All guides / Public sector bids
Can AI help write government tenders?
Yes. AI can help sort your evidence and turn it into a tender draft. Give it facts your team has checked. Keep prices, promises and the final bid with the person who owns the tender. Check the buyer’s rules before you use the tool.
Try one question from an old bid first. The interview below helps you work out what you can show, where the gaps are and who should review the answer.
Try the promptRead the rules. Show the evidence. Check the promises.
Read what the buyer asks for
Put the bid question and word limit in one place. Add the scoring guide. Add later updates from the buyer. Check that your team is using the same version of the rules.
More on this step
PPN 017 is government guidance on AI use in bids. It includes questions the buyer can ask you. Read the actual tender and answer its questions with care. Keep a note of how you used AI and who checked the draft. Do not assume that every buyer uses the same form.
Read the source: Cabinet Office: PPN 017 on AI use in bids
Make a sheet of facts you can prove
For each part of the question, list the claim you want to make. Add the record that supports it and the person who can confirm it. Check dates on training, certificates and past work.
More on this step
Leave a blank when evidence is missing. Ask the tool to work from this sheet and show any gaps. If it adds a claim that sounds good, find its source. Remove the claim if no one can support it. A strong-sounding sentence is not evidence of what your firm can do.
Check every promise before sending
Read the draft with the question beside it. Does it cover each part? Check the named roles, steps, prices and dates. Ask the delivery team whether it can do what the draft promises.
More on this step
Give one bid owner the final say. Let them see the proof sheet and any open questions. Keep late changes visible for a fresh check. Use the route set by the buyer. Send the bid before the deadline. Use the government’s supplier guidance to learn about the central digital platform and the Procurement Act.
Read the source: Cabinet Office: guidance for suppliers
Not sure where to start?
Talk through your tender evidence
Choose a past question or a public example. Use invented company details. Keep private buyer documents out until you have approval to use them.
Try this in your AI app
ChatGPT
Claude
Gemini
CodexUsing Gemini? The button copies the prompt. Paste it into the new chat.
Help me prepare a tender answer backed by facts our delivery team has checked.
Ask a question, then wait for my reply. First ask for the buyer’s question, the word limit and the scoring points we are allowed to share. Ask which rules we need to check, including any questions about AI use.
Explore how our business would do the work. Ask for the steps, the people involved and the records that back our claims. Mark missing evidence as a gap. Do not invent projects, results, certificates or staff skills.
Repeat what you believe we can prove. Ask me to check your summary before you go on. Then build a table with each part of the question, our evidence, its owner and what is missing.
Suggest an outline for the answer using only those facts. Leave prices and promises for our bid owner to agree. Finish with the checks we need before submitting. Do not claim the outline will win the contract.
Watch Warren's story
Warren gave AI the context of his work
Warren had tried ChatGPT but did not know how to make it useful in his hotel role. We built a profile of his work and set up a Claude Project with hotel standards. After Leader Lab coaching, he built a reporting dashboard of his own.
Read the source: Warren’s Leader Lab story
Testimonial · 0:24
"I’ve learned more than I have playing around with it in the last couple of months."
Warren Lavender · Luxury hospitality
Worked example
Use the delivery team’s own facts
A facilities management firm bids to look after a public building. The question asks how it will cover staff absence. The team gives AI its checked steps and the roles that carry them out. The draft must leave a gap for any cover time that has not been agreed.
A construction firm uses its own project records to draft a tender answer about past work. The bid owner checks the dates, scope and named roles. They remove any claim that belongs to a different job. The people who will deliver the work check each promise before it is submitted.
Made-up example.
Tender question
Keep the buyer’s rules and scope together.
Delivery facts
Use checked cover steps and project records.
Bid approval
Confirm the promises with the delivery team.
A few more things to know
Check which files you may share
Use public files or made-up notes for a test. Before using live records, check the account and terms with your data lead. The ICO gives guidance on AI, personal data and keeping it secure.
Keep private customer records apart from public project descriptions. Ask through your normal route before naming a client or sharing a result. A small source pack is easier for a reviewer to trace. Include only the files the answer needs, with clear names and current dates.
Read the source: ICO: AI and data protection
Measure the work, then decide
Time the full trial from finding evidence to the final review. Record which facts were missing and how long staff spent fixing the draft. Compare that with writing the answer in your usual way.
A shorter writing stage may still leave more work for the technical lead. Ask whether the method helped the team produce a sound answer. Treat winning a bid as a separate result. Keep a useful proof sheet and the review notes for your next decision.
Do we have to say that AI helped?
Read the buyer’s instructions and answer the disclosure questions it asks. PPN 017 gives examples buyers may use. Your record of drafting and checking will help you answer with care.
What if we lack a strong project example?
Show the proof you do have and explain its limits. Let the bid owner decide whether to proceed. Never pass off an invented job as work your business has delivered.