Most arguments for running an AI audit before you build anything are theoretical. Here’s one that isn’t.
AI consultant Liam Ottley, who runs the AI automation agency Morningside AI, has documented one of his firm’s audits publicly: an eight-week engagement with a mid-market manufacturer of a few hundred employees. His team interviewed staff across the business, mapped how work actually moved through it, and came back with 15 separate automation opportunities, each scored and costed individually.
The headline numbers, by his own account: an estimated implementation cost of roughly $250,000 USD across the recommended builds, against a projected $700,000 in savings once those systems replaced the manual work they targeted. Not every recommendation needed building immediately. The report ranked them, so the client could fund the highest-value ones first and treat the rest as a roadmap.
Why the numbers work out this way
The math isn’t complicated, and that’s the point. Once you know how many hours a task costs someone monthly, and what that person’s time is worth, the savings calculation is arithmetic. What’s hard, and what most businesses skip, is the measuring: actually sitting down with the people doing the work and finding out what they do, how often, and how long it takes.
Skip that step and you’re left guessing at both sides of the equation. You don’t know if a given task genuinely costs $50 a month or $5,000. You don’t know if a $30,000 build pays for itself in six weeks or six years. An audit exists to replace that guess with a number you can defend.
What this looks like at SME scale
A few hundred employees and an eight-week engagement is a different scale than most small and mid-sized businesses need. The principle holds regardless of size: find the tasks that are genuinely expensive in hours, score how automatable they are, and cost the fix before committing to it.
At the scale this business operates, an audit runs the key decision makers on a team over ten working days rather than dozens of people over two months. The numbers are smaller, proportionally, but the shape of the finding is usually the same: a handful of tasks account for most of the wasted time, and at least one of them is cheap enough to fix immediately.
That’s usually where the AI Audit earns its keep before the bigger build even starts. A ten-day review that turns up one automation saving a team member five hours a week already pays for itself inside a couple of months, before you’ve built anything more ambitious.
The honest caveat
Fifteen opportunities and a seven-figure savings estimate came from a business with hundreds of employees and a lot of repeated, high-cost manual work to find. A smaller business with three people wearing five hats each won’t produce numbers on that scale, and it shouldn’t need to. What scales down cleanly is the method: interview the people doing the work, measure the real cost of each task, score it honestly, and only build what the numbers actually justify.
If you want that same rigor applied to your own team, without the eight-week timeline or the seven-figure price tag, the AI Audit runs ten working days and ends with a scored list, not a guess.