---
title: "One-page AI business case (with its evidence gaps)"
slug: caso-de-negocio-una-pagina
lang: en
situation: medir
format: plantilla
minutes: 15
updated: 2026-10-19
---

# One-page AI business case (with its evidence gaps)

> For anyone who has to convince a CFO with recovered minutes and would rather show the gaps before someone else finds them.

## When to use it

- A pilot went well and you are asked for "the business case" to scale it.
- The vendor gave you a savings figure and nobody has recalculated it.
- You are going to committee and know the first question will be "and how do you know that?".

## The sheet

| Block | What to write | Illustrative example |
|---|---|---|
| 1. Decision | One sentence: what is being approved, for how long and who signs. | "Extend the quoting assistant to three branches for 12 months. Signed by: sales director." |
| 2. Baseline | Volume, time per task, review and exceptions measured before, with date and sample. | "600 quotes a month, 12 min each, measured in September over 80 cases." |
| 3. Costs that go in | Licences, integration, supervision (hours of whoever reviews), training and exit (what it costs to switch off). | "Licence + CRM connection + 2 min of review per quote + training + exporting data on exit." |
| 4. Verified benefit | What was already measured in the pilot, with the same scope as the baseline. | "From 12 to 6 min per quote, 80 cases, six weeks." |
| 5. Estimated benefit | What is projected from the verified part, with the assumption written down. | "60 hours a month released if volume and adoption hold." |
| 6. Assumed benefit | What someone hopes for and nobody has measured. Kept separate. | "More quotes won by replying sooner." |
| 7. Sensitivity | What happens if the benefit is half: does the decision still hold? | "At 30 hours a month it covers the licence but does not justify the integration this year." |

## How to fill it in 15 minutes

1. Write block 1 first. If it does not fit in one sentence, it is not a decision yet; it is an idea.
2. Fill the baseline with data you already have. If there is none, write it: "no baseline" is a gap, not an embarrassment.
3. Split the benefit across blocks 4, 5 and 6; the temptation is to put everything under "verified".
4. List all five costs even if one is zero. Supervision and exit are the ones that almost never appear.
5. Run the test with half the benefit. If the decision does not hold, say so before the CFO does.
6. The most common mistake: turning released hours into money saved. A released hour is capacity; it is only a saving if a real cost falls or that capacity is redeployed to something measurable.

## The five questions the CFO will ask

1. What baseline and sample support the benefit?
2. Who reviews the outputs and handles the exceptions, and what does that time cost?
3. Which cost actually changes, or where is the released capacity redeployed?
4. What happens if adoption is lower or the benefit is half?
5. What evidence would make us stop the plan, and what does it cost to exit?

## The line that is almost always missing

Block 6. Almost every business case blends what was measured with what is hoped for and presents it as a single number. Separating the assumed benefit does not weaken the case: it makes it defensible, because the CFO will find that blend anyway.

## If this becomes a big decision

If the case commits budget that is hard to reverse and carries your name, the free tools at vicsol.org/decision-clarity/ help you map its evidence gaps. They do not replace a formal financial 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.