Operate and scale

Keep, adjust or stop: reviewing an AI use case

Anyone inheriting or reviewing AI use cases already running.

One-page sheet 10 min to use From video 20 · IA sin humo Updated 2026-10-19

Vic Sol · Decision DeskOne-page sheet · 10 min to use

Keep, adjust or stop: reviewing an AI use case

For anyone who inherits or reviews AI use cases that are already running and that nobody has evaluated calmly.

When to use it

You arrived in an area with three or four "AI pilots" and nobody knows which one is still alive.
License renewal is coming and someone has to decide what stays.
You were asked to reorganize with AI and want to map duplicated work before touching anyone.

The sheet

One row per use case. If it does not fit in one row, it is still an idea.

Use case	Monthly cost	Metric it moved	Evidence	Verdict
[What it does, for whom, since when]	[Licenses + hours of human review]	[One number, before and after]	[Where the data lives]	Keep / Adjust / Stop
Illustrative: draft replies to complaints, customer service, since March	Licenses for 6 people + 5 h/week of review	First-response time: from 2 days to 1	Ticket system report, April to September	Keep
Illustrative: automatic summaries of sales meetings	Included in the suite + 0 h of review	Nothing measured	Nobody reads them, per internal survey	Stop

The duplicated-work question (answered with the team):

What office work is repeated in my area? [Two people or two systems produce the same thing: ____________]
Where, in the operation, is it worth investing more? [A process that touches the client and waits today: ____________]

Reviewed by: ____________ · Date: ________ · Next review: ________

How to fill it in 10 minutes

List today's use cases, by name and start date. If nobody knows the date, that is already a signal.
Write the full cost: licenses plus hours of human review. The second part is almost never counted.
Ask for one metric per row, with the before and the after. "It saves us time" is not a metric.
Point to where the evidence is. If it is the opinion of whoever proposed it, write "opinion" and treat it as Adjust, not Keep.
Give the verdict. Adjust means changing scope, data or who reviews, with a date to measure again.
The most common mistake: reviewing the use cases without asking the duplicated-work question. The sheet tells you what to stop; the question tells you where to reinvest.

The line that is almost always missing

The evidence column. When a case has been running for months, everyone assumes it works because nobody turned it off. A Keep with no data behind it is inertia. And if the review touches people, neither the sheet nor the AI makes that decision: a leader does, with data and taking responsibility.

What the news says

Fact: according to Reuters, cited by Fox Business on 13 October 2026, Kroger, one of the largest supermarket chains in the United States, plans to eliminate about 3,000 corporate positions over three years and is bringing AI tools and training to more than 400,000 operations employees. The company talks about costs and duplicated work, not only AI.

My reading: the easy reading is "AI takes jobs". The useful one is that the company cuts where it believes work is repeated and invests where the customer is. If you are going to reorganize with AI, first map that repeated work with your team. Otherwise you cut blind.

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.

Keep, adjust or stop: reviewing an AI use case

For anyone who inherits or reviews AI use cases that are already running and that nobody has evaluated calmly.

When to use it

- You arrived in an area with three or four "AI pilots" and nobody knows which one is still alive.
- License renewal is coming and someone has to decide what stays.
- You were asked to reorganize with AI and want to map duplicated work before touching anyone.

The sheet

One row per use case. If it does not fit in one row, it is still an idea.

 Use case  ·  Monthly cost  ·  Metric it moved  ·  Evidence  ·  Verdict 

 [What it does, for whom, since when]  ·  [Licenses + hours of human review]  ·  [One number, before and after]  ·  [Where the data lives]  ·  Keep / Adjust / Stop 
 Illustrative: draft replies to complaints, customer service, since March  ·  Licenses for 6 people + 5 h/week of review  ·  First-response time: from 2 days to 1  ·  Ticket system report, April to September  ·  Keep 
 Illustrative: automatic summaries of sales meetings  ·  Included in the suite + 0 h of review  ·  Nothing measured  ·  Nobody reads them, per internal survey  ·  Stop 

The duplicated-work question (answered with the team):

- What office work is repeated in my area? [Two people or two systems produce the same thing: ____________]
- Where, in the operation, is it worth investing more? [A process that touches the client and waits today: ____________]

Reviewed by: ____________ · Date: ________ · Next review: ________

How to fill it in 10 minutes

1. List today's use cases, by name and start date. If nobody knows the date, that is already a signal.
2. Write the full cost: licenses plus hours of human review. The second part is almost never counted.
3. Ask for one metric per row, with the before and the after. "It saves us time" is not a metric.
4. Point to where the evidence is. If it is the opinion of whoever proposed it, write "opinion" and treat it as Adjust, not Keep.
5. Give the verdict. Adjust means changing scope, data or who reviews, with a date to measure again.
6. The most common mistake: reviewing the use cases without asking the duplicated-work question. The sheet tells you what to stop; the question tells you where to reinvest.

The line that is almost always missing

The evidence column. When a case has been running for months, everyone assumes it works because nobody turned it off. A Keep with no data behind it is inertia. And if the review touches people, neither the sheet nor the AI makes that decision: a leader does, with data and taking responsibility.

What the news says

Fact: according to Reuters, cited by Fox Business on 13 October 2026, Kroger, one of the largest supermarket chains in the United States, plans to eliminate about 3,000 corporate positions over three years and is bringing AI tools and training to more than 400,000 operations employees. The company talks about costs and duplicated work, not only AI.

My reading: the easy reading is "AI takes jobs". The useful one is that the company cuts where it believes work is repeated and invests where the customer is. If you are going to reorganize with AI, first map that repeated work with your team. Otherwise you cut blind.

If this becomes a big decision

If a row's verdict means stopping something that carries your name or reorganizing a team, the free tools at vicsol.org/decision-clarity/ help you structure it. They do not replace a formal review.

---

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.

Download the prompt (.md)

Act as an operations director with 15 years of experience who inherits areas with several AI pilots already running and has to decide, with data, which ones keep going, which get adjusted and which get stopped.

Context I am giving you:
- Company / area: [e.g. customer service, 40 people, services company]
- AI use cases that exist today, by name and since when: [ ]
- What I know about their cost (licenses, review hours): [ ]
- Metrics someone has reported, if any: [ ]
- Why I am reviewing this now: [renewal, reorganization, new role]

What I need: produce the review sheet with exactly these columns, one row per use case:
1. Use case (what it does, for whom, since when).
2. Monthly cost (licenses plus hours of human review).
3. Metric it moved (one number, before and after).
4. Evidence (where the data lives; if it is an opinion, write "opinion").
5. Verdict: Keep, Adjust or Stop, with one line of justification.

After the table, answer two questions with what I gave you:
- What office work looks duplicated in this area?
- Where, in the operation, would it be worth investing more?

Output format: one table, then the two questions, then the list of data still missing to confirm each verdict. Maximum 400 words. No introduction.

Rules:
- Do not invent figures or metrics. If data is missing, write [DATA PENDING] and tell me how to get it.
- Mark every assumption as an assumption.
- Do not propose staffing changes; a leader decides that with data, not this exercise.
- 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 row have a metric with a before and an after number, or did it fall back to "saves time"?
Does the evidence column separate data from opinion, or did it give Keep to everything that has been running for months?
Do the two duplicated-work questions get a concrete answer or only generalities? If generalities, give it more context and ask again.

If this becomes a big decision

If the decision is hard to reverse and carries your name, the free Decision Clarity tools help you structure it. They do not replace a formal review.

Open the free tools

Keep, adjust or stop: reviewing an AI use case

When to use it

The sheet

How to fill it in 10 minutes

The line that is almost always missing

What the news says

Generate this sheet for your case

This resource is a step in

Related in the same situation

The AI changes worth your attention.

If this becomes a big decision