A clause lost in the tender documents

The problem

The firm depends on public tenders, and every tender comes with dozens of pages of documents. Going through each one took hours, so the team did not read it all. On a recent one, a special insurance policy required to sign the contract slipped through, and that cost them.

What we did

Build

We set up an assistant in the AI account the firm already has. It takes the specifications and the tender rules for a new tender and flags whatever departs from the firm’s standard: insurance, bonds, penalties, deadlines, licences and price revision. It also points out contradictions between the documents and the clarifications worth asking for.

Behind it is a fixed list of 42 questions, checked on every tender. The output is a report with the deviations ranked by severity, each one with the clause and the page it sits on.

Before handing it over, we agreed an acceptance test: on a past tender, the assistant had to catch a clause chosen in advance.

The result

• In the test on a past tender, the assistant put the insurance policy whose absence would have voided the award at the top of the list. That was the clause agreed for the acceptance test.
• On the same tender, it also flagged critical deviations, points to check and contradictions between the documents.
• The team no longer needs hours to go through a tender. With the deviations ranked by severity, it decides quickly, on each tender, whether to go ahead and what it needs to look at again first.

The aim is to see, before deciding to bid, the clauses that could lose a contract already won or tie up cash.