Omer Schneider
Most AI maturity assessments produce one number for the whole company. That number is usually wrong, not because the assessment is badly built, but because it's answering a question no real organization can honestly be asked. Maturity isn't a company-wide property. It's a department-by-department one, and averaging it into a single score hides the one thing worth knowing: where to focus next.
The four stages
| Stage | What it looks like | What's missing |
|---|---|---|
| 1. Ad Hoc | Isolated experiments. Individual employees use AI on their own initiative. | No shared strategy, no governance, no policy or budget attached to any of it. |
| 2. Emerging | Pockets of adoption. A few teams run trial projects, usually pushed by one or two internal champions. | Still uncoordinated. What works on one team doesn't transfer to the next. |
| 3. Scaling | Coordinated rollout. Shared standards, training, and policy start appearing across departments. | ROI is just starting to be measured, usually inconsistently. |
| 4. Embedded | AI is core to how the organization runs. It's a default step in workflows, not an add-on. | Continuous optimization, spend tied directly and immediately to outcomes. |
Maturity is not a single number
An enterprise doesn't have one AI maturity stage. Engineering might be operating at Stage 3 while Finance is still at Stage 1. Any assessment that produces one company-wide score is averaging away the information you actually need, which is where to focus next. Two companies can land on the exact same overall score and be facing completely different problems, one needs governance, the other needs a measurement standard, and a single number can't tell you which. An IBM study published in 2025 found 85% of organizations rated themselves "data-driven" or "AI-first," while an objective assessment placed only 11% in those higher-maturity stages. Self-reported maturity and actual maturity are measuring two different things.
The money is in the transitions, not the stages
The leap from Emerging to Scaling is where fragmented pilots either get a shared measurement standard or calcify into permanent silos. The leap from Scaling to Embedded is where "ROI is being measured" turns into "spend is tied directly to outcomes," and that's usually a bigger jump than it sounds, because it requires the underlying data infrastructure to actually support it, not just the intent to measure.
Where to go from here
Most organizations significantly overestimate which stage they're actually in, because the assessment gets done at the company level instead of the department level. A more honest exercise scores each function separately and asks a narrower question: which single transition, in which department, would move the most money if it happened this quarter.
Frequently asked questions
Can a company be in more than one AI maturity stage at once?
Yes, and most are. Engineering might be at Stage 3 (Scaling) while Finance is still at Stage 1 (Ad Hoc). A single company-wide maturity score averages this away and hides where the actual opportunity is.
What's the difference between the Scaling and Embedded stages?
At Scaling, shared standards and training exist across departments but ROI measurement is still inconsistent. At Embedded, AI is a default step in workflows and spend is tied directly and immediately to outcomes, not just tracked.
Where does most of the value get created in a maturity model like this?
In the transitions between stages, not the stages themselves. The Emerging-to-Scaling jump is where fragmented pilots either get a shared measurement standard or calcify into silos; that transition is usually worth more than reaching any single stage on its own.