From exported procurement data to evidence-backed intelligence.

The five stages of the Nita process

Each stage is deliberately practical. Procurement teams rarely have a clean, single source of truth, so the process starts with the material that already exists rather than a data model you have to build first.

1. 1

   1. Provide exported procurement data

   Send contract documents, supplier master data, spend extracts, purchase order reports and any renewal trackers your team maintains. Exports from an ERP, a contract repository, a finance system or a shared folder are all acceptable, in the formats those systems already produce — PDF, Word, CSV or spreadsheet.

2. 2

   2. Nita structures the data

   The supplied material is normalised into a consistent procurement model: suppliers are reconciled to a single record, contracts are linked to the suppliers and spend they relate to, and commercial fields such as contract value, term, renewal mechanism and notice deadline are captured in a comparable form.

3. 3

   3. AI analyses contracts, suppliers and spend

   Analysis reads contract terms alongside supplier spend and renewal position. That means pricing and indexation clauses are considered against actual spend, volume commitments against actual volumes, and termination rights against the date a decision must be taken.

4. 4

   4. Your team asks questions

   Procurement users ask questions in plain English — which contracts have notice deadlines in the next 90 days, where spend is fragmented across similar suppliers, which agreements allow annual price increases — and receive answers drawn from the structured dataset rather than from a general-purpose model's assumptions.

5. 5

   5. Findings and recommended actions

   Outputs identify the finding, the source material it came from, the commercial implication and a recommended next action. A procurement professional reviews and decides. Nita prepares the analysis; it does not take commercial decisions or contact suppliers.

No direct systems integration is required to start

Nita does not currently connect directly to customer ERP, CLM, SRM, finance or procurement systems. This is deliberate for early engagements: an export can usually be produced in a day, whereas an integration involves IT scheduling, security review and change control before any analysis happens.

Working from exports means analysis can start on a defined dataset — one category, one supplier group, or a set of high-value agreements — and the value can be assessed before anyone commits to wider technical work.

• Use the export formats your systems already produce
• Start with a defined scope rather than an enterprise rollout
• No changes to your existing procurement systems
• Refresh the dataset with a new export when the position changes

What procurement data can be included

Contract documents

Master agreements, statements of work, order forms, variations and renewal letters. These carry the terms that matter commercially: renewal mechanism, notice deadline, pricing and indexation, termination rights, volume commitments and benchmarking rights.

Supplier information

Supplier master records, category assignment, ownership, criticality ratings and any risk or performance information your team already holds.

Spend data

Invoice or purchase order extracts by supplier and period, so contracted commercial terms can be compared with what is actually being spent.

Procurement records

Renewal trackers, category plans, savings pipelines and sourcing schedules — the working documents that hold context no system captures.

Why structuring the data matters more than the model

A language model applied to a folder of contracts can summarise individual documents. It cannot tell you which renewal represents the largest commercial exposure, because that requires the contract, the spend against it and the notice deadline to be held in one comparable structure.

Procurement data orchestration — reconciling supplier identities, linking contracts to spend, and normalising commercial fields — is what makes portfolio-level questions answerable. It is the part of the work that most AI tools skip, and the reason results are often disappointing.