The one-page scope doc that stops AI project creep

You had a clear idea. "We'll use AI to handle customer follow-ups." Six weeks later you're debating whether the AI should also manage your inventory alerts, write your newsletter, and maybe run your loyalty program. Nothing is live yet. Sound familiar?

Scope creep on AI projects is sneaky because AI can do a lot — which makes it tempting to keep adding jobs to its plate before the first one is even working. The fix isn't willpower. It's a single sheet of paper you fill out before the project starts.

1. Write one problem statement — no conjunctions allowed

Your problem statement gets one sentence. It can't contain "and," "also," or "plus." If you catch yourself writing "AI will handle customer replies and post to Instagram," you have two projects. Pick one.

A tight problem statement acts as a filter for every decision that follows. When someone suggests adding a feature, you hold it against the statement. If it doesn't directly solve that one problem, it goes on a "later" list — not into this project.

2. Define done in three bullet points

Before you write a single prompt or pick a tool, answer: what does success look like in week four? Write it as three measurable bullets, for example:

• Customer enquiries answered within 5 minutes, 7 days a week
• Response accuracy checked and approved on 20 real conversations
• Handoff to a human triggered correctly at least 90% of the time

Three bullets forces prioritisation. If you can't describe done in three points, you don't know what you're building yet. That's okay — but figure it out now, not halfway through.

3. Name the single owner

AI projects stall when everyone has opinions but no one has final say. Put one name on the document — the person who decides when a feature is in scope, when the project is done, and when to stop. In a small business, that's usually you. Write your name anyway. It makes the role feel real and stops the "well, what if we also…" conversations from going circular.

If you're working with a contractor or a tool vendor, they need to know who the decision-maker is too. One owner, one name, no committee.

4. List what is explicitly out of scope

This is the most underused section of any project document. Write a short list of things the AI will not do in this phase. Be specific and slightly ruthless:

• Will not handle billing disputes
• Will not post to social media
• Will not integrate with the inventory system (yet)

The word "yet" is important. You're not saying never — you're saying not now. This gives your team (or your future self) permission to say no without feeling like good ideas are being killed. They're being queued.

5. Set a four-week hard stop

Open-ended AI projects drift. Give this one a fixed end date — four weeks from today is a good default for a first AI workflow. At that date you review the three success bullets from tactic two and make one of three calls:

1. Ship it. It works. Move it to business-as-usual.
2. Extend it. One specific thing needs more work. Reset the clock once.
3. Kill it. The problem turned out to be different. Document what you learned and start fresh with a better problem statement.

Four weeks is short enough to stay focused and long enough to see real results. Anything that doesn't fit inside four weeks is either a second project or a sign the scope is still too big.

---

Filling out this document takes about 20 minutes. Skipping it costs weeks. Keep it to one page — if it spills onto a second page, cut until it doesn't.

If writing and maintaining this kind of operational thinking sounds like exactly the work you never have time for, that's what a Sidekyk is for. Message your Sidekyk on WhatsApp, describe the AI project you're planning, and it'll help you draft the scope doc, track the four-week timeline, and flag the moment things start to drift — so you stay focused on running your business. Find out more at sidekyk.ai.