When I started reading McKinsey’s 2026 State of AI report, for a second I thought I had opened last year’s file by mistake. The headline numbers felt so similar to 2025, like a déjà vu:

• 80% of individuals say they are more productive with AI. (It was 88% in 2025)
• Only 37% of organizations see that show up in EBIT. (2025? 39%)
• Actual AI-driven workforce declines sit at 14%, against the 32% that were expected a year earlier.

Two years. Same story. Same gap.

(Worth noting for the careful reader: the 2026 survey gathered 1,719 responses across 97 nations, and McKinsey cautions that some measurement bases shifted between years, so year-to-year comparisons deserve a little care.)

The adoption gap did not close this year. It moved down a level, and it has been building there quietly.

The Number That Actually Moved

Buried under the familiar headlines is the one figure that genuinely shifted. The report puts it plainly:

“Among midlevel managers and individual contributors, 47 percent say they have experienced one of these negative effects, compared with 31 percent of executives and senior managers.”

Those negative effects are stress, fatigue, and the overwhelm of keeping pace with a faster stream of output. And since only 13% of respondents see AI as a direct threat to their careers, this strain is not fear of replacement. It is operational load.

At first this seems backwards, because the 80% figure above confirms that most professionals feel AI has helped them. So why is the layer in the middle straining?

The answer sits in what we ask of them. Individual contributors get to enjoy the productivity gains. Middle managers are the ones expected to convert everyone else’s gains into results the business can point to, usually without the time, tools, or authority to redesign how the work actually gets done.

Why the Pressure Moved Down

Take a closer look, and two things are quietly loading onto this layer at once.

The first is the budget. One in five respondents say their organization is now limiting AI use because of operating costs, the tokens and agents behind the scenes. As McKinsey senior fellow Michael Chui notes, CFOs have discovered that AI is anything but “too cheap to meter” once agents start running multi-step work. More of the AI budget is moving toward compute rather than new headcount, which leaves managers absorbing tighter constraints with the same team size.

The second is the review burden. As we explored in last week’s article on the Supervision Gap, managers are the ones catching AI-assisted work at the point it reaches their desk, often without ever getting the hands-on reps of producing that work with AI themselves.

Supervising output you have never personally produced is demanding work, and this year’s data finally puts an emotional cost on carrying it.

There is a third, quieter mechanism, and it turns the pressure into a loop. Faster individuals get assigned more work, more output lands on the manager’s desk for review, and even less time remains to redesign anything. AI does not remove the bottleneck here. It relocates it.

Think of it as two gears. AI spins the small gear, the individual task, dramatically faster. If the big gear, the workflow around it, does not turn, all that extra speed becomes friction, review burden, and eventually burnout (for both the manager, and the individual contributor).

Leave that combination alone long enough, and managers stop pushing for the redesign leadership actually needs. They start “rubber-stamping” instead, because that is the option that does not cost them time they already do not have.

A Lesson From a Championship Coach

John Wooden won 10 national championships. His first practice at UCLA started with socks. 🧦

New players arrived expecting drills and strategy. Instead, Wooden sat them down and taught them how to put on their socks and tie their shoes properly, because a wrinkle means a blister, and a blister loses games. He spent his whole career watching the fundamentals nobody else was watching, and the scoreboard took care of itself.

His most famous line could sit above this entire report:

“Don’t mistake activity for achievement.”

That is exactly what the sameness of the 2025 and 2026 numbers is telling us. Activity is loud: licenses, logins, messages sent. Achievement is quieter, and much of it sits with the people nobody is measuring.

What the 6% Do Differently

If the story so far sounds heavy, the same report also contains the way out. McKinsey classifies about 6% of respondents as AI high performers, organizations attributing at least 5% of EBIT to AI, and three of their habits stand out in this year’s data:

• Nearly three-quarters of high performers (75%) have fundamentally redesigned workflows because of AI, up from 55% last year. Everyone else barely moved, from 20% to 25%. The distance between the two groups is widening, not closing.
• They treat checking as a system, with defined processes for when human experts must validate AI output. In last year’s survey, 65% of high performers had this in place, against 23% of everyone else.
• Their senior leaders are three times more likely to visibly own the AI agenda (48% versus 16%).

Notice what that list is made of: management practices, not tool choices. And most of those practices sit one level below the executive team, with the same middle managers carrying the strain. The layer under the most pressure is also the layer with the most leverage.

Ending Thought:

The most useful way to read McKinsey’s 2026 report is as a story about where pressure goes when it cannot show up in the headline numbers. The gap between adoption and impact did not shrink this year. It sank one level down, into the layer least equipped to carry it and most relied upon to close it.

McKinsey’s own commentary points the same way: the limiting factor is increasingly the organization’s ability to absorb change. The 80% productivity gain is real, and it is already inside your building. The open question is whether the organization can absorb it.

The good news is that this is a solvable, human problem. Managers do not need another tool. They need time, reps, and the authority to change how the work gets done, and every one of those is within a leadership team’s gift.

If your organisation is working on the layer between the people who use AI and the people accountable for its results, Radiant Institute designs manager-focused AI enablement programs for exactly that. Reach out to explore how we can help.