The first thing most organisations want to know when they engage an AI consultant is: which tools should we be using? It's a reasonable question. It's also almost never the right one to start with.
Before recommending a single tool, the first stage of every engagement I run is an AI audit. Not a theoretical framework audit. A practical, operational investigation into how AI is actually being used — or not used — inside the organisation right now.
WHAT AN AI AUDIT IS NOT
It is not a survey asking people whether they've heard of ChatGPT. It is not a review of which software licences you currently hold. It is not a benchmarking exercise against industry averages.
Those things produce data. An AI audit produces understanding.
The most important discovery in most AI audits is what people are already doing — without telling anyone.
WHAT IT ACTUALLY IS
An AI audit starts with conversations — structured, confidential conversations with people across the organisation at every level. What tools are people actually using day to day? What are they using them for? Where are they getting good results and where are they frustrated? What are they avoiding, and why?
In almost every audit I've conducted, the same pattern emerges. There is a layer of informal, undisclosed AI usage happening across the organisation — people using tools they found themselves, for tasks they've quietly integrated into their workflow, without any governance or oversight. Alongside that there is usually a layer of people who are curious but uncertain, waiting for permission or guidance that has never arrived.
The audit maps both. It identifies where AI is creating genuine operational value, where it's being misused or creating risk, and where significant opportunity is being left on the table.
WHAT COMES OUT OF IT
The output of an AI audit is not a list of tool recommendations. It's a clear picture of the organisation's current AI reality — and a prioritised view of where the highest-value interventions are.
That picture is what makes everything else possible. The integration work, the governance framework, the strategic roadmap — all of it is built on understanding what's actually happening, not what the organisation thinks is happening.
It typically takes about a week. It is, consistently, the most revealing week of the entire engagement.
Found this useful? Mark works with organisations to turn AI thinking into operational capability — through the Production-First AI System™.
MARK HORTON
Strategic AI Advisor · Organisational AI Capability