- Maturity runs from ad-hoc experiments to AI being core to operations. Most firms sit lower than they assume.
- The evidence: in Publicis Sapient's 2026 survey of large organisations, 73% said AI is used regularly or across most business processes, but only 10% said it is core to how the business operates.
- You climb one process at a time. There is no platform you can buy that jumps you two stages.
An AI maturity assessment places your business on a scale from occasional experiments to AI embedded in how you operate, and names the one thing blocking your next step. It is worth doing because self-assessment is unreliable in a predictable direction: firms rate themselves higher than the evidence supports. The headline number makes the point. When Publicis Sapient surveyed AI decision-makers at large organisations in 2026, 73% said AI is used regularly or across most business processes, but only 10% said it is core to how the business operates. Using AI and being mature in it are different things, and the gap between them is where most of the value still sits unclaimed.
This is the stage above readiness. The readiness assessment checks whether you can start; maturity tracks how far you have actually got and what the next honest move is. It pairs with the AI audit, which supplies the opportunities you climb with.
The five stages, plainly
| Stage | What it looks like |
|---|---|
| 1. Experimenting | Individuals trying tools ad-hoc. Nothing owned, nothing measured, no process changed. |
| 2. Pockets | A few teams use AI tools regularly, but each in isolation. Gains are personal, not organisational. |
| 3. In production | At least one real process runs on AI with a human checking output and a number attached to it. |
| 4. Embedded | Several processes run on AI, governed and measured. It is part of how work moves, not a bolt-on. |
| 5. Core | AI is central to the operating model. Removing it would break how the business runs. |
The stages are not a technology choice, they are an operating one. The jump that matters most is from stage 2 to stage 3, because that is where value stops being a personal productivity anecdote and starts showing up in the accounts.
Where firms actually land
The distribution is humbling. Beyond that small core, BCG's 2026 Applied AI Index puts only 7.5% of companies in its "future-built" group and a further 41% actively scaling AI, together nearly half the market. Everyone else is still waiting for the payoff. McKinsey's 2026 State of AI survey points the same way: 44% of respondents say AI is now scaling across their enterprise, up from 38% a year earlier. So the honest read is that many businesses are at stage 2, believing they are at stage 4. And the block is rarely the technology: 22% of Publicis Sapient's respondents named the way their organisation operates as the main barrier to AI success. This is happening against enormous spend, with Gartner forecasting worldwide AI spending of $2.59 trillion in 2026, up 47% year on year. Money is not the constraint. Operating discipline is.
Those surveys cover large organisations. For the UK as a whole, the official picture starts lower and splits sharply by size. The ONS found AI in use at 28% of businesses with 0 to 9 employees, against 49% of businesses with 250 or more employees (ONS: AI in UK businesses). So most of the smallest firms report no AI use at all: they're not on the ladder yet. The same release found that 41% of UK businesses with 10 or more employees report no barriers to adoption. Our read: when a firm sees nothing in its way and is still at stage 2, what's missing is usually the step of putting one process into production, not a tool or a budget.
How you climb a rung
You do not buy your way up a stage, you earn it one process at a time. The move from pockets to production is a single high-value workflow: put AI into it in production, keep a person checking the output, prove the number, then move to the next. This is the shadow-mode-then-cutover pattern we use, described in the piece on designing human oversight that does not become the bottleneck. Trying to leap from stage 2 to stage 4 with a company-wide programme is exactly the move that produces stalled pilots and no return.
Why the honest stage matters
Knowing your real stage stops you buying the wrong thing. A stage-2 business does not need an enterprise AI platform; it needs one process shipped into production. A stage-4 business does not need another pilot; it needs governance that scales. The assessment is worth the hour because it points your next pound at the move that actually changes your stage, rather than the one that looks most like progress. Where a build is the right next step, the bespoke software guide covers how to buy it without the overrun.
Common questions
What is an AI maturity assessment?
It places your business on a scale from ad-hoc experimentation to AI embedded in daily operations, and names the specific thing blocking the next step. It is less about a score and more about knowing which move actually matters right now.
What are the stages of AI maturity?
A practical ladder runs from experimenting, to using tools in pockets, to running one process on AI in production, to embedding it across several processes, to AI being core to how the business operates. Most firms are lower than they think.
What stage are most businesses at?
Lower than the headlines suggest. Most use AI somewhere, but far fewer have made it central: among large organisations surveyed by Publicis Sapient, 73% said AI is used regularly or across most business processes, but only 10% said it is core to how the business operates. BCG's 2026 index counts 7.5% of companies in its "future-built" group and a further 41% actively scaling AI, together nearly half the market, which leaves the rest at the pilot or pockets stage.
How do you move up a stage?
One process at a time, not one grand programme. You pick a single high-value workflow, put AI into production on it with a human checking the output, prove the number, then repeat. Maturity is earned in steps, not bought in a platform.
If you want a straight read on which rung you are actually on and which single move takes you up one, that is a short conversation. Run through the readiness checklist first if you want to self-score, or start with us at contact.
Sources
Every figure in this article links back to the source below it was checked against.
From the Orwell Lab archive, rebuilt in our current design. Figures checked against the sources listed above on 30 September 2026.
Discuss your own requirements ↗