Most AI maturity models are theatre. They produce a tidy score, a colourful chart, and a slide for the board — and they change nothing, because they measure how AI-enthusiastic an organisation feels rather than whether it can actually deliver. A maturity assessment is only worth doing if its output is a list of things to fix.

This article describes the five-dimension model we use. It is not designed to flatter. It is designed to find, before a project starts, the specific gaps that would otherwise surface mid-build as expensive surprises.

Five dimensions that decide delivery

1. Data

The most decisive dimension. Is the data AI work needs available, accessible, well understood, and good enough? Is its quality known rather than assumed? Is there governance covering how it is used? Organisations consistently overrate themselves here, because data that is adequate for existing reports can still be far short of what an AI system needs.

2. Technology and infrastructure

Can the organisation build, deploy, and operate AI systems? This covers cloud foundations, the ability to integrate with existing systems, and the platform capability to run models reliably in production. A team can be capable of building a model and still have no dependable way to operate it — a gap that stays invisible until the pilot needs to ship.

3. Skills and talent

Does the organisation have, or can it reach, the people to deliver and sustain AI work — engineering, data, and the product judgement to aim it well? Crucially, this includes the skills to operate and maintain systems after launch, not only to build them. Many organisations can produce a pilot and cannot support what they produced.

4. Governance and risk

Is there a clear way to decide what should and should not be built, to manage risk, and to meet regulatory obligations? Without this, AI work is either reckless or paralysed. With it, teams can move quickly because the boundaries are known. Governance is an enabler of speed, not a brake on it.

5. Strategy and leadership

Is there a real strategy connecting AI work to business goals, and leadership that understands enough to support it well? Where this is weak, organisations get scattered experiments with no compounding value. Where it is strong, each project builds on the last.

A maturity assessment is only worth doing if its output is a list of things to fix.

Scoring that points at action

We rate each dimension on a simple four-level scale, and the level names are written to make the next move obvious.

  1. Absent — the capability essentially does not exist; building it is a prerequisite, not a parallel task.
  2. Emerging — early, inconsistent capability; it exists in places but cannot yet be relied on.
  3. Established — a solid, dependable capability that supports real AI work.
  4. Advanced — a genuine strength the organisation can build on and lead with.

The single number some models love is the least useful output. What matters is the shape across the five dimensions, because that shape tells you what to do.

Reading the shape

A profile that is Established or better across all five means the constraint is choosing the right use case, not building capability — move to delivery. A profile strong everywhere except one Absent or Emerging dimension means that weak dimension is the whole plan: fix it before committing to AI delivery, because it will otherwise stall the project regardless of how strong everything else is.

An evenly low profile is, oddly, the clearest result of all. It says the honest first step is foundation-building, and that any AI project started now would mostly serve to prove the foundations are missing — at considerable cost.

Used this way, a maturity model is not a scorecard. It is a diagnosis, and its value is entirely in the work it tells you to do before you begin.