How to Prepare Procurement Data for AI

The short answer

To prepare procurement data for AI, assemble four things: the contract documents you hold, a supplier master export, a spend extract by supplier and period, and whatever renewal or category trackers your team maintains. Then reconcile supplier identity across those sources so that spend and contracts can be attached to the same supplier, and record what is missing.

You do not need clean data. You need known data — a defined scope, a consistent supplier identity, and an honest list of gaps. Waiting for a data cleansing programme to finish before starting analysis is the most common reason nothing happens.

1. Define the scope before the export

Start with a slice you can complete rather than the whole estate. A single category, a top-twenty supplier list by spend, or the agreements renewing in the next twelve months are all defensible scopes, and each produces findings that justify the next step.

A narrow scope also makes gaps visible quickly. If you cannot locate contracts for eight of your twenty largest suppliers, that is a material finding in itself and one worth knowing in week one rather than month three.

2. Request the four core exports

Spend extract

Invoice or purchase order lines by supplier, period, entity and cost centre, covering at least twelve months and preferably twenty-four so a trend is visible. Include the raw supplier name as it appears in the finance system rather than a cleaned version — the inconsistency is diagnostic information.

Supplier master

Supplier records with identifiers, legal or registered name, trading name, category, owner and any criticality or risk rating. Parent and subsidiary relationships where the system records them.

Contract documents

Master agreements plus the attachments that change the commercial position: statements of work, order forms, variations, renewal letters and price lists. A master agreement alone frequently omits the pricing that is actually in force.

Working trackers

Renewal trackers, category plans and savings pipelines. These hold context that exists nowhere in a system, and they also reveal what the team currently believes — which is useful to test against the contracts.

3. Reconcile supplier identity first

Supplier reconciliation is the single highest-value preparation step. The same supplier typically appears under a trading name in the contract repository, a registered entity in finance, an abbreviation in a spreadsheet, and as separate records for regional subsidiaries. Until those are reconciled, total spend is understated and concentration risk is invisible.

Work top-down by value. Reconciling the suppliers representing eighty per cent of spend delivers most of the analytical benefit; the long tail can be treated as a category rather than as individual relationships.

Where a match is genuinely uncertain — similar names, different registration details, unclear group relationships — flag it for a person to confirm. A wrong merge produces a confidently incorrect concentration figure, which is worse than an acknowledged gap.

4. Link contracts to the spend they govern

A contract with no spend attached cannot be assessed for exposure, and spend with no contract attached is a finding in its own right. Linking the two is what allows a notice deadline to be ranked by value, an indexation clause to be tested against what has actually been charged, and a volume commitment to be compared with actual volumes.

Expect three outcomes from this exercise, all of them useful: contracts matched to spend, contracts with no material spend against them, and spend with no agreement in place. The third is usually the largest and the most immediately actionable.

• Contracted spend under a current, identified agreement
• Contracts with negligible or no spend against them
• Uncontracted spend with no agreement on file
• Spend against an agreement that has already expired

5. Capture the commercial fields consistently

For analysis across a portfolio, the same fields must exist for every agreement in a comparable form. At minimum: parties, effective date, term, renewal mechanism, notice period and resulting notice deadline, contract value, pricing basis, indexation, volume or spend commitments, termination rights and benchmarking rights.

Consistency matters more than completeness. A field recorded as unknown for a quarter of agreements still supports portfolio analysis; the same field recorded three different ways across three teams does not.

6. Judge readiness honestly

Data is ready when the scope is defined, suppliers within it are reconciled, contracts are matched to spend where they exist, and the gaps are documented. It is not ready when nobody can say which agreements are in force, or when supplier records cannot be reconciled at all.

Nita carries out the structuring, reconciliation and extraction described above on the exports you supply, and reports gaps as gaps rather than filling them by inference. The preparation worth doing before that is mostly deciding scope and locating the documents.

• Scope defined and agreed
• Suppliers representing the majority of spend reconciled
• Contracts located for the suppliers in scope
• Spend covering at least twelve months
• Known gaps written down rather than assumed away