← All articles
AI Strategy6 min read

What is an AI readiness audit? (And what it actually finds)

An AI readiness audit checks your data, workflows, systems and team before you build. Here is what it examines, what it finds, and what to fix first.

What is an AI readiness audit? (And what it actually finds)

An AI readiness audit is a structured assessment of your data, processes, systems, people, and risks before you invest in AI. It finds where automation will actually pay, where data is too messy to trust, and which ideas should wait until foundations improve.

What does an AI readiness audit examine?

A proper audit looks at five areas, in this order:

  • Processes — which workflows are repetitive, rule-based, and costly enough to justify automation
  • Data — where your information lives, how clean it is, and who can access it
  • Systems — which tools you run, whether they connect through APIs, and what integration would take
  • People — who would use the AI, who would own it, and what skills exist in-house
  • Risk — privacy, compliance, security, and the cost of a wrong output in each workflow

The sequence matters. We regularly see companies start with the tool they want to buy and work backwards. An audit starts with the work itself, because the right tool for a broken process is usually no tool at all.

What does an audit actually find?

In our experience, audits surface the same categories of finding again and again. The first is data that exists but cannot be used: duplicate records, free-text fields where structured data should be, and important information trapped in inboxes and spreadsheets. AI built on that data does not fail loudly; it produces confident output from unreliable input, which is worse.

The second finding is that the best automation candidate is rarely the glamorous one. Teams arrive wanting a customer-facing agent and leave with a plan to automate document processing or internal reporting first, because that is where the volume, the clear rules, and the measurable savings actually are.

The third finding is ownership. Many organisations have enthusiasm for AI and nobody accountable for it after launch. An audit names the owner, the reviewer, and the person who decides when an output is good enough — before any build begins.

What should the audit deliver?

Insist on a written output you can act on without the auditor in the room. At minimum:

1. A ranked list of use cases, scored by value, effort, and risk 2. A plain statement of the data and integration gaps blocking each one 3. A recommended first project with a defined scope and success measure 4. A list of ideas to defer, with the reason for each deferral 5. The governance basics: ownership, human review points, and data-handling rules

If the deliverable is a slide deck of trends and a maturity score with no ranked use cases, you have bought a presentation, not an audit.

How long does an AI readiness audit take?

For a small or mid-sized organisation, a focused audit is a matter of weeks, not quarters. The time goes into interviews with the people who do the work, inspection of the actual data and systems, and scoring the candidate use cases honestly. Audits drag when access to systems and staff is drip-fed; they move quickly when leadership gives the process a sponsor and a deadline.

When we run this work, it sits inside our wider AI consulting and build practice, so the findings translate directly into a scoped first build instead of a report that gathers dust.

Frequently asked questions

Do we need an audit if we already know what we want to build? Usually yes, in a lighter form. Knowing the destination does not confirm the data, integrations, and ownership are in place, and a short audit that validates those three things is cheaper than discovering gaps mid-build.

What is the difference between an AI readiness audit and a strategy workshop? A workshop aligns people around ideas; an audit inspects evidence. The audit checks real data quality, real system access, and real process volumes, then ranks use cases on what it finds rather than on enthusiasm.

Can we do an AI readiness audit ourselves? You can run a useful internal version if you score use cases honestly and inspect your data without optimism. The common failure is grading your own data generously, so an outside check on the top candidate use case is worth considering.

Field note: A recurring moment in our audit work: a team is certain its records are "basically clean," and then we open the actual data together and find the same customer recorded several different ways across systems. Nobody was careless — the mess accumulated quietly over years. We treat that moment as the audit earning its fee, because every AI plan built on the assumption of clean data would have inherited the problem on day one.

Related service AI products

Building something worth shipping?

We take on a small number of AI product engagements. Tell us what you are building — we reply quickly.