How long does an audit take?

Usually weeks rather than months. Long enough to observe the work through a normal cycle including a bad week, short enough that the findings are still true when they are delivered.

What do we get at the end?

A ranked shortlist with the reasoning, an explicit list of what not to automate, and the measure each item should move. If the deliverable is a slide deck of AI trends, that was not an audit.

Can we do this ourselves?

Partly, and it is worth trying — mapping one process honestly with the person who runs it will teach you a lot. What outside help adds is having seen which of those patterns turn out to be worth automating and which look promising and are not.

Glossary

What is AI audit?

An AI audit is a structured review of how a business actually works — where hours, errors and delays accumulate — that ends in a prioritised list of what to automate first, what to fix manually, and what to leave alone.

The word suggests an assessment of your AI. For most businesses it is the opposite: an assessment of your processes, from which the AI question follows.

What it examines

Not your technology stack. Your work. Where a request enters the business, every hand it passes through, where it waits, where it gets re-typed, and where it goes wrong often enough that someone has built a habit around checking.

The most valuable findings are usually unglamorous. A step that exists because of a system retired three years ago. An approval nobody can explain. A report produced weekly that nobody opens.

Why it has to be observed, not described

Ask how a process works and you get the official version. Watch it and you get the real one, which is reliably different — the documented path plus three workarounds people invented and never mentioned because they no longer notice them.

Those workarounds are the audit's best material. They mark exactly where the process fails, and they were discovered by the people who do the work.

What a good audit produces

A shortlist, ranked by volume against effort, not by how interesting each item is
An explicit "not this" list, with the reasoning — usually more useful than the shortlist
The measure each item should move, chosen from numbers you already track
Steps that should be deleted rather than automated, which is the cheapest improvement available

The failure mode to avoid

An audit that recommends everything. A list of forty opportunities is not a plan; it is a way of avoiding the responsibility of choosing. The value is in the ranking and the exclusions, because those are the parts that required judgement.

Questions

How long does an audit take?: Usually weeks rather than months. Long enough to observe the work through a normal cycle including a bad week, short enough that the findings are still true when they are delivered.
What do we get at the end?: A ranked shortlist with the reasoning, an explicit list of what not to automate, and the measure each item should move. If the deliverable is a slide deck of AI trends, that was not an audit.
Can we do this ourselves?: Partly, and it is worth trying — mapping one process honestly with the person who runs it will teach you a lot. What outside help adds is having seen which of those patterns turn out to be worth automating and which look promising and are not.

Wondering whether this applies to your business? That is a thirty-minute call, not a research project.

Book a Consultation Call