AI maturity models: what they measure, where they mislead, and what to use instead
How the main AI maturity models work, where they help the board, and where a readiness view serves you better.
BY STEFAN MARITZ··7 MIN READ
TL;DR
- An AI maturity model places an organisation on a staged scale of AI capability, usually from early experiments to AI embedded in how the business runs. It is useful for a board conversation and less useful for deciding what to do next.
- The best-known models come from MIT CISR, which uses four stages, and Gartner, which scores seven pillars. In MIT CISR's research, organisations in the first two stages performed below their industry average financially, and those in the last two performed above it.
- Maturity scores tend to average across dimensions, which hides the weakest one, and the weakest dimension is the one that limits results.
- We prefer to score readiness: whether strategy, data and systems, people, process, and tooling are each ready for a specific next step. It produces a to-do list where a maturity score produces a label.
- Use a maturity model to set direction with the board, and a readiness view to decide the next quarter's work.
Every leadership team wants to know where it stands on AI compared with its peers. A maturity model gives a tidy answer: you are at stage two of four. The trouble starts on Monday, when nobody is quite sure what stage two means they should do.
This post explains what an AI maturity model is, compares the most widely cited ones, sets out where they help and where they mislead, and describes the readiness approach we use instead when the question is what to do next.
What is an AI maturity model?
An AI maturity model is a staged framework that describes how organisations progress in their use of AI, from early experimentation to AI being part of how decisions get made across the business. You assess yourself against the stages, find your position, and read off the capabilities the next stage requires.
The idea comes from earlier maturity models in software and process management. The appeal is the same: a shared language for a complicated topic, and a sense of direction.
Which AI maturity models are most widely used?
Two come up most often in leadership conversations, and they work quite differently.
The MIT CISR Enterprise AI Maturity Model describes four stages: experiment and prepare, build pilots and capabilities, industrialise AI across the enterprise, and become "AI future-ready". It draws on a survey of 721 companies, where 28% were in the first stage, 34% in the second, 31% in the third, and 7% in the fourth. Organisations in the first two stages had financial performance below their industry average, and those in the last two performed above it.
The Gartner AI Maturity Model scores seven pillars: AI strategy, the use-case and product portfolio, governance, engineering, data, ecosystems and operating models, and people and culture.
| Model | Structure | Best for | What to watch for |
|---|---|---|---|
| MIT CISR Enterprise AI Maturity Model | Four stages, linked to financial performance | A board-level view of overall direction | A single stage hides uneven progress across functions |
| Gartner AI Maturity Model | Seven pillars, scored separately | Larger organisations with a dedicated AI or data function | Breadth can make it heavy for a company of 200 people |
| Praxes AI Readiness Diagnostic | Five dimensions scored for readiness | Mid-sized organisations deciding where to start | It is a starting point, and it does not benchmark you against an industry |
Choose MIT CISR if you need a shared picture for the board. Choose Gartner if you have the internal capacity to work through seven pillars. Choose a readiness view if your question is what to do in the next quarter.
Where do AI maturity models help?
They help most at the top of the organisation. A board that has been reading about AI wants to know whether the company is behind, on track, or ahead, and a staged model gives that conversation a structure. The MIT CISR link between stage and financial performance is useful here, because it moves the discussion from enthusiasm to evidence.
They also give a vocabulary. When the chief executive and the head of IT both describe the company as "building pilots", they are at least describing the same thing.
Where do AI maturity models mislead?
They mislead when the score becomes the goal. Three patterns stand out.
- Averages hide the bottleneck. An organisation with strong strategy and tooling but weak data scores "medium" overall. The weak data is what will stop every project, and the average makes it look manageable.
- Stages reward activity. Moving from stage one to stage two can look like running more pilots. More pilots are not progress if none of them change how work gets done, which is the pattern behind why AI pilots fail.
- Organisations are uneven. Marketing may be at stage three while finance has not started. MIT CISR's own researchers suggest assessing the stage that describes most of your AI activity, which acknowledges that unevenness without resolving it.
An analogy from operations helps. A plant's average equipment age tells you something about its condition. It tells you nothing about which machine will fail next week, and that is the question the maintenance manager needs answered.
What is the difference between AI maturity and AI readiness?
Maturity describes how far along the organisation is overall. Readiness describes whether it can succeed at a specific next step. A company can be fairly mature and still not ready for a particular use case, because the data that use case depends on is poor or the team affected has not been brought along.
That is why we score readiness across five dimensions: strategy, data and systems, people, process, and tooling. The output is a list of what to fix first, in order. A full AI readiness assessment applies the same logic in more depth.
How to assess where your organisation stands
- Pick one stage label for the board. Use the MIT CISR stages to agree, in one meeting, the stage that describes most of your AI activity today.
- Score the five readiness dimensions separately. Ask three or four leaders to score each from one to five without conferring.
- Find the lowest score and the widest disagreement. The lowest score is your bottleneck. The widest disagreement is where leaders see the business differently, which matters just as much.
- Turn the bottleneck into one or two actions for the quarter. Name an owner and a date for each.
- Check that people's work is changing. Re-score in three months, and ask teams whether their workflows have changed. If the score moved and the work did not, the score was measuring activity. Measuring AI adoption properly keeps this honest.
Frequently asked questions
What is an AI maturity model?
An AI maturity model is a staged framework describing how organisations progress in their use of AI, from experimentation to AI embedded in decision-making. It helps leaders agree where they stand and what capabilities come next.
What are the stages of AI maturity?
The MIT CISR model uses four stages: experiment and prepare, build pilots and capabilities, industrialise AI across the enterprise, and become AI future-ready. Other models use different numbers of stages or score separate pillars instead.
How do you measure AI maturity?
Assess the organisation against the model's stages or pillars, ideally with several leaders scoring independently. Pair the result with a readiness view of each dimension, so the weakest area is visible.
What is the difference between AI maturity and AI readiness?
Maturity describes overall progress. Readiness describes whether you can succeed at a specific next step, and it points to the dimension most likely to block it.
Is a higher AI maturity score always better?
A higher score is only useful if it reflects changed work and business results. Organisations can raise a maturity score by running more pilots without changing how work gets done.
If you want a scored view of your readiness across the five dimensions, the AI Readiness Diagnostic is free and produces a report you can bring to your leadership team.
WORK WITH PRAXES
Ready to move from reading about AI to getting your organisation ready for it? Start with the diagnostic.
Take the AI Readiness Diagnostic