How to Write an AI Project Brief That Gets You an Accurate Quote
If you send AI vendors a one-line brief — 'we want an AI chatbot' — you will get back wildly different quotes, because each vendor is guessing at a different project. A good brief is not long; it is specific on seven points. Specify those, and the quotes you get back become comparable, accurate, and fast. Here is what to include, and a template you can copy.
The 7 things a brief must specify
- The outcome — the measurable business result, not the technology. 'Cut first-response time from 4h to under 10 min', not 'use AI'.
- The use case — one bounded job: what goes in, what comes out, who uses it.
- The data — what data the system will use, where it lives (CRM, docs, database), roughly how much, and who can grant access.
- The integrations — which systems it must read from or write to, and whether those have APIs.
- Constraints — compliance (GDPR, DORA, EU AI Act), data residency, latency, and any hard 'must not' rules.
- Success criteria — the metric and target that define 'done', and how it will be measured.
- Budget range and timeline — even a broad range. It lets the vendor scope to your reality instead of guessing.
A copy-ready template
- Outcome we want: …
- Use case (in / out / users): …
- Data sources and rough volume: …
- Systems to integrate with: …
- Compliance and constraints: …
- Success metric and target: …
- Budget range and ideal go-live: …
That is the whole brief. Half a page, seven answers. It does not need a 20-page RFP — those mostly add procurement overhead without adding the clarity that actually matters.
Why the vague brief costs you money
A vague brief produces one of two bad outcomes. Either every vendor quotes a different imagined project and you cannot compare them — so you pick on price, which selects for the vendor who understood the least. Or the vendor quotes low to win, then change-requests the real scope back in once the project starts. A specific brief removes both: the vendor scopes the actual project, the quote reflects it, and the fixed price holds.
“The brief is the cheapest lever in the whole project. Ten extra minutes specifying the outcome, the data, and the success metric saves weeks of mis-scoped work — and tells you immediately which vendors actually read it.”
The bottom line
A good AI project brief is short and specific: one outcome, one use case, the data, the integrations, the constraints, the success metric, the budget range. It is also a vendor test — a serious partner will respond to a specific brief with a discovery call and sharper questions, not an instant quote. If you cannot yet fill in all seven lines, that is useful information too: it usually means the project needs a discovery workshop before it needs a quote.