Our AI project brief template asks thirteen questions. Each one is there because a missing answer tends to cause a specific problem later: proposals you cannot compare, a quote that grows, or a tool nobody looks after. This page explains why we ask each question and what a useful answer looks like, so you can fill in your AI project brief with confidence.
Key takeaways
- Short, honest answers beat polished ones. "We don't know yet" is fine.
- Send the same brief to every consultant so the replies are comparable.
- The sections on measurement, ownership and after go-live are the ones people skip and later regret.
The business
Why we ask: a consultant who knows your size, customers and current tools can suggest something that fits. Without it, you get generic proposals.
A good answer: two or three sentences. "A 14-person accountancy practice serving local sole traders. We use Microsoft 365, Xero and a shared inbox."
For the bigger picture, see AI for small business.
The problem
Why we ask: the clearest briefs describe one problem, not a wish to "use AI". A consultant can only price and plan what they understand.
A good answer: what happens today, who it affects, roughly what it costs in time or money, and what you have tried. "Two of us spend about a day a week answering the same client questions by email."
How to write an AI project brief has more on framing it.
The outcome wanted
Why we ask: "better customer service" cannot be finished. A described outcome gives everyone the same finish line.
A good answer: something you could check at the end. "Routine client questions answered within an hour, with a person approving each reply."
How you will measure success
Why we ask: without a starting figure, nobody can show the work made a difference. This is the section people most often leave blank.
A good answer: one to three measures and where they stand now, even roughly. If you do not measure it yet, say so: a good consultant will help you take a baseline first.
Data available
Why we ask: most AI work depends on your information. How much there is, where it lives and how tidy it is changes the cost and the approach more than almost anything else.
A good answer: the type of data, a rough size, where it is kept and who can grant access. Flag anything personal or commercially sensitive, so the consultant can plan for it and you can check with your own adviser before anything is shared.
People and oversight
Why we ask: AI changes someone's job. Projects stall when the people affected are not involved, or when nobody is checking what the AI produces.
A good answer: who does the work now, who will use the new way of working, who checks output before it reaches a customer, and what must never be left to the AI.
Security and confidentiality
Why we ask: you want every supplier to tell you, in writing, how they will look after your information. Putting your rules in the brief means they answer against them.
A good answer: any rules you already follow (a client contract, Cyber Essentials, an insurer's conditions), what must not leave the business, and the security questions you want answered. Our list of security questions to ask an AI supplier is a starting point, and AI security for small businesses explains the basics.
Budget band
Why we ask: a band lets consultants propose something realistic without anchoring on one number. It also stops you receiving three proposals at three wildly different scales.
A good answer: tick the band that fits, or "not sure yet" and ask for options at different sizes. Remember running costs such as subscriptions and usage charges. How much does an AI consultant cost? explains how pricing works.
Timeline
Why we ask: a fixed date changes what is possible, and so do your busy periods. A consultant who knows your year-end is in March will not plan a launch for it.
A good answer: when you want to start, any hard deadline and the reason for it, and when your team will be short of time.
Decision-maker
Why we ask: proposals go nowhere when the person reading them cannot say yes. Naming who decides also tells the consultant who to keep informed.
A good answer: who signs off, who runs it day to day, and who else needs to agree. In a small business, that may all be you.
Ownership, handover and exit
Why we ask: this is the question that protects you from lock-in. If the work, the prompts, the settings and your data sit only in the supplier's accounts, leaving later is hard.
A good answer: what you expect to own, what you want handed over (documents, logins, settings) and how you would move to another supplier or tool. Our red flags guide covers what to watch for.
After go-live
Why we ask: AI tools need looking after. Suppliers change their models and prices, and what worked at launch can drift. Many projects have no plan for month two.
A good answer: who fixes problems, who makes changes, who you call, and how you will check in six months that it still works as well as it did.
Questions for the consultant
Why we ask: the same questions to every consultant make the replies comparable. The ones in the template cover experience, approach, cost, people, your time, ownership, data and how you will know it worked.
A good answer: keep ours and add your own. How a consultant answers, especially "what would make you recommend stopping?", tells you a lot. See questions to ask an AI consultant for the reasoning behind each one.
Once your brief is ready
Send it to two or three consultants, give them the same deadline and compare what comes back. You can find consultants in our directory, and how to choose an AI consultant walks through the rest of the process.