Measure and budget

Almost nobody compares cost against value: the quarterly review

Finance, operations or IT with more than three live AI tools.

Editable template 15 min to use From video 15 · El costo real de la IA Updated 2026-10-19

Vic Sol · Decision DeskEditable template · 15 min to use

Almost nobody compares cost against value: the quarterly review

For finance, operations or IT teams with more than three live AI tools, where nobody has reviewed, case by case, what each one delivered.

When to use it

A licence renewal arrives and nobody remembers which number it was supposed to improve.
AI spend grows every quarter and the committee only sees the total.
You are asked for "more AI budget" and do not know which of the current uses works.

The sheet

One row per use case, not per tool: the same licence can hold three use cases with three verdicts.

Use case	Owner	Monthly cost (licence + usage + review hours)	Metric it was meant to move	Evidence (data point, date, source)	Verdict
[What it does and for whom]	[From the business area, not IT]	[Three items; the third, estimated]	[A business number, not "productivity"]	[Before and after, with a date]	Keep / Adjust / Stop
Illustrative: support ticket summaries	Head of support	Licence + usage + 6 h/month of review	First-response time	From 4 h to 2.5 h over six weeks, per the ticketing system	Keep

Verdict rule. Keep: there is dated evidence that the metric moved. Adjust: the metric exists but was never measured, or review hours eat the benefit. Stop: after a full quarter it moves no number at all.

Date of this review: ____________ · Next review: ____________ · Convened by: ____________

How to fill it in 15 minutes

List use cases, not invoices: one row for each thing an area does with the tool.
Fill the cost with three items: licence, usage and the hours of whoever reviews what comes out. The third is almost never in the budget; estimate it and label it.
Write the metric it was meant to move. If the answer is "productivity" or "efficiency", it is not a metric yet.
Look for the evidence: one dated data point is worth more than ten opinions. If there is none, leave the cell blank; the blank is information.
The most common mistake: giving "keep" to everything that is used a lot. Usage is not value. A popular use case that moves no number is a popular expense.
Set the date of the next review before you close; the sheet only works if it is repeated.

The line that is almost always missing

The evidence column. Companies know what each tool costs and even have a metric in mind; what is missing is the dated data point showing it moved because of the tool. Without that column, the verdict is an opinion formatted as a table.

What the study says

Fact: KPMG published its Global AI Pulse survey for the third quarter on 24 September 2026, covering 2,131 senior leaders in 20 countries. 61% review AI costs when approving a project and 59% monitor them in operation. Only 12% consistently assess AI value against cost. The sample is large enterprises; its investment figures do not apply to a small or mid-sized company.

My reading: reviewing cost at approval is easy. Comparing cost against value every quarter is uncomfortable, because it forces you to switch things off. It is like approving a team's budget and never checking what it delivered.

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.

Almost nobody compares cost against value: the quarterly review

For finance, operations or IT teams with more than three live AI tools, where nobody has reviewed, case by case, what each one delivered.

When to use it

- A licence renewal arrives and nobody remembers which number it was supposed to improve.
- AI spend grows every quarter and the committee only sees the total.
- You are asked for "more AI budget" and do not know which of the current uses works.

The sheet

One row per use case, not per tool: the same licence can hold three use cases with three verdicts.

 Use case  ·  Owner  ·  Monthly cost (licence + usage + review hours)  ·  Metric it was meant to move  ·  Evidence (data point, date, source)  ·  Verdict 

 [What it does and for whom]  ·  [From the business area, not IT]  ·  [Three items; the third, estimated]  ·  [A business number, not "productivity"]  ·  [Before and after, with a date]  ·  Keep / Adjust / Stop 
 Illustrative: support ticket summaries  ·  Head of support  ·  Licence + usage + 6 h/month of review  ·  First-response time  ·  From 4 h to 2.5 h over six weeks, per the ticketing system  ·  Keep  ·    ·    ·    ·    ·  Verdict rule. Keep: there is dated evidence that the metric moved. Adjust: the metric exists but was never measured, or review hours eat the benefit. Stop: after a full quarter it moves no number at all.

Date of this review: ____________ · Next review: ____________ · Convened by: ____________

How to fill it in 15 minutes

1. List use cases, not invoices: one row for each thing an area does with the tool.
2. Fill the cost with three items: licence, usage and the hours of whoever reviews what comes out. The third is almost never in the budget; estimate it and label it.
3. Write the metric it was meant to move. If the answer is "productivity" or "efficiency", it is not a metric yet.
4. Look for the evidence: one dated data point is worth more than ten opinions. If there is none, leave the cell blank; the blank is information.
5. The most common mistake: giving "keep" to everything that is used a lot. Usage is not value. A popular use case that moves no number is a popular expense.
6. Set the date of the next review before you close; the sheet only works if it is repeated.

The line that is almost always missing

The evidence column. Companies know what each tool costs and even have a metric in mind; what is missing is the dated data point showing it moved because of the tool. Without that column, the verdict is an opinion formatted as a table.

What the study says

Fact: KPMG published its Global AI Pulse survey for the third quarter on 24 September 2026, covering 2,131 senior leaders in 20 countries. 61% review AI costs when approving a project and 59% monitor them in operation. Only 12% consistently assess AI value against cost. The sample is large enterprises; its investment figures do not apply to a small or mid-sized company.

My reading: reviewing cost at approval is easy. Comparing cost against value every quarter is uncomfortable, because it forces you to switch things off. It is like approving a team's budget and never checking what it delivered.

If this becomes a big decision

If a verdict means stopping a contract that is hard to reverse and it carries your name, 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 a financial controller with investment-committee experience who reviews every quarter what each technology expense delivered, and who accepts neither "productivity" as a metric nor "it is used a lot" as evidence.

Context I am giving you:
- Company and sector: [ ]
- AI tools we pay for and their approximate monthly cost: [ ]
- Use cases by area (what each area does with each tool): [ ]
- Business metrics we already track in those areas: [ ]
- Internal measurements or reports on each tool's effect: [paste or write "none"]
- Hours people spend reviewing what the AI produces, if known: [ ]

What I need: build the quarterly cost-against-value review table, with one row per use case (not per tool) and these columns in this order:
1. Use case.
2. Proposed owner in the business area.
3. Monthly cost split into licence, usage and review hours.
4. Business metric it was meant to move.
5. Available evidence, with date and internal source, or a blank cell.
6. Verdict: keep, adjust or stop, with a one-line justification.
After the table, list the use cases with no evidence and tell me which data point would need to be measured for 30 days to reach a verdict.

Output format: one table and one short list. Maximum 400 words. No introduction.

Rules:
- Do not invent figures or metrics. Where data is missing, write [DATA PENDING] and tell me how to get it.
- If the only evidence is that the tool is used a lot, the verdict cannot be "keep"; mark it "adjust" and explain what is missing.
- Label every review-hours estimate as an estimate.
- 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 separate the use cases, or leave one row per invoice?
Does the evidence column hold dated data, or opinions written to look like data?
Does any "keep" verdict rest only on the tool being popular? If so, ask it to re-evaluate.

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

Almost nobody compares cost against value: the quarterly review

When to use it

The sheet

How to fill it in 15 minutes

The line that is almost always missing

What the study says

Generate this sheet for your case

Read the decision note on this topic

This resource is a step in

Related in the same situation

The AI changes worth your attention.

If this becomes a big decision