Public sector buyers are starting to ask a different question in tenders. Not just what will you deliver, but how was this bid produced? Specifically, was AI used, how was it used, and what checks were applied?

That brings the process behind the submission into scope. The issue is not whether suppliers use AI – many will – the issue is whether the answer being evaluated can be trusted. AI disclosure is therefore less about technology and more about assurance.

If AI is used in bid drafting, the risk does not lie with the tool itself. The risk is unchecked content: fabricated examples, inconsistent claims, confident statements without evidence, or generic text that does not reflect actual capability.

This is incredibly important in public sector procurement, where written responses are often used as a proxy for operational reality. Evaluators are judging capabilities based on what is on the page. If AI has shaped that content, it is reasonable for buyers to ask how it was verified.

A credible answer does not need to be long but needs to establish control: how AI was used, who reviewed the output, what checks were applied, whether any sensitive information was protected, and who is accountable for the final submission.

If AI forms part of the service delivery, buyers are testing something more fundamental: whether its use is governable. For example, they’ll need to know what function it performs, the data it interacts with and where – if any – human judgement sits.

This is where weak answers will fail as suppliers describe the tool; buyers are looking for the control environment. What is bounded? What is reviewed? What happens when outputs are wrong? Who is responsible? Claims about efficiency or innovation do not answer those questions. A functional description, tied to data use and human oversight, does.

There is also a commercial implication. Unclear answers create friction: more clarifications, more due diligence, more contractual safeguards, and greater perceived risk.

Transparency will reduce that friction.

That means a big change for people considering bidding for public sector work:

• If AI is used in preparing the bid, disclose it and show the checks.
• If AI is used in delivery, describe its function and the governance around it.

Do not overclaim, do not default to generic language, and do not describe the software when the buyer is asking about responsibility.

This direction of travel is moving towards buyers being less concerned with whether suppliers use AI, and more with whether they can use it safely, transparently and accountably in a public service context.

As with most AI conversations, this is about capability, not technology.