What Is an AI Maturity Model?

An AI maturity model is a framework that scores your organization’s AI capability across several dimensions and places you on a staged ladder — typically four or five stages running from ad hoc experimentation to AI embedded in how the company operates. It is a benchmarking tool. It tells you where you sit relative to other organizations, and it tells you which capabilities are furthest behind.

What it does not tell you is what to do on Monday. That distinction is the whole subject of this guide, because the published models are mostly written for organizations far larger than the companies asking about them, and the gap shows up the moment you try to act on a score.

What are the stages of AI maturity?

Most models use four or five stages, and they describe the same progression under different names. Gartner’s AI maturity model groups organizations into five: Foundational (ad hoc experimentation with limited coordination), Emerging (early pilots and growing executive interest), Operational (AI embedded in select processes with defined ownership), Scaled (deployed across functions with measurable ROI), and Transformational (AI reshapes decision-making and operating models).

MIT’s Center for Information Systems Research uses four, and its version is more useful because it publishes how many companies are actually in each one. Peter Weill, Stephanie Woerner, and Ina Sebastian mapped the MIT CISR Enterprise AI Maturity Model onto a 2022 survey of 721 companies, followed by 2024 interviews with executives at nine enterprises. Stage 1 is experiment and prepare. Stage 2 is build pilots and capabilities. Stage 3 is industrialize AI throughout the enterprise. Stage 4 is AI future-ready, where AI is embedded in all decision-making and the company sells services built on its own AI capability.

Model	Stages	Dimensions scored	Built for
Gartner AI Maturity Model	5 — Foundational to Transformational	7 — strategy, data, governance, engineering, operating model, culture, AI product/value	Gartner clients; full report is a paid deliverable
MIT CISR Enterprise AI Maturity Model	4 — experiment to future-ready	Processes, technology, organizational culture	Large enterprises
CMU SEI AI Adoption Maturity Model v1.0	Assessment-based roadmap	Published June 2026	Software and defense organizations
MITRE AI Maturity Model	Staged	Multi-dimensional	Federal and public-sector programs
GSA AI Capability Maturity Model	Staged	Organizational, operational, technical	US government agencies
Cisco AI Readiness Index	4 bands — Pacesetters to Laggards	6 pillars	Enterprises, self-scored

How many companies actually reach the top stage?

Very few. In the MIT CISR data, 7% of enterprises were at stage 4, and 62% were still in the first two stages — 28% at stage 1 and 34% at stage 2, with 31% at stage 3. Cisco’s 2025 readiness index found a similar shape from a different angle: 13% of organizations fully prepared, 36% in its Chasers band, and 48% in the second-lowest band.

Those distributions are the most useful numbers on this topic, and they are usually quoted for the wrong reason. The vendor framing is that most companies are behind and should buy something. The more accurate reading is that the top stages describe a small and structurally distinct group. MIT CISR also found that companies in stages 1 and 2 performed below their industry’s financial average while those in stages 3 and 4 performed above it — a real correlation, but one that runs through company scale and capital as much as through AI competence. A 180-person distributor is not one workshop away from selling AI capability as a service.

Who were these models built for?

Four of the six most-cited AI maturity models were written for large enterprises or government agencies. Gartner’s is a client toolkit gated behind a work-email form. MITRE’s, the GSA’s AI Capability Maturity Model, and CNA’s AI Maturity Model for Government Agencies are all public-sector instruments. Carnegie Mellon’s Software Engineering Institute published its AI Adoption Maturity Model v1.0 in June 2026 for organizations running formal software engineering practices.

That provenance is not a criticism of the models. They are careful, and they do what they were designed to do: give a large organization with many business units a common scale so that the finance function and the supply chain function can be compared on the same axis. The problem is one of transfer. A mid-market company has one of most things — one operations lead, one system of record, one person who knows how the quoting process really works. Scoring it on a scale whose upper stages assume a scalable enterprise architecture, a pervasive test-and-learn culture, and proprietary models trained on internal data will return a low number every time, and the low number will be accurate and useless in equal measure.

When is an AI maturity model the wrong tool?

It is the wrong tool whenever you need a decision rather than a benchmark. If the question is “what should we build first,” a maturity score cannot answer it. The score is an average across dimensions, and averages hide the thing you needed to know — that the data is fine, the governance is irrelevant at your size, and the actual constraint is that two people re-key the same information between two systems every morning.

Here is the stance we will defend: below roughly 500 employees, a maturity ladder measures your size more than your competence. The honest way to use one at that scale is narrow. Pick the two or three dimensions you have decided to invest in, score only those, and re-score them in twelve months. That gives you a trend line, which is what a maturity model is genuinely good for. Scoring all seven pillars produces a number that will not move for years, because most of the pillars describe an organization you are not trying to become.

And a higher score does not mean the money comes back. MIT’s NANDA initiative studied more than 300 AI initiatives for its 2025 report The GenAI Divide: State of AI in Business and found that roughly 95% of organizations deploying generative AI saw no measurable impact on the P&L. Maturity measures whether a company could run AI work. It does not measure whether any particular project was worth running.

What should a mid-market company measure instead?

Measure the workflow, not the organization. The unit that pays back is a specific process with a named owner: where the hours go, who does the work today, what the handoff costs, and what would change if part of it stopped being manual. That produces a decision — start here, or fix this policy first — where a maturity score produces a position.

Our own AI Opportunity Assessment works this way deliberately. It scores readiness across five dimensions, because a readiness read is genuinely useful before you spend, but the output is a workflow and a dollar range with its assumptions written down, not a stage. Roughly half the time the recommendation involves no AI at all — a policy change, or configuring a system already being paid for. A maturity model has no way to return that answer, because “you don’t need this” is not a stage on the ladder.

Use both if you like. Just use them for what they are: the maturity model for the board slide that shows movement year over year, the workflow assessment for the decision about what to fund next quarter.

Where to go next

If you are trying to work out whether to run any of this at all, start with what an AI readiness assessment measures and the scoring method behind it. If you want to see how the free vendor tools compare before spending anything, we ranked them in AI readiness assessment tools ranked. The rest of the series sits on the AI assessment guides hub, and our own pricing is published on the pricing page.

Sources

• MIT Sloan, What’s your company’s AI maturity level?, 25 February 2025 — the four MIT CISR stages and the 28% / 34% / 31% / 7% distribution, from a 2022 survey of 721 companies and 2024 interviews at nine enterprises.
• Gartner, AI Maturity Model and AI Roadmap Toolkit — the five stages and seven pillars, read from the live page in August 2026.
• Carnegie Mellon University Software Engineering Institute, The AI Adoption Maturity Model v1.0, 30 June 2026.
• US General Services Administration, AI Capability Maturity, AI Guide for Government — the federal AI CMM.
• Cisco, AI Readiness Assessment Tool — the four bands and the 2025 global distribution.
• MIT NANDA, The GenAI Divide: State of AI in Business 2025, reported by Sheryl Estrada, Fortune, 18 August 2025.

Method note. Stage names, pillar counts, and gating were read from each model’s live page in August 2026, not from a secondary summary. The MIT CISR stage distribution is the researchers’ own reported figure from their 2022 survey; it is a survey result, not a current market measurement. Where a model is published behind a form or a client paywall, that is stated rather than inferred.