An AI Maturity Model for Growing Businesses

An AI maturity model describes the stages a business passes through as AI moves from personal experiments to shared, governed work. For a growing business, five stages fit: individual use, sanctioned tools, shared knowledge, written workflows, and agents with approval gates. Most businesses of 10 to 200 people sit at the first or second.

What is an AI maturity model?

An AI maturity model is a staged description of how an organisation's use of AI develops, from first experiments to AI built into everyday decisions. It helps leaders see where they are, what the next stage requires, and which investments are premature. Its value is in the sequence, not the labels.

Maturity models are popular because AI adoption feels shapeless. Everyone uses something, nobody can say what it returns, and spending decisions get made on enthusiasm. A model replaces "are we doing AI?" with "what is the next thing we need?"

A good model also tells you what not to do yet. Building an AI agent before the business has written down its product facts, for example, produces an agent that confidently invents them.

What do the established AI maturity models say?

The best-known research comes from enterprise studies. MIT CISR's model has four stages: experiment and prepare, build pilots and capabilities, develop AI ways of working, and become AI future ready. Its 2024 research found most enterprises in the first two stages, with financial performance below industry average there and well above it in stages three and four.

That research surveyed 721 companies. Survey data from McKinsey points the same way. According to McKinsey's 2025 State of AI survey, 88% of organisations use AI in at least one business function, yet only about a third have begun scaling it across the enterprise.

These models are useful but built for a different reader. Stage two in MIT CISR's model, for instance, involves consolidating data silos and investing in APIs. That is real work for a bank; a 40-person consultancy needs stages it can recognise from its own Tuesday.

What are the five stages of AI maturity for a growing business?

The five stages are individual use, sanctioned tools, shared knowledge, written workflows, and agents with gates. Each one adds something the business owns rather than something an individual knows. The jump from the second stage to the third, where business knowledge is written down for every AI to read, is the one most businesses miss.

Stage one is where most businesses start, often without deciding to. According to Microsoft and LinkedIn's 2024 Work Trend Index, 78% of AI users bring their own AI tools to work. Whatever those people learn stays in their personal accounts.

Stages three to five map to the three layers of a Company OS: data, workflows and agents. The explainer on the three layers covers how they fit together.

How do you tell which stage your business is at?

Ask five questions and stop at the first "no". Are staff on business AI accounts? Is there a written AI policy people have read? Are product facts and customer language written down where every AI can read them? Are recurring jobs written as workflows? Do any agents run them, with a named approver?

The questions are deliberately concrete. "Do we have an AI culture?" invites a hopeful answer. However, "can someone show me the file our AI reads for product facts?" does not.

Watch for a common mismatch. Some businesses run an impressive agent pilot while staff still use personal accounts with customer data. That is stage one with a demonstration on top, and the pilot rarely survives contact with the rest of the business. A structured AI readiness assessment covers the same ground in more depth.

How do you move up a stage?

Move one stage at a time and finish each before starting the next. At stage one, move people onto business plans. From stage two, write down what is true about the business. Once that exists, pick the workflow the team runs every week and write it down. Agents come once workflows run reliably.

Each stage has an obvious first move:

1. Individual to sanctioned. Move anyone using personal AI accounts with company data onto business plans, then publish a one-page policy.
2. Sanctioned to shared knowledge. Write down product facts, objections, customer language and brand voice, each with an owner.
3. Shared knowledge to workflows. Choose the job that pays back fastest, usually a weekly one, and write it step by step.
4. Workflows to agents. Let an agent run a proven workflow, with read-only access first and a named approver at the end.

The third step is where the returns usually start. Shared knowledge makes every person's AI better at once, so output stops sounding generic. The guide to the company brain explains what goes into it.

Skipping stages is the usual failure. According to the MIT NANDA research, most systems studied did not retain feedback, adapt to context or improve over time. Those are exactly the capabilities that stages three and four build.

What does AI maturity look like in results?

Maturity shows up in business measures, not in the number of tools. A more mature business produces first drafts faster, wins more pipeline from the same team, lowers cost per lead and appears more often in AI answers. Track the measures that matter to you at each stage, so every step up is justified by evidence.

Be honest about the ceiling, too. A 30-person business does not need stage five everywhere. Many will run agents on two or three workflows and keep the rest at stage four, which is a sound place to stop.

Skein's strategy work is a fixed-scope diagnostic: it listens, maps, ranks and plans, and ends in a ranked shortlist of three to five opportunities, a 90-day plan and a 12-month roadmap; the AI strategy page shows the steps. That suits a business that knows it is at stage one or two and wants a sequence it can trust. A business that has already mapped its workflows may only need help building the next one.

Frequently asked questions

What is an AI maturity model?

An AI maturity model is a staged description of how an organisation's use of AI develops, from scattered personal experiments to shared, governed work built into everyday decisions. It helps leaders see where they stand, what the next stage needs, and which investments would be premature at their current stage.

What are the stages of AI maturity?

For a growing business, five stages work well: individual use, sanctioned tools, shared knowledge, written workflows, and agents with approval gates. Enterprise models such as MIT CISR's use four stages, from experiment and prepare to becoming AI future ready, but they assume enterprise data teams and budgets.

How do I assess my company's AI maturity?

Ask five concrete questions in order and stop at the first no. Are staff on business accounts? Is there a policy people have read? Are business facts written down for AI? Are recurring jobs written as workflows? Do agents run any of them with a named approver?

Can a small business skip AI maturity stages?

Skipping usually fails. An agent built before business knowledge is written down will invent facts, and a workflow without sanctioned accounts leaks data. Moving one stage at a time is faster overall, because each stage gives the next one something reliable to build on and evidence to justify it.

Does every business need to reach the top stage?

No. Many growing businesses will run agents on two or three high-value workflows and keep the rest as written workflows that people run with AI help. The right stopping point is where the next step costs more than it returns, which differs for every business and every workflow.

Related reading