Why your AI pilot does not scale: pilot exit checklist
For pilot owners stuck in "pilot" for more than a quarter who want to know what is missing before asking for budget to scale.
When to use it
The pilot went well, everyone applauded, and six months later it is still a trial.
You are about to ask for budget to scale and do not know what you will be asked.
You are asked to approve a new pilot and want to request in writing what the last one lacked.
The sheet
Seven conditions; the first three are the ones that, in what I have seen, are almost always missing. A "No" means the pilot is not ready to ask for money, but ready to close that gap.
#
Condition
Control question
Evidence that meets it
Met
1
Owner in the business
Who in the area that will use it answers for the result? Not someone in technology.
A name and a role from the area.
Yes / No
2
Number it was meant to improve
Which metric was defined before starting, and how much did it move?
Dated baseline and pilot measurement with the same scope.
Yes / No
3
Money for the day after
What does it cost to connect systems, train people and give support, and who budgeted it?
Approved or requested budget line, with an amount.
Yes / No
4
Process redesigned
Did the handoffs between people change, or was a tool just placed on top?
Process map before and after.
Yes / No
5
Data access
Does the tool reach real data with approved permissions, or did it work on a prepared sample?
Permissions and connections approved by IT and legal.
Yes / No
6
Review load measured
How many hours go into reviewing what the AI produces, and who pays for them?
Review hours per week, measured, not estimated.
Yes / No
7
Exit plan
If it is switched off, which data is recovered, which contract is closed and who takes the work back?
One page with steps and an owner.
Yes / No
Score: ___ of 7 · If 1, 2 or 3 is missing, do not ask for budget yet · Date: ________ · Owner: ____________
How to fill it in 10 minutes
Fill it with the pilot owner and someone from finance. Ten minutes, no slides.
Start with condition 1. If the only name that appears is from IT, you already know why it did not scale.
For condition 2, look for the baseline. If the metric was defined after the pilot, mark "No" even if the number looks good.
For condition 3, write the cost of the day after even if it is an estimate; if nobody budgeted it, the pilot stays where it is.
The most common mistake: using this sheet only at the end. Ask for the first three conditions in writing before approving the pilot; at the end you only verify them.
The line that is almost always missing
Condition 3. A pilot is approved on the cost of the trial and stays a pilot because of the cost of what comes next: connecting systems, training people and giving support. Nobody budgeted it because nobody asked. Without that, the pilot is not a project; it is an expensive experiment.
Vic Sol · Decision Desk · vicsol.org/resources/ · Free to use with attribution. This sheet structures a conversation; it is not legal, financial or audit advice.
Open the .docx in Word (or the .xlsx in Excel) or print it before the meeting.
Fill the fields with the team; leave the blanks visible, they are the discussion.
Close with one owner, one date and one next action. Nothing else.
Prompt
Generate this sheet for your case
Paste this prompt into ChatGPT, Claude or Gemini together with your proposal, process or policy. It reproduces the sheet with your data and flags what is missing.
Act as an operations director who has watched several AI pilots stay pilots and who, before approving budget to scale, asks in writing who the owner is, which number improved and what the day after costs.
Context I am giving you:
- Pilot (what it does, in which area, since when): [ ]
- Who led it and who uses it today: [ ]
- Metric defined before starting, if any, and the result: [ ]
- Which systems it touches and which data it worked on (real or a prepared sample): [ ]
- Review hours it requires, if known: [ ]
- Pilot cost and estimated cost to scale, if any: [ ]
- Pilot documents: [paste report, presentation or notes]
What I need: assess the pilot against the seven exit conditions, in this order: owner in the business, number it was meant to improve, money for the day after, process redesigned, data access, review load measured, exit plan. For each one:
1. Met: Yes / No / Unknown.
2. Evidence you found in what I gave you, or its absence.
3. What it would take to move it to "Yes", with a suggested owner by role.
Close with a verdict: ready to ask for budget / close gaps first / close the pilot, and the reason in two lines.
Output format: a table Condition · Met · Evidence · What is missing, and the verdict. Maximum 400 words. No introduction.
Rules:
- If the metric was defined after the pilot, condition 2 is "No" even if the number is good.
- If the only owner is in technology, condition 1 is "No".
- Do not invent evidence. Where it is missing, write [DATA PENDING] and tell me how to get it.
- If critical information is missing, ask me at most 3 questions before answering.
- Answer in plain, direct business English, with no sales adjectives.
What to check in the answer:
Did it mark "Yes" on condition 1 with an IT name or with "the innovation team"?
Does the money for the day after have an amount, even an estimate, or was it left as "to be defined"?
Is the verdict consistent? With 1, 2 or 3 as "No", it should not say "ready to ask for budget".
Vic Sol