Vic Sol · Decision DeskOne-page sheet · 5 min to use
Not everything is yes or no: the trust scale by task type
For anyone debating "do we let AI do this?" with their team and feeling there are only two answers.
When to use it
The committee argues approve everything or reject everything, and nobody proposes the middle.
A task has been "in pilot" for months and nobody knows what would expand it.
Someone said no out of fear, or yes to everything, and the conversation needs order.
The sheet
Deciding on AI is not just yes or no. Many of the best decisions are a third answer: yes, but changing something (scope, data or contract). This scale turns that "something" into five levels of autonomy; the last column is the question that moves a task up one level.
Level
The answer
What it means in operations
Illustrative task
Question that moves it up a level
1
No, for now: it stays with a person
AI does not take part; reviewed in six months.
Deciding a dismissal; signing a contract.
What would have to be true to try it in one small area?
2
Yes, changing the scope: one area, not the whole company
Pilot in one team; one person reviews all output.
First draft of client replies in one team.
Is the data it uses clean and approved?
3
Yes, changing the data: sort the information first
Clean, permitted data; full review.
Summarising contracts for a person to review.
Is there an easy exit, and a number to track?
4
Yes, changing the contract: an easy exit
More areas; sample review; can be switched off cheaply.
Classifying support tickets and proposing a reply.
Three months of sample review, no incidents and one number that improved?
5
Yes, in full: it operates with an owner
AI executes; a named person answers for the result; quarterly review.
Transcribing and summarising internal meetings.
No level 6: does it still have an owner and a review date?
Reading rule: a task moves up one level at a time, only when someone answers the question with evidence, not enthusiasm.
How to fill it in 5 minutes
List the tasks in the area where AI is already used or someone wants to use it.
Place each one on a level using the "what it means in operations" column, not "the answer".
Answer each level's question with what exists today: clean data, easy exit, a number; yes or no.
Check the tasks at level 1 or 5: do they belong there, or did nobody want to discuss the middle?
The most common mistake: approving a task straight to level 5 because "the tool already does it". The level is set by your operation, not the vendor.
The line that is almost always missing
The last column. A committee places tasks on levels and leaves them there forever. The question that moves a task up turns the scale into a decision that advances: it protects you from saying no out of fear and yes to everything. If a task cannot answer it, it stays where it is; that is a decision too.
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 a chief operating officer with 15 years of experience who presents AI decisions to a demanding committee and knows that many of the best decisions are neither yes nor no, but "yes, but changing something": the scope, the data or the contract.
Context I am giving you:
- Company and area: [ ]
- Tasks where AI is already used or someone wants to use it (one per line): [ ]
- For each task, what exists today: clean and approved data [yes/no/not sure], easy exit from the vendor [yes/no/not sure], a number being tracked [which one or "none"]
- The committee's last AI decision and how it ended (yes, no, or yes with changes): [ ]
What I need: place each task on a five-level autonomy scale and deliver a table with exactly these columns: Task · Level (1 to 5) · Why that level · What would change to move up one · Question that moves it up a level.
The levels are: 1 No, for now (stays with a person) · 2 Yes, changing the scope (one area, full review) · 3 Yes, changing the data (data sorted first, full review) · 4 Yes, changing the contract (easy exit, sample review) · 5 Yes, in full (operates with a named owner and quarterly review).
Output format: one table, maximum 350 words, followed by one line for each task placed at level 1 or 5 explaining whether it really belongs there. No introduction.
Rules:
- The level is set by what exists in the operation today, not by what the tool promises.
- A task moves up only one level at a time; do not propose jumps.
- Do not invent data; where something is missing, write [DATA PENDING] and tell me how to get it.
- Flag every assumption as an assumption.
- 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:
Does every task have a concrete question for moving up, or just a description of the level?
Did it place tasks at level 5 without a named owner? If so, move them to 4 and ask it to explain what is missing.
Did it separate what is verified from what is assumed in the "why that level" column?
Vic Sol