Prove · How I Work
Proof-or-pause: why your AI consultant’s contract should have an exit clause
April 2026 · Jason Lee
Here is a test you can run on any AI consulting proposal in under a minute: find the paragraph that describes how the engagement ends.
In most proposals it doesn’t exist. The engagement renews by default, scope expands by “alignment,” and termination requires you to initiate an awkward conversation with someone who now attends your leadership meetings. That structure isn’t an accident. It is the business model: consulting revenue compounds through inertia, and inertia is engineered by making “continue” the path of least resistance.
In a regulated business buying AI help, that structure works directly against you, for a specific reason: AI initiatives are unusually easy to keep alive without results. The work is novel, the metrics are unfamiliar, and “these things take time” is always plausible. An open-ended engagement plus an unfalsifiable initiative is how organizations spend eighteen months and six figures with nothing in production — the exact pattern the stalled-pilot statistics describe.
The clause
My contracts contain the opposite structure, and I publish it because I want it compared:
Every engagement runs a fixed 90-day cycle. At day 90, you receive a CFO-ready ROI report — hours recovered, dollars saved, risks retired, against baselines set in week one. Leadership then makes an explicit decision: scale, continue, or stop. No auto-renewal. If the numbers don’t justify continuing, we stop — and you keep everything built: code, documentation, policies, prompts.
That’s the proof-or-pause clause. Four properties worth noticing:
- The decision is forced, not available. Anyone can technically cancel a consultant. The clause makes not deciding impossible — the default outcome is a decision meeting, not a renewal invoice.
- The evidence is contractual. The ROI report isn’t a courtesy deliverable; it is the input the decision clause requires. Which means baselines must exist, which means they get built in week one, which quietly fixes the number-one reason pilots stall.
- You keep the assets. Ownership transfer removes the hostage dynamic where firing the consultant means losing the work. The engagement is designed to make you capable, not dependent.
- The incentive points the right way. My only mechanism for revenue past day 90 is a report your CFO believes. I cannot bill through inertia; I can only bill through results. That is precisely the incentive you want in the person choosing which of your workflows to automate.
“Doesn’t this just mean you’re confident?”
Partly. But confidence is cheap to claim; the clause makes it expensive to fake. A consultant who won’t structure the engagement this way is telling you something about their expected results — listen. And a buyer should understand what the clause costs me: I carry real revenue risk every quarter, which is why the engagement is fixed-fee, priced for senior work, and limited to two concurrent clients. The clause and the price are the same decision viewed from two sides.
Questions to ask any AI consultant, this one included
- What happens at the end of the engagement, and what forces it to happen?
- What baseline will the results be measured against, and who sets it?
- If we stop, what do we keep?
- What does your renewal depend on — my inertia, or my numbers?
If the answers are vague, the proposal has told you how it ends: slowly, expensively, and without a production system.