What an AI audit covers, and what you should have in your hands at the end

01What it looks at

• Data: what you hold, where, in what condition and who owns it.
• Systems: the ERP and tools you run, and how easily anything can connect to them.
• Process: how work is actually done, as distinct from how it is drawn.
• People: skills, appetite and how AI is already being used unofficially.
• Governance: rules, responsibilities and who decides.
• Market: what competitors and customers are doing that changes your position.

02How it is done

1. Interviews with leaders and with the people who do the work, each asked the same questions.
2. A look at the data itself, not only at descriptions of it.
3. A walk through two or three real processes from start to finish.
4. A read-out in which the findings are challenged before they are final.

The people closest to the work know where the problems are. An audit that talks only to leadership finds only what leadership already believes.

03The four things you should receive

• A readiness assessment: where data, systems, process and people stand today.
• A threat assessment: which parts of the business AI puts under pressure, and how soon.
• An opportunity map: where AI would save or earn the most, with the effort each needs.
• A roadmap: what to do first, next and later, with owners and measures.

04What a good audit does not do

• Recommend only the auditor's own products
• Hide behind a maturity score with no actions attached
• List a hundred use cases with no order
• Avoid saying what should not be done

05How to prepare

• Name a sponsor on the leadership team.
• Give access to the people who run the processes, and tell them to be candid.
• Share real data extracts, under a confidentiality agreement.
• Say which decisions the audit must help you make.

Questions