AI maturity model
Also known as: AI maturity framework, AI maturity levels
What is AI maturity model?
An AI maturity model is a staged framework that shows where an organisation stands in its use of AI and what the next step is. The levels are described as observable practices, so a leadership team can place itself without commissioning an assessment first. The value sits in the discussion the scale forces, not in the label it hands out.
Why does an organisation need a maturity model?
A maturity model gives leadership a shared scale for the argument about how far to take AI. Without a scale the discussion becomes an exchange of impressions in which everyone describes a different level using the same words. The scale also exposes the gap between use and management. Statistics Finland found that 38 percent of Finnish companies with at least ten employees used AI in spring 2025, while only 15 percent had documented guidelines for that use. On a scale, the gap becomes a state with a next step.
The risk is the model turning into a scoreboard. Once climbing a level becomes the goal, it displaces the business outcome the level was meant to represent. As a diagnostic the model earns its place, as a trophy it wastes everyone’s time. The top level is also not automatically the right target, because the target level follows from strategy.
The Sofokus AI maturity model
The Sofokus AI maturity model has four levels, and one question separates them: where does the benefit arise and who leads it. The model is built around the everyday reality of Finnish organisations, and it is presented in Sofokus’s AI guide for business leaders.
| Level | Where the benefit arises | Who leads it |
|---|---|---|
| 1 Unmanaged | With individual employees, at random | Nobody |
| 2 Individual | As time saved by individuals, which does not add up across the organisation | Whoever provides the tools and the rules |
| 3 Process | In named processes, measured | Process owners |
| 4 Organisation | As organisational capacity that reaches the result | Leadership, as an investment portfolio |
At the unmanaged level AI is already in use, but nobody has a full picture of where, by whom and with which tools. At the individual level the organisation has taken individual use in hand with chosen tools, written rules and training, yet the work still runs through the same process as before. The process level builds AI into named processes so that skipping the AI step would be an error, and the effect is tracked with process metrics. By the organisation level, production capacity has come loose from headcount, and the effect is reported as a business figure alongside other result figures.
Sofokus derived the levels bottom-up from observable signs. Every sign takes the form “you have this” or “you do not have this”, so a leader places the organisation while reading. A level is a state, not a grade. Levels 1 and 2 are the usual starting point. The narrowness shows up in measurements elsewhere too. In the U.S. Census Bureau’s business survey, 57 percent of the companies using AI used it in no more than three business functions.
Gartner’s maturity model
Gartner’s AI Maturity Model has five cumulative levels: Awareness, Active, Operational, Systemic and Transformational. The scale describes an organisation’s use of AI as a whole, and it is the best known maturity scale in the field. Gartner’s own criteria and toolkit sit behind a paywall, so what circulates publicly are the level names and broad descriptions.
The models differ in what they ask. Gartner’s scale describes capability developing broadly, while the Sofokus model asks one thing: where the benefit arises and who leads it. The practical consequence shows in what follows from a level. A descriptive scale reports a position, and a scale tied to observable signs also names the next visible change. Some models publish their criteria openly, the MITRE AI Maturity Model among them. The choice depends on whether you need shared vocabulary for a discussion or a checklist for the work.
Maturity, AI-nativeness and AI-ification answer different questions
AI-opas johtajalle separates three concepts that blur together in everyday conversation. AI-nativeness is the starting point, the assumption an operation is built on. AI-ification is the movement, an organisation’s own act of redesigning work around what AI makes possible. Maturity is the position, where on the scale the organisation stands right now. The framing belongs to the guide and is not an established industry split.
Keeping them apart pays off, because each answers a different question. Nativeness answers what assumption you build on, AI-ification answers what is being done right now and who owns it, and maturity answers where you stand relative to where you are heading. An organisation can reach a high level without being AI-native, and AI-ification can run for years before the level moves.
In practice
A hundred-person professional services firm first places itself at level 3, because AI is in wide use and the staff survey reports weekly benefit. The observable signs overturn the placement. Not one process has changed so that skipping the AI step would be an error, and no process has a baseline recorded before the tools arrived. The firm is at level 2, and its next step is to name two processes, at least one of them billable, and measure them before and after. The placement changed because the scale asked about observable facts rather than impressions.
Frequently asked questions
What level is our organisation at?
You find the level by reading the level descriptions and asking which one describes your working week as it actually is. The observable signs decide it: is there a written AI guideline in onboarding, is there a process where skipping the AI step would be an error, and is the effect reported as a business figure. When the answer stays unclear, the level is the lower one.
How does the Sofokus model differ from Gartner’s?
Gartner’s five-step scale describes the maturing of AI use broadly, and its criteria sit behind a paywall. The four-level Sofokus model asks one question, where the benefit arises and who leads it, and ties its levels to observable signs. Gartner’s scale suits building shared vocabulary, while a scale tied to signs suits deciding what to do next.
Is an AI maturity model the same as being AI-native?
No. AI-nativeness describes a starting point, an organisation built around AI rather than one that uses it. Maturity describes a position on a scale at a given moment. A native organisation usually sits high on the scale, but a high level does not make an organisation native, because the same level can be reached by redesigning an existing operation.
Why does the Sofokus model have four levels instead of five?
The signs settled into four states, and every boundary between them is separated by a question a leader can answer without help. A fifth level would have required splitting the top level, and its two halves, measured result impact and a changed revenue model, can appear independently of each other. The signs therefore do not form a staircase. In practice the top category would also be close to empty.
Is the highest level always the goal?
No. The target level follows from strategy, not from the model, and for many organisations the sensible target is the next level rather than the top one. The top level is also a state that has to be maintained: as the field moves, the bar for the same level keeps rising. Choosing a target level is a leadership decision, and the model exists to make that decision explicit.
Sources
- Statistics Finland: Use of information technology in enterprises 2025 – AI adoption rate and the share of companies with documented guidelines.
- U.S. Census Bureau: Business Trends and Outlook Survey, CES-WP-26-25 – how broadly AI is used across business functions.
- BMC: Gartner’s AI Maturity Model – the five Gartner level names as publicly reported.
- MITRE AI Maturity Model – a maturity model with a fully public criteria set.
