly sample.
    Dormant supplierBlocks for payment after 24 months without spend. Reversible, under an approved policy.Reactivation. Procurement.
    BOM unit of measure and costConverts pack to each. Fills a zero cost from the last purchase order. Repoints orphan parts by specification.,
    Standard cost roll-upUpdates the standard when the move is within 5%.Above 5%: inventory revaluation. Financial Controller.
    Price move on a quoteWithin 10% of the last PO: accepted and logged.Above 10%. Category Manager / CPO: accept, negotiate or resource.
    Supply riskWatches credit scores and lifecycle notices every day.Sole source at risk, end-of-life part. CPO with Engineering: dual-source, last-time buy or alternate.
    Related-party signalMatches supplier bank accounts to employee payroll accounts.Always. CFO with Internal Audit.

    How it runs: a loop, not a project

    Cleansing is a project. Stewardship is a loop. It runs every day, whether or not anyone is watching the data.

    Detect

    Every new and changed record, every day: duplicates, gaps, conflicts, drift, risk signals.

    Diagnose

    Checks the source systems: the registry, the contract, the last PO, the item master. It does not guess.

    Fix or route

    Below the Line, AI fixes it. Above the Line, it builds the evidence pack and routes it to the named owner.

    Prove

    Every change is logged with its rule, its source and its decider. That gives audit a trail and the CFO a control.

    Prevent

    The same checks run as a gate at onboarding and at each engineering change, so new dirt is stopped before it reaches the ledger.

    The first 90 days

    Supplier master first because it protects cash. Bills of material second because they protect margin. Then the gate, so the baseline stays clean.

    DAYS 0–15

    Baseline scan

    A read-only extract of your supplier master and your top 20 bills of material. You get a data-health score, the duplicates, the unverified bank changes and the cost roll-up exposure, each with a £ value attached.

    DAYS 15–30

    Draw the Line

    A working session with the CFO and CPO: which changes AI makes on its own, which ones a named human signs, at what thresholds, and who that human is. It is written down and approved.

    DAYS 30–60

    Steward live on the supplier master

    Daily run, fixes below the Line, evidence packs above it, and the log reviewed weekly. Duplicate-payment and bank-change exposure is closed first.

    DAYS 60–90

    BOMs, then the gate

    Unit of measure, cost, lifecycle and price drift across the bills of material. Then the same checks move upstream into supplier onboarding and engineering change, so new dirt is stopped at the door.

    Bring your supplier master. We'll show you the baseline moving.

    A 45-minute working session with Madeleine: your data problem on the table, your Line drawn for master data, and an honest read on where AI keeps it clean and where your people still sign.

    Book a 45-minute session →

    STORYC LTD (company no. 17293375), trading as AI-Augmented Consulting™ (UK TM application UK00004409146). The AI Augmentation Line™, Evidence Lock™, the AI Augmentation Profiler™ are trade marks of StoryC Ltd. The Augmentation Line® is a registered trade mark of Story-C Ltd (UK00004411978, classes 35 and 41), of which the AI Augmentation Line™ is the form in use. The AI Augmentation Profiler® is a registered trade mark of Story-C Ltd (UK00004411967, classes 35 and 42). Locked Signal™ is a trade mark of StoryC Ltd. EVA³ is used under the registered trade mark E³VA® (UK00003724166, series of two marks, classes 9 and 35), in the name of Madeleine Joubert and used under licence by StoryC Ltd. (UK application in progress). All supplier, part and employee data on this page is fictitious. Privacy · storycltd.co.uk