How to Answer "Did You Use AI?" in a Proposal or Tender

The question has moved from occasional to standard. Procurement questionnaires, RFP response forms, and client onboarding packs now routinely ask whether AI was used in producing the response, whether it will be used in delivery, and what safeguards apply.

Two bad answers dominate. The first is a denial that is not quite true, which is a risk you have created for no benefit. The second is a page of hedged language that reads as evasion and prompts three follow-up questions.

The good answer is short, specific, and distinguishes between two things most questionnaires conflate.

The Distinction That Does The Work

Assistance. A person wrote the substance and used a tool to edit, rewrite, or check it. Human authorship, human judgment, human accountability. The tool changed the form.

Generation. A model produced the substance from a prompt, and a person reviewed it to some degree. The tool supplied the content.

Almost every client concern is about the second. Almost every actual professional use is the first. If your answer does not separate them, the reader applies their generation-shaped worry to your assistance-shaped practice, and you spend the next email correcting it.

State the distinction, then say which one you do.

What A Good Answer Contains

Four elements, in about six lines.

What you actually do. Named tools, named use cases.

Where confidential material is processed. This is the question underneath the question. Clients ask about AI because they are worried about their information reaching a third party.

Who is accountable. A person, not a process.

What you do not do. Explicit boundaries are more reassuring than assurances, because they are checkable.

Notice what is not on the list: a compliance claim. Do not write “fully compliant with the EU AI Act”. It is a legal conclusion, it invites a follow-up you may not want, and it is not what they asked. The Act’s own regulatory framework page makes clear that obligations attach to specific roles and risk categories, so a blanket compliance claim is not even the right shape of statement. Describe your practice and let them assess it. The reasoning is in AI disclosure obligations in 2026.

The Example

Before:

We are committed to the responsible use of artificial intelligence. All AI usage within our organisation is governed by a comprehensive AI policy and is subject to human oversight at all stages. We are fully compliant with applicable AI regulations and do not use AI in any way that would compromise client confidentiality. Please let us know if you require further information.

After:

This proposal was written by our team. AI tools were used for editing and rewriting drafts our consultants wrote, in the same way we use a spelling checker. No section was generated from a prompt.

In delivery, we use AI the same way. Where a document contains your confidential information, the model runs locally on the consultant’s machine and the text is not transmitted to any provider.

For non-confidential internal drafting we use a hosted provider, listed in our subprocessor schedule with the applicable data processing agreement.

Accountable owner: Priya Raman, Head of Delivery. Position last reviewed 6 August 2026.

We do not use AI to make decisions about your matter, to generate legal or technical conclusions, or to process your data for any purpose other than producing your deliverables.

Five short paragraphs. Every sentence is checkable. It answers the confidentiality question directly, names a person, and states boundaries.

The first version says nothing and takes just as long to read. “Comprehensive AI policy” and “human oversight at all stages” are the kind of phrases that make a reader look harder rather than less hard.

Why Defensiveness Reads Badly

Procurement teams read a lot of these. The pattern they have learned is that the length of the AI paragraph correlates inversely with the specificity of the practice behind it.

A supplier who says “we use AI to edit drafts, and confidential text is processed locally” sounds like a supplier who has thought about it. A supplier who produces four paragraphs about responsible AI principles sounds like a supplier who has not, and is covering.

Confidence here comes from specificity, not from reassurance. That is the same principle that makes any client communication land, and it applies with particular force in how to write a proposal executive summary.

When The Answer Is Genuinely Awkward

Two cases worth handling honestly.

You used generation for part of it. Say which part, and say what review it received. “Section 6, the market overview, was drafted with AI assistance from public sources and reviewed by our analyst” is a fine answer. Concealing it and being asked later is not.

The client’s form asks you to certify no AI was used. Some forms still do. If that is not true, do not certify it. Answer with what you actually do and let them decide, because a false certification is a contractual problem you are creating voluntarily for a deal you might have won anyway.

A surprising number of clients accept the accurate answer readily. The ones who do not were going to find out eventually.

The Context For This

A Wrivio Context for procurement answers could say:

Rewrite this as a precise answer to a client procurement question. Neutral, confident register, complete sentences. Keep every claim scoped exactly as written, including limitations and named people. Do not add claims of compliance or certification, do not add principles or commitments that are not in the original, and do not generalize a specific statement. Keep the result no longer than the input.

Press Ctrl+Shift+Space, paste the draft, and check the diff. Two things to watch for in this genre: the rewrite adding “fully compliant” or similar, and the rewrite padding a specific answer with responsible-AI language. Both are the register pulling the model toward the bad version, and both are visible in the diff.

There is a mild irony in using a rewriting tool on your answer about using rewriting tools. It is also the honest demonstration of the distinction: you wrote the substance, the tool fixed the phrasing, and you are saying so.

Common Questions

Do I have to disclose AI use in a proposal?

If the client asks, answer accurately. Absent a question or a contractual term, using AI to edit your own writing is generally treated as tooling rather than something requiring proactive disclosure, though your professional body may have its own rules.

What if the client’s form asks me to certify no AI was used?

Do not certify something untrue. Describe what you actually do and let them decide. A false certification creates a contractual problem that is worse than the deal risk.

Should I say we are compliant with the EU AI Act?

Avoid it. It is a legal conclusion rather than an answer to the question, and it invites scrutiny you do not need. Describe your practice instead.

How long should the answer be?

Around five short paragraphs. Length in this section reads as covering rather than as thoroughness.

Download Wrivio for Windows to draft proposal responses with a local model, so client-confidential material stays on your machine while you write about keeping it there.