The AI Maturity Model: 4 Stages for Growing Companies
An AI maturity model shows how far your company has moved from people trying AI on their own to AI running inside how the business works. For a company of 25 to 500 people, there are four stages: Scattered, Experimenting, Operating and Compounding. Most mid-market companies sit in the first two, and the jump that matters most is from Experimenting to Operating.
You do not need a consultant to place yourself. You need an honest look at three things: who owns AI, where it shows up in real work, and whether you can point to a number it moved. This post walks through each stage, the signs you are in it, and what it takes to reach the next one.
What is an AI maturity model?
An AI maturity model is a ladder of stages that describes how deeply AI is built into a company's work, decisions and results. The label is not the point. The point is knowing which move comes next, so you stop spending on things that belong two stages ahead.
The big consultancies have their own versions, and they all tell the same story: most companies are stuck low on the ladder. BCG's research found that only 5% of companies are "future-built" and getting value from AI at scale, while 60% report minimal revenue and cost gains despite substantial investment. McKinsey's 2026 survey shows the same gap from another angle. 44% of respondents say AI is scaling across their enterprise, but just 6% qualify as high performers who attribute 5% or more of EBIT to AI.
Those studies lean toward large enterprises. The four stages below are written for the mid-market, where there is no AI team and the CEO is usually the one who has to make the call.
What are the four stages of AI maturity?
Stage 1: Scattered
Scattered means AI is used by individuals, not by the company. A few people pay for their own tools, nobody knows what data goes where, and leadership has no shared view of what is working.
This is more common than most CEOs think. Microsoft and LinkedIn's Work Trend Index found that 78% of AI users bring their own AI tools to work, rising to 80% at small and medium-sized companies. If you have not set a direction, your people already have, one browser tab at a time.
Signs you are here: no named owner for AI, no approved tools list, and the honest answer to "where are we using AI?" is "not sure."
Stage 2: Experimenting
Experimenting means the company is trying AI on purpose, but in pilots that sit beside the real work instead of inside it. There is budget and energy. There are demos. There is rarely a number on the P&L that moved.
This is where pilot sprawl and AI fatigue set in. Each department runs its own trial, nothing scales, and people start to see each new tool as one more thing to learn.
Signs you are here: five or more pilots, no clear success metric on most of them, and a vendor list that keeps growing.
Stage 3: Operating
Operating means a small number of AI use cases are live inside core workflows, each with a business owner and a metric. People use them daily because the process now assumes they will. AI stops being a project and becomes part of how a team works.
The difference from Experimenting is not better technology. It is redesign. In McKinsey's survey, nearly three-quarters of high performers report fundamentally redesigning workflows because of their AI use, compared with one-quarter of other respondents.
Signs you are here: two or three use cases you could show a board member, each tied to a KPI you already report, with usage you can actually see.
Stage 4: Compounding
Compounding means each AI win makes the next one cheaper and faster. Data, playbooks, skills and owners carry over. The company has a steady rhythm for choosing, shipping and retiring AI use cases, and leaders expect AI to show up in every plan.
BCG's numbers show why this stage is worth reaching. Its future-built companies see twice the revenue increase and 40% greater cost reductions than laggards. The gap widens over time because leaders keep building on what already works.
Signs you are here: a ranked AI backlog reviewed by the leadership team, data and tools reused across functions, and new hires trained on AI-shaped workflows from day one.
How can you tell which stage your company is in?
Answer five questions honestly with your leadership team. Most companies land lower than they expected, and that is useful.
- Who owns AI? Nobody (Scattered), a committee or IT (Experimenting), named business leaders per use case (Operating), or the leadership team as a standing agenda item (Compounding).
- Where does AI show up? In personal tabs, in pilots, in core workflows, or in every plan.
- Can you name a number it moved? No, maybe, yes for a few use cases, or yes and you track it monthly.
- What happens when a pilot ends? Nothing, it slowly fades, it graduates into the workflow or gets killed, or it feeds the next bet.
- How do people learn AI? On their own, at a one-off workshop, inside their daily workflow, or as part of onboarding.
If your answers spread across two stages, you are in the lower one. Maturity is set by the weakest link, and the weakest link is usually ownership.
How do you move from one stage to the next?
Each move needs one specific shift. Trying to skip a stage is the most common and most expensive mistake.
- Scattered to Experimenting: set a direction. Pick approved tools, write a one-page policy on what data can go where, and ask each leader for the three most repetitive tasks on their team. This is weeks of work, not months.
- Experimenting to Operating: cut and commit. Stop most pilots. Choose one core bet and one or two quick wins, give each an owner and a metric, and redesign the workflow around them. Our guide on how a CEO should prioritize AI projects walks through the scoring.
- Operating to Compounding: build the rhythm. Keep a ranked backlog, review it monthly, and make each new use case reuse what the last one built: the data, the prompts, the training and the owner's playbook.
Notice that none of these moves is about picking a better model. BCG's rule of thumb is that 70% of the focus should go to people and processes, 20% to technology and 10% to algorithms. The hard part is people and process. That is also where the value is.
Why does the jump from Experimenting to Operating matter most?
Because that is where the money starts. Pilots cost money and attention. Operating use cases pay them back. Most mid-market companies stall here, with plenty of activity and little to show for it, and the reason is usually the same: nobody made the hard call about what to stop.
That call belongs to you. IT can tell you what is possible. Your leaders can tell you what hurts. Only the CEO can trade one priority for another and make it stick. McKinsey found that high performers are twice as likely as others to say their senior leaders demonstrate commitment to AI initiatives. Commitment here does not mean speeches. It means choosing, funding and following up.
What should a CEO do this quarter?
- Place yourself honestly. Run the five questions with your leadership team and agree on one stage.
- Name one next move. Only the move for your current stage. Ignore advice written for companies two stages ahead.
- Use AI on your own week. You cannot lead a change you have not felt. A good place to start is an AI chief of staff for your own calendar, inbox and meeting prep.
- Set one number. Pick the KPI that will tell you the next stage is real, and check it every month.
If you want an outside view of where you stand, Journey offers a free CEO AI audit that maps your company against these four stages and points to the next move. Either way, the work is the same: be honest about the stage, pick one move, and see it through.
Frequently Asked Questions
An AI maturity model is a set of stages that shows how deeply AI is built into a company's work, decisions and results. For mid-market companies, a practical version has four stages: Scattered, Experimenting, Operating and Compounding. Its main use is to show which move comes next, so you stop investing in things that belong to a later stage.
The four stages are Scattered, where individuals use AI on their own; Experimenting, where the company runs pilots beside the real work; Operating, where a few AI use cases live inside core workflows with an owner and a metric; and Compounding, where each AI win makes the next one faster and cheaper because data, skills and playbooks carry over.
Ask five questions: who owns AI, where it shows up in real work, whether you can name a number it moved, what happens when a pilot ends, and how people learn it. If your answers spread across two stages, you are in the lower one. Ownership is usually the weakest link and sets your real stage.
Moving from Scattered to Experimenting can take a few weeks: pick tools, set a data policy and gather use cases. Moving from Experimenting to Operating usually takes one or two quarters, because it means stopping pilots and redesigning workflows. Compounding is not a finish line. It is a rhythm you keep running.
The CEO should own the direction and the trade-offs, because only the CEO can decide what gets stopped to make room for AI. Each live AI use case should then have one named business owner, not IT, with the metric in their goals. IT supports security, data and tools, but the business owns results.
