Gartner surveyed 353 data and analytics leaders at the end of 2025 and found that only 39% were confident their organisation’s current AI investments would have a positive impact on financial performance.
Read that from the CEO’s chair. Six out of ten of the people responsible for spending the AI budget do not believe it will show up in the numbers. They are not saying so in the steering committee, because there is no acceptable vocabulary for saying it. The maturity model is that vocabulary.
What the model actually measures
Gartner’s AI maturity model sorts organisations into five stages — Foundational, Emerging, Operational, Scaled, Transformational — and scores them across seven pillars: strategy, data, governance, engineering, operating model, culture, and AI product and value. Each is rated from Level 1, planning and beginning, to Level 5, leadership.
Most explainers walk you up all five stages as though the journey were evenly spaced. It isn’t.
| Transition | Nature | What it actually demands |
|---|---|---|
| Foundational → Emerging | easy |
Buy something, run a proof of concept, present it |
| Emerging → Operational the cliff | the only one that counts |
Named ownership, governance, a definition of acceptable output, an escalation path, a data decision, and a number that says whether to keep paying |
| Scaled → Transformational | slow |
Expensive, gradual, and largely a problem for organisations that already know what they are doing |
Foundational to Emerging is easy. You buy something, you run a proof of concept, you present it. Scaled to Transformational is slow, expensive and largely a problem for organisations that already know what they are doing. Emerging to Operational is not a step at all. It is a cliff, and it is the only transition on the ladder that decides whether any of the money comes back.
Emerging to Operational is not a step at all. It is a cliff — and it is the only transition on the ladder that decides whether any of the money comes back.
Emerging means pilots. Operational means at least one AI capability running in production, with named ownership, under governance, producing a measured result. Everything that makes the second sentence true is absent from the first. A pilot needs a champion and a budget line. A production system needs someone accountable for it at 3am, a definition of acceptable output, an escalation path, a decision about what data it may touch, and a number that tells you whether to keep paying for it.
None of those are technical problems. All of them are agreements that have to exist before the work starts.
Why organisations stall there
The stall usually has a specific and unglamorous cause: nobody agreed in advance what “operational” would mean in this particular company.
So the pilot ends and the conversation that follows is a debate rather than a decision. Was it good enough? Compared to what? Good enough for whom to sign off? The absence of a definition guarantees that the answer is another pilot, and a sixth pilot is indistinguishable from a first one on the balance sheet.
Gartner’s own data suggests how widespread the underlying gap is. In a mid-2025 survey of 360 IT leaders, only 23% said they were very confident in their organisation’s ability to manage security and governance when deploying generative AI tools. Governance is one of the seven pillars, and it is one of the three that genuinely gates the cliff — alongside operating model and AI product and value. Data and engineering are necessary but they are not what stops most organisations. Strategy and culture are lagging indicators; they improve after you cross, not before.
What the organisations that cross it do differently
Gartner assessed 432 organisations across six countries and compared the high-maturity group against the low. The differences are not about tools.
High-maturity organisations keep AI initiatives running in production for three years or more at more than twice the rate of the low-maturity group. Their business units trust and are ready to adopt new AI solutions at roughly four times the rate. Around two thirds run genuine financial risk analysis, ROI analysis and customer impact measurement on their AI work. More than nine in ten have appointed a dedicated AI leader. Close to six in ten have centralised AI strategy, governance, data and infrastructure.
And the spending pattern is the clearest signal of all. Gartner found that organisations reporting successful AI initiatives invest up to four times more, as a percentage of revenue, in foundational areas — data quality, governance, AI-ready people, change management — than organisations reporting poor outcomes.
Four times more on the things that don’t demo.
The two decisions
For the CIO, the decision is procedural and it has to happen before the next pilot is funded, not after it concludes: write the gate criteria in advance. What must be true for this to go to production, who signs it off, what number decides whether it stays. If those cannot be written down, the pilot is not ready to start, and running it anyway commits real money to producing a debate.
For the CEO, the decision is harder because it is a refusal. Stop authorising pilot number six. The instinct when confidence is low is to buy more conviction — another tool, another vendor, another proof of concept with a better demo. The maturity model exists to make the alternative sayable.
We are not stalled because we chose the wrong technology. We are stalled because we never defined what finished looks like.
The model will not tell you how advanced you are. It will tell you which of the seven pillars is currently keeping your organisation on the wrong side of the only step that counts.