AI maturity model: MIT CISR and Gartner stages and how to assess a company
An AI maturity model describes the successive stages a company goes through, from first experiments to AI embedded in processes and decisions. The article compares the MIT CISR and Gartner models, shows a worked example of assessing a company across several areas, and points out the limitations of such assessments.
Amounts and values in the examples are illustrative and show how the calculation works. In projects, we use the client's data.
Dominik Dębski · Lemur Media · · reading time: 4 min
Reviewed by Dominik Dębski, 1 November 2026
What is an AI maturity model?
An AI maturity model is a scale of stages that shows how far a company has moved from testing tools to the sustained use of AI in processes, products and decisions. The model is used to establish the starting point and next steps, and to compare areas of the company with one another. The most frequently cited models come from the research centre MIT CISR and the analyst firm Gartner.
Which stages does MIT CISR describe?
The MIT CISR model (Peter Weill, Stephanie L. Woerner, Ina M. Sebastian, December 2024) identifies four stages of enterprise AI maturity. The shares of companies and financial results come from a 2022 survey of 721 companies; the results are the difference in growth and profit against the industry average, in percentage points.
| Stage | Share of companies | Growth | Profit |
|---|---|---|---|
| 1. Experiment and Prepare | 28% | −12,6 p.p. | −9,6 p.p. |
| 2. Build Pilots and Capabilities | 34% | −3,5 p.p. | −2,2 p.p. |
| 3. Develop AI Ways of Working | 31% | +11,3 p.p. | +8,7 p.p. |
| 4. Become AI Future Ready | 7% | +17,1 p.p. | +10,4 p.p. |
Stage 1 covers education, usage rules and access to data. Stage 2 covers piloting AI use cases and process automation. Stage 3 covers a shared AI platform and a "test and learn" approach. Stage 4 covers AI embedded in decisions and processes and selling AI-based services. The results show correlation, not proof of causation.
What does the Gartner model look like?
Gartner assesses AI maturity across seven areas and assigns a company to one of five levels. The areas are strategy, data, governance, engineering, operating model, culture, and AI product and value. The levels are Foundational, Emerging, Operational, Scaled and Transformational. The assessment result includes a map of current and target maturity in each area and the largest gaps.
How do you assess a company?
Each area is assessed separately, and the overall score is a weighted average with weights set by the board. The example concerns an e-commerce company with 250–999 employees, on a 1–4 scale corresponding to the MIT CISR stages.
| Area | Weight | Score | Weighted score |
|---|---|---|---|
| AI strategy and goals | 20% | 2 | 0,40 |
| Data | 25% | 3 | 0,75 |
| Governance and rules | 15% | 1 | 0,15 |
| Processes and operating model | 25% | 2 | 0,50 |
| Skills and culture | 15% | 2 | 0,30 |
| Total | 100% | 2,10 |
A score of 2.10 corresponds to the pilot stage. The lowest score is for governance and rules, so an AI usage policy and a system register are needed before scaling AI use cases. Good data (score 3) is an advantage the company can exploit first.
How do you move to the next stage?
Each transition requires a different kind of work: first rules and data, then pilots with a measured effect, and finally a shared platform and process changes.
| Transition | Main actions | Readiness signal |
|---|---|---|
| Stage 1 → 2 | AI usage policy, system register, data access, portfolio of AI use cases | List of use cases with value estimates and owners |
| Stage 2 → 3 | Pilot charters with decision gates, shared components, monitoring | At least several use cases in production with a measured effect |
| Stage 3 → 4 | Process and role changes around AI, AI-based products and services | AI affects revenue or margin as shown in the management report |
Lemur Media uses the LEMUR AI Readiness Index as a baseline assessment: it combines areas similar to those in the MIT CISR and Gartner models with an assessment of data, skills and governance readiness. The result shows which transition is nearest for the company and which initiatives should be in the portfolio for the next two quarters.
What are the limitations of maturity models?
- The overall score hides differences between areas; decisions are based on the gaps in individual areas.
- Board self-assessment tends to be inflated; evidence is needed, e.g. the number of systems in production and the measured effect.
- A higher stage is not a goal in itself: the goal follows from the growth strategy and the unit economics of AI use cases.
- Different models use different names and numbers of stages, so results from two models cannot be compared directly.
- The MIT CISR data come from a 2022 survey, so we treat the shares of companies at each stage as a historical benchmark.