CASE STUDY · AGRIBUSINESS

Migrating a contract back catalogue without trusting the AI.

An AI tool read the contracts. A person checked every field before anything reached the live system.

Region
Australia
Sector
Agribusiness and fresh produce group
Delivery scope
Coupa application managed services across functional (L2) and technical (L3) tiers; contract lifecycle legacy migration
Technology
Coupa Contracts, SAP, AI-assisted document extraction, Valtatech Application Managed Services
Contract LifecycleCoupaSAPManaged servicesData migration
2 tierL2 functional and L3 technical support
BatchedLoaded in reviewable stages
FlaggedMissing values reported, never guessed

The challenge

An Australian agribusiness runs procurement on Coupa with SAP behind it, supported by Valtatech under application managed services across functional and technical tiers.

Years of legacy agreements sat outside that platform as PDFs. Until they exist in Coupa as structured records, none of the things a contract platform is for — renewal dates, spend ceilings, ownership, assignment terms — can be reported on or enforced.

Contract migration is where these programmes usually stall. Re-keying hundreds of agreements by hand is slow enough that it never gets funded. Letting a tool do it unsupervised produces a contract register nobody trusts, which is worse than having no register at all.

What we did

Valtatech used an AI tool to convert the contract PDFs into a Coupa-compatible format for bulk upload, carrying the legal agreements and attachments through as part of the load.

The tool produces two files, and the second one is the point. Alongside the extracted metadata it generates a missing criteria report — an explicit list of what it could not find or could not resolve. It flags gaps rather than filling them with a plausible guess, which is the failure mode that quietly corrupts a contract register.

A manual review then populates the required custom fields and validates every value against what actually exists in the platform: active users, active suppliers, valid legal entities, available commodity names and payment terms. A value that looks right but is not configured will either fail the upload or, worse, load as something wrong.

The load runs in batches rather than one pass, so each tranche is reviewed and corrected before the next begins.

THE OUTCOME

A legacy contract back catalogue migrated into structured, reportable records, with every field validated by a person before it went live.

Valtatech is software vendor agnostic. The technology named here is what this client selected; we start from requirements, operating model and existing systems. See how we approach Source to Pay →

The AI does the reading, which is the part that does not scale by hand. It does not do the deciding. That split — machine extraction, human validation before anything commits — is the same principle behind our forward deployed analysts on VIA and PayGuard.

Facing something similar?

Talk to us about what this would look like in your business.

Start a conversation

← All case studies