Ask a leadership team how mature their AI program is and you will get a single number. "We're a level 3." "We're maybe a 4." That single number is the reason so many AI programs quietly stall.
AI maturity is not one number. It is two, and they move independently.
Breadth and depth are different axes
Breadth is how widely AI touches the organization. How many people, how many teams, how many workflows have AI in them at all.
Depth is how deeply any given workflow has been rebuilt around AI. Not "we use a chatbot sometimes" but "this process was redesigned so that AI is load-bearing and the old way is gone."
These two do not rise together, and conflating them produces two very different failure modes wearing the same score.
The two ways to be "a level 3"
Wide and shallow: Everyone has a license. Everyone has touched the tool. Adoption dashboards look great. But no single workflow has actually changed, so none of the work is faster, cheaper, or better. This org has breadth with no depth. It feels mature and delivers almost nothing.
Narrow and deep: A few teams have rebuilt their core workflows around AI until the new way is just how the work happens. The impact is real and measurable, but it is confined. This org has depth with no breadth. It is delivering value but cannot yet claim organizational maturity.
A single maturity number cannot tell these two apart. And they need opposite next moves. The wide-and-shallow org needs to go deep on a few workflows and stop counting logins. The narrow-and-deep org needs to spread its proven patterns. Give them the same generic "advance to level 4" advice and you will misfire on both.
Why most orgs overrate themselves
Most organizations think they are at stage 4. Most are at stage 2. The gap is not arrogance, it is the wrong measuring stick.
People rate maturity by capability — tools purchased, pilots launched, training delivered. But capability is what your tools can do. Maturity is what your people actually do on a Tuesday. The honest measure is workflow change, and when you switch to it the number usually drops two levels.
That drop is not bad news. It is the start of real progress, because you are finally measuring the thing that pays rent.
A test you can run this week
For each major workflow, ask two questions:
- Breadth: Is AI present in this workflow at all? (yes / no)
- Depth: If AI disappeared tomorrow, would this workflow break, or would people barely notice? If they would barely notice, you have breadth without depth.
Plot the answers on two axes. The shape tells you the move. Wide and shallow: pick three workflows and go deep until removal would hurt. Narrow and deep: take a proven pattern and widen it.
The deeper point
Maturity is a verb, not a purchase order. The most mature AI team I ever saw had the smallest tool budget. They had not bought the flashiest stack. They had picked three workflows and rebuilt them until the new way was simply how the work happened.
Stop reporting one number. Report two, and the right next step stops being a guess.
This draws on the AI Usage Maturity Model in LEVEL UP, a measurement framework for leaders building real AI programs.
Top comments (0)