← Back to thinking
Thinking

The AI opportunity audit, in practice

A worked example of the AI Opportunity Audit template, run across a real, multi-department retail business this week.

3 of 9 areas that got a clear yes on the first pass

The situation

A business I work with asked everyone to show how they're using AI, and to start thinking about training the wider team. Not an AI vendor pitching a platform. An actual internal ask, this week, aimed at a business with around nine distinct departments: web development, design, UX, sales and customer service, buying, marketing, warehouse operations, a production team that personalises physical products to order, and bookkeeping.

The instinctive response to a brief like that is a tools list. Here's ChatGPT, here's how to write a prompt, off you go. I used the AI Opportunity Audit template instead, partly because it's what I'd tell a client to do, and it felt dishonest not to use it on my own doorstep first.

What the audit actually asks

Not "where could AI help." Almost anywhere, if you're vague enough about it. The template forces three sharper questions per area: what does this area actually do day to day, what specific task could AI plausibly touch, and what would it need to be given to do that task well rather than generically. That third question is the one a tools list skips entirely, and it's the one that decides whether the output is usable or just plausible-sounding.

Where it was an easy yes

Customer service was the clearest fit, and the reason why is worth stating plainly: the context it needs already exists. A customer's history, what's already been promised, the tone the team uses when owning a mistake, all of it is real and sitting somewhere, even if nobody had written it down as "the material an AI reply would need." Marketing and product copy were similar: strong existing examples of what good sounds like, a real brand voice, just never assembled into something a model could actually use.

Those became the two areas to start with, not because the tasks were simpler, but because the context was already closest to ready.

Where the honest answer was no, or not yet

Warehouse operations was the clearest "no." The work is physical: picking, packing, checking stock against a system. There's no AI tool that sits inside that role in any way that isn't contrived. Where AI genuinely touches that part of the business is one level up, in demand forecasting, which is a different department's problem to solve, not a training need for the warehouse floor.

The production team that personalises physical orders to order was a narrower, more interesting case. Not a training track, but one specific, useful task: checking a customer's personalisation text against the artwork before anything gets printed, which is exactly the kind of narrow, high- value, low-risk use case that gets missed when the exercise starts from "who needs AI training" instead of "where does a mistake actually happen here."

Bookkeeping got a deliberately narrow yes: drafting routine correspondence and turning figures into plain language for a query to the external accountant, with an explicit line that anything touching the figures themselves stays with a person. Writing that boundary down mattered more than the use case itself.

The pattern underneath it

The same missing piece showed up in more than one area independently: a written brand voice guide, and a plain record of the exceptions and judgement calls that currently only live in a few people's heads. Neither is an AI problem. Both are the actual blockers to nearly everything else on the list working well, in customer service, marketing, and buying alike. That's usually what an audit like this actually finds: not nine separate training needs, but one or two shared gaps wearing nine different costumes.

What this replaces

Not a training deck. A short, honest map of where to start, where to wait, and where the answer is genuinely no for now. That's a more useful thing to hand a business than an enthusiastic rollout plan, and it took an afternoon, not a consultancy engagement, because the template did the actual thinking in advance.

If you're facing a version of this brief yourself, the AI Opportunity Audit template is the same worksheet, blank, ready to run against your own business.

---
date: 2026-09-15
updated: 2026-10-02
title: The AI opportunity audit, in practice
slug: ai-opportunity-audit-in-practice
summary: A worked example of the AI Opportunity Audit template, run across a real,
  multi-department retail business this week.
stat: 3 of 9
stat_label: areas that got a clear yes on the first pass
---

If this sounds like your business, the audit is the quickest way to find out what your AI is missing: one workflow, one 90-minute session, and a written brief on what to fix first.

Book the audit

Prefer to talk first? Get in touch — no obligation.