Using AI to write tender responses

What government policy says

The Cabinet Office's PPN 017, Improving transparency of AI use in procurement, applies to central government departments, their executive agencies and non-departmental public bodies; other public bodies may choose to follow it. It says suppliers' use of AI is not prohibited during the commercial process, and that buyers should understand the risks, as they would if a supplier had used a bid writer.

• Buyers may ask whether you used AI or machine learning tools, including large language models, in any part of your tender, and where.
• They may ask whether AI is part of the product or service you are offering.
• The note says those questions should not be scored or taken into account in assessing the tender: they are for information.
• Buyers can check what a tender says through clarification questions, supplier presentations and site visits.
• It warns that AI-generated content may include inaccurate or misleading statements, and that suppliers should not use the buyer's confidential information to train AI tools.

Councils, NHS bodies and universities set their own approach, and may add their own declaration. Read the instructions to tenderers and answer any AI question honestly.

Where AI genuinely helps

• Understanding the documents. Summarising a long specification, listing every question with its word limit and weighting, and building a compliance checklist.
• Structure. Turning a question's bullet points into subheadings, so the evaluator finds each element where they expect it.
• Drafting from your own evidence. Give it your case study, your method and your numbers, and ask for a first draft in the question's structure.
• Editing. Cutting to the word limit, removing jargon and filler, and checking the answer covers everything the question asked.
• A second marker. Asking it to score your draft against the buyer's published descriptors and say what is missing.

Where it loses marks

Evaluators award marks for specific evidence against a published scheme (how answers are scored). An AI tool asked to “write an answer about our approach to quality” produces a fluent answer that any supplier could have submitted, which is exactly what scores a 2. It also tends to fill gaps with plausible detail. A made-up accreditation, client or figure in a tender is a misrepresentation, and it can become a contractual promise if you win.

Confidentiality

Tender documents can contain confidential or personal information, and your own answers contain your pricing and methods. Use a tool whose terms say your inputs are not used to train its models, switch off any setting that shares your conversations, and never paste personal data from a buyer's documents. If the tender says its documents are confidential, treat an AI tool like any other third party you would share them with.

The firms that use AI well spend the time it saves on the parts it cannot do: understanding the buyer, choosing the right tenders and gathering better evidence. Deciding whether to bid is still the decision that matters most.