Your company bought the licenses. Someone ran a lunch-and-learn. The all-hands deck had a slide that said "AI-first." Six months later you check the logs and a quarter of the seats have never been used. The leadership read on this is usually "our people are resistant to change." That read is almost always wrong, and it sends you spending money in the wrong place.
The confusion is that two different things wear the same word. Capability is what your tools can do. Adoption is what your people actually do. The gap between them is your AI strategy, and most organizations have no plan for the gap at all because they cannot see that it exists.
Capability is easy to buy. Adoption is not.
Capability is a purchasing decision. You evaluate vendors, you benchmark, you sign a contract, and on day one the capability is fully present. It shows up on a line item. It is legible to a budget.
Adoption is a behavior change, and behavior change does not respond to purchasing. You cannot buy the moment when a project manager stops opening a blank document and starts opening the tool instead. That moment happens, or fails to happen, inside an actual workflow on an actual Tuesday, under deadline, with the old habit sitting right there as the path of least resistance.
So you can have 100 percent capability and 20 percent adoption at the same time, and the org chart will tell you you are "doing AI." The logs will tell you the truth.
Training does not move the number. Workflow does.
Here is the part that costs the most to learn the hard way. Training, by itself, does not move adoption. You run the workshop, everyone nods, and within two weeks usage settles back to where it was. Not because the training was bad. Because training teaches a capability and then returns people to a workflow that does not require the capability.
If the workflow does not change, the maturity does not change. The default path still routes around the new tool. People are not being stubborn. They are being efficient inside the system you actually built, which is different from the system you described in the deck.
The lever that works is embedding the tool into the path of the work, so that using it is the easy way and not using it is the friction. Make the AI step the default in the template, the checklist, the ticket flow, the review. Adoption follows the path of least resistance, every time, in every org, without exception.
Most teams misread their own stage
A useful frame: think of AI maturity as five stages, roughly from "experimenting individually" to "AI is structurally embedded in how decisions get made." Run any leadership team through it and a pattern shows up. Most of them will place the organization at stage four. Most of them are actually at stage two.
The gap is not ego. It is a measurement error. Leaders see the capability they purchased and read it as adoption they achieved. They see the pilot that went well and miss that the pilot never propagated past the three enthusiasts who volunteered for it. Capability is visible from the top. Adoption is only visible from inside the work.
This is why the honest first move in any AI strategy is not buying more capability. It is going and looking at what people actually do, in the actual workflow, when no one is presenting. That is the only place the real stage lives.
What to do with this on Monday
Three moves, in order:
Measure adoption, not licenses. Stop reporting seats purchased. Start reporting the percentage of a specific, named workflow that now runs through the tool. "60 percent of first-draft proposals start in the tool" is a strategy metric. "We have 400 licenses" is a procurement metric wearing a strategy costume.
Pick one workflow and change the default. Do not try to move the whole org. Take a single high-frequency workflow and rebuild it so the AI step is the path of least resistance. Make the old way the one that requires extra clicks. Watch what happens to usage.
Stop funding training as your adoption plan. Training is fine as a supplement. It is a disaster as a strategy. If your entire adoption plan is "we will train them," you are funding the thing that reliably does not move the number.
The organizations pulling real value from AI are not the ones with the most capability. Plenty of low-capability orgs are getting more out of modest tools than high-capability orgs are getting out of frontier ones. The difference is entirely in the gap, and the gap is a workflow design problem, not a tooling problem and not a people problem.
So the question worth sitting with: in your organization, what is the actual adoption rate of the AI capability you have already paid for? Not the license count. The behavior. If you do not know the number, that is your AI strategy talking.
I write about the gap between AI capability and AI adoption, and the maturity model behind it, at IamHITL.com.
Top comments (0)