The Spec · AI Project Brief Template v1.0
The brief that should exist before any code is written.
A fillable brief used by Jacarenda Labs to scope every AI engagement. Fill it out before you talk to a vendor: the quality of the brief predicts the quality of the build. Save a copy as PDF when you’re done.
What problem is this solving, and what happens if we don’t?
One paragraph. Use plain language. If the problem can’t be stated in one paragraph it isn’t scoped yet.
The workflow we’re changing.
Describe the existing workflow step by step. Where does it start? What hands does it pass through? Where does it end?
Acceptance criteria.
What does “working” look like? Be concrete and testable. Numbers, examples, and edge cases beat adjectives.
Data in, data out.
Where does the data live today? Where does the output need to land? Name the systems.
Who’s involved?
Sponsor signs the cheque. Owner runs the workflow. Reviewer tells us if the output is right or wrong. End-user lives with it.
What this is, and what this isn’t.
Naming what’s out of scope is half the brief. Be explicit about edge cases, future phases, and adjacent workflows you are deliberately not addressing.
What could go wrong, and what’s the blast radius?
If the AI is wrong 1 in 100 times, who notices and what does it cost? Knowing the failure cost determines how much human review the workflow needs.
Before we build.
The same 10 statements that drive the workflow audit. Tick each that is true today. The more ticks, the faster we can move.
- We have at least one workflow that eats hours a week of manual work.
- That workflow has a predictable trigger: it happens at known moments.
- We could write down the steps someone takes to complete this workflow.
- The people doing this workflow could describe what “done” looks like.
- The data this workflow needs already exists somewhere in our systems.
- There is a person who would know if the AI output was wrong or right.
- We have a budget holder who can decide on a fixed-scope investment.
- A short delivery window is fast enough to matter to our business.
- We are willing to hand a draft output to a human reviewer before it goes live.
- We can imagine measuring whether this workflow actually saved time after launch.
How fast, and at what cost?
Realistic ship date. Budget envelope. Trade-offs you’re willing to make.