Change Capacity and the Pace of AI Adoption: What Gartner Measured, and Where Our Argument Begins

Your company bought an AI tool this year, and somebody is now asking why the time savings have not appeared. The savings are real at some companies, and the gap between buying and using is where the rest of them go. Buying is going well, and using is where it stalls.

MIT Sloan Management Review and BCG surveyed 2,102 executives in November 2025 and found 35% already using agentic AI, with another 44% planning to deploy it soon. Few of those companies had restructured any part of the work around it. Gartner, 6 months later, reported that only 20% of executives believed their workforce was truly AI-ready, and that 19% of employees had saved no time at all.

Those 2 readings are a long way apart, and something causes that. That distance is usually explained as a skills problem, which is reasonable and incomplete. A second explanation fits the same evidence better, because the workforce has run out of room, and room is a different thing from skill.

What the Numbers Show

Running out of room is measurable, and Gartner measured it. Gartner's measurement found employee willingness to support enterprise change falling from 74% in 2016 to 43% in 2022. Over the same 6 years the average employee went from 2 planned enterprise changes a year to 10.

Read those 2 numbers together and the picture is a workforce absorbing change at 5 times the rate it was built around. The second number explains the first, because willingness is a behavioral measure of how much is already in flight at the moment somebody asks you for one more thing. When it drops, programs do not fail loudly, and they take longer and return less than the business case assumed.

The Part That Gets Misread

Read as a morale finding, this is a common and expensive error. Morale is how people feel about the work, and willingness is what they have left to give a new initiative after the current ones have taken their share. Any well-liked company still has a capacity problem at initiative 11.

Why the Skills Explanation Is Incomplete

So a capacity problem explains what a skills problem cannot, and the difference between them decides whether a training budget does any good. IBM found in 2023 that limited AI skills and expertise was the barrier most often named, by 33% of respondents, which is a large enough share to settle that question. The finding is real, and it describes a symptom more than a cause.

That cause is capacity, because a company with none cannot build skills either, and skill building is itself a change. Building them competes for the same calendar and the same attention. So a skills budget in a saturated window buys completions and little behavior.

Behavior is what McKinsey's 2023 work on generative AI was pointing at, treating reskilling and capability building as conditions on capturing the productivity gains. The condition underneath that one is simpler and easier to miss. Capability building needs room on the calendar before it needs budget.

Change Fatigue in Employees Is the Real Constraint on AI Adoption

Change fatigue in employees is the real constraint on AI adoption, and the room it takes away is simple to describe. AI arrives as change number 11 in a year that already holds 10, and nobody at the company removed a single one of those 10 to make space. Each of those 11 competes for the same hour, and the new tool does not come with an extra one.

Once you see it this way, what you measure before a deployment changes. That question stops being whether the workforce has the skill and becomes how many changes it is already carrying. Teams that run AI skills assessments to map workforce gaps usually find the variance between departments is wider than the variance inside one.

What Sequencing Changes by 2027

Given that variance, the move available to you is an unglamorous one. That move is cutting the number of simultaneous changes, which is what separates the companies reporting real AI productivity from the rest. Such companies look slower on paper for 2 quarters and finish ahead.

Finishing ahead comes from sequencing, which beats volume once the workforce is past its limit and makes it an operating decision. A chief executive can approve 10 initiatives in a planning cycle, and only the person running the work can decide which 3 of them move at once. That is the hopeful part, because capacity is yours to control.

Control starts with knowing where capacity currently is, by team and by role. Knowing that is harder than it sounds, because most organizations plan a major initiative as though readiness were uniform. Readiness never is uniform, and one department absorbs the change in 3 weeks while another is still 2 initiatives behind.

Where to Start

Counting those initiatives is cheaper than any purchase and takes an afternoon. Count the live change programs your people carry this quarter, and add the ones nobody formally closed. Those 2 numbers added together tell you more than any readiness survey will.

Those numbers lead to a second step, which is making the capacity you have visible to the people spending it. That visibility is what KC LMS gives you, with what is assigned, what is overdue and who is carrying it, by name and by team, before anybody approves the next thing. Each manager sees the same picture in KC Performance, alongside the goals that person is being measured on. Managers of roles with no obvious KPI notice the difference first.

Both views at once are what let you say no to initiative 11 with a reason attached. The reason is simple enough to say out loud, because 3 changes run properly beat 10 run badly. AI deserves one of those 3 slots, and the 11th slot kills it.

Frequently Asked Questions

1. What is change capacity?

Change capacity is how much new work a workforce can absorb before its willingness to support the next initiative drops. Gartner measured that willingness falling from 74% in 2016 to 43% in 2022, over the same period that the average employee's count of planned enterprise changes rose from 2 to 10. It is an operating variable, and it moves with the number of things running at once.

2. Why would change fatigue affect AI adoption specifically?

AI is a change in how people do their daily work, so it draws on the same capacity every other initiative draws on. Gartner reported in May 2026 that only 20% of executives believed their workforce was truly AI-ready and that 19% of employees had saved no time from AI deployment. Those figures describe tools in place that nobody has had the room to adopt.

3. Is this not simply a skills gap?

Skills are part of it. IBM's 2023 index found limited AI skills and expertise was the most commonly named barrier, at 33% of respondents. The point underneath that figure is that building skills is itself a change program, so a workforce with no remaining capacity struggles to complete the training that would close the gap.

4. What should a company measure before approving an AI rollout?

Count the change programs your people are carrying this quarter, including the ones nobody formally closed. Compare that count by department rather than for the organization as a whole, because readiness is rarely uniform. Then decide which initiatives can wait a quarter so the one you care about has room.

5. Does slowing down mean falling behind?

The evidence points the other way. Companies reporting real AI productivity are the ones that ran fewer changes at once and finished them. Running 3 initiatives properly produces more adopted capability than running 10 that each stall halfway.

References

1. Harvard Business Review. Employees Are Losing Patience With Change Initiatives. Gartner research. May 2023.
2. Gartner Newsroom. Press release on people-centric AI strategy. May 2026.
3. Boston Consulting Group. Agentic AI Blurs the Line Between Tool and Teammate. With MIT Sloan Management Review. November 2025.
4. McKinsey Global Institute. The Economic Potential of Generative AI. June 2023.
5. IBM Institute for Business Value. Global AI Adoption Index. 2023.