Industrial AI practice  ·  Gulf  ·  Arabic and English Reply within one working day
Multimodal invoice validation

A Gulf telecom operator's finance function - Multimodal invoice validation

Telecommunications / Finance · Gulf

← All documented proof

The shape of this problem

A human reads it and types it in again

The problem

AP teams were manually comparing supplier PDFs, iSupplier entries, and Oracle ERP records. Three-way matching is the kind of work that is too rule-bound to be interesting and too variable to script: the same supplier's invoice arrives in a different layout each quarter, and the discrepancies that matter are small ones buried among formatting differences that do not. Because the check was manual it was also quietly sampled under load - which is exactly when errors are most likely.

What we built

We used Gemini vision and Vertex AI to extract invoice fields, reconcile all three sources, flag mismatches, and support correction notifications back to the supplier.

What moved

Measure Before After
Invoices reconciled across supplier PDF, iSupplier and ERP a sample under load 100%
Average validation time per invoice 12 minutes 45 seconds
Repeat mismatches from the same supplier after correction notices down 67%

† Measured against the prior level of the same measure. The source publishes the size of the movement, not the figure it moved from.

Average validation time per invoice

Before 12 minutes After 45 seconds

Repeat mismatches from the same supplier after correction notices

Prior level 0 100 200 down 67%

Measured against the client's own prior process, indexed to 100. The source publishes the size of the movement, not the absolute figure it moved from.

Figures are drawn from the practice's own delivery records for the engagement named, measured against the process that preceded it. They have not been through third-party audit, and none is presented as an average across clients.

What it turned on

Closing the loop back to the supplier is what stopped the same mismatch recurring every month - detection alone would have relocated the work rather than removed it. Relevant to finance operations whose control is technically three-way matching and practically a sample.

Service line
AI Document Intelligence
The proof is telecom and financial-services documents, not a construction subcontractor invoice. The machinery is identical; say which it was.

Start here

Which of the four is yours?

Tell us the documents and the monthly volume and we will send the two closest records, with the proof behind each and the honest note where the match is partial.

You get a reply within one working day, from the engineer who would do the work - not a sales sequence.