# AI Opportunity Audit

A working document, not a tools list. Use it to map where AI actually helps
across a business, one area at a time, before anyone runs a training session
or buys a subscription. Fill in what applies. If an area doesn't earn a "yes"
in section 3, leave it out — that's a finding, not a gap in the document.

---

## Before you start

List every distinct area of the business you're auditing — departments,
teams, or functions, whatever the real unit is here. Don't group anything
together yet for convenience; a team that gets folded into "operations" at
this stage usually gets skipped over later.

-
-
-
-
-

---

## The per-area worksheet

Copy this block once for each area on your list above. Filling in one
properly is worth more than filling in all of them vaguely.

### Area:

**1. What does this area actually do, day to day?**
Not the job title. The real tasks, especially the ones nobody's written
down because they didn't need to be.

-

**2. Where could AI plausibly help?**
Specific tasks, not "AI could help with admin." If you can't name a task,
you haven't found a fit yet.

-

**3. What would it need to work well?**
This is the section most audits skip, and it's the one that decides whether
the output is useful or generic. For each task above:

- **Facts/spec** it needs to be accurate:
- **Tone or standard** it needs to match:
- **A real example** of this done well already:
- **What it needs to be told to leave out or avoid:**

**4. Is this actually a fit?**
Answer honestly, using the checks in the next section. If the answer is no,
say so and move on — a clear "not yet, and here's why" is more useful to
whoever reads this than a forced use case.

- Fit / Not yet / Indirect only (delete as appropriate)
- Why:

---

## The honest no

Not every area belongs in an AI rollout, and pretending otherwise is how you
end up with a tool nobody in that role actually opens. Use these checks
before writing "fit" in section 4:

- **Is the work mostly physical?** If the task is manual, hands-on work,
  AI likely doesn't sit inside the role itself — it might still help one
  level up (forecasting, documentation), which is worth noting separately
  rather than forcing training onto people whose actual job doesn't change.
- **Is the risk of an unreviewed mistake high and immediate?** Anywhere
  output reaches a customer, a regulator, or real money with no review step,
  scope the use case narrowly (one specific task, checked) rather than
  opening the door broadly.
- **Does the context this area would need actually exist anywhere?**
  If nobody could hand over the facts, tone, and exceptions this area
  would need even if they tried, that's not an AI problem to solve yet —
  it's a documentation problem to solve first.

An area that fails all three isn't a training gap. It's an accurate finding,
and it's worth stating plainly rather than working around it to make every
department look included.

---

## Once every area's done

Look across all the worksheets for the pattern, not just the individual
entries:

- Which two or three areas have the clearest fit *and* the context to
  support it already available? Start there, not everywhere at once.
- Which areas came back "indirect only" or "not yet"? Note them honestly —
  revisiting in six months is a real plan; pretending they're covered isn't.
- Where the same missing context shows up in more than one area (a brand
  voice guide, a documented set of exceptions), that's usually worth fixing
  once, centrally, rather than solving per department.

---

*This is the same worksheet behind an AI-adoption audit run across a real,
multi-department retail business — see [how it actually played out](/thinking/ai-opportunity-audit-in-practice)
for the worked example. The gap between a generic AI rollout and a useful
one is almost always what this document forces you to write down.*