Turning AI Adoption Into Cognitive Fitness: What the Data (and Practice) Shows

_{^{Visual created using Gemini AI by Google}}

Recent data reveals a precise ‘productivity sweet spot’ for AI: using it for exactly 7% to 10% of the workday. According to ActivTrak’s 2026 workforce report, people in this range reach 95% productivity. However, only 3% of users actually fall within that range. This highlights a gap between having a tool and truly mastering the work.

Three percent. How did we end up here?

We’re all still figuring out what to measure here. Licenses, tokens, prompts processed are the easiest things to count, so they’re the ones showing up on dashboards. That’s understandable. But those numbers tell you the tools are being used, not that the work is getting better.

That’s the gap I keep circling. Adoption is one thing. Letting AI change the work is another.

The data underneath

Here’s what’s wild about the ActivTrak (2026) finding. The same period that produced peak productivity in the 7–10% band also produced a three-year low in focus efficiency (60%) and a 23% rise in disengagement risk¹. So most users are using AI more, focusing less, engaging less.

Deloitte’s AI Institute (2026) saw the same shape from a different angle. Tool access expanded by 50%, but 84% of companies have not redesigned jobs or workflows to take advantage of AI capabilities². Access is not redesign.

Put those two together and a pattern shows up. The tools moved fast. The redesign of the work around them is moving slower, which makes sense; most organizations are figuring this out in real time, with no playbook. So adoption metrics look healthy, while the lived experience underneath stays uneven.

A muscle, not a metric

Cal Newport’s framing of “cognitive fitness” has been kicking around in my head since I started this project. He treats deep concentration as a trainable, yet depletable capacity. Like a muscle. You can build it, and you can burn it out.

Sit with that idea for a second, and the question we should be asking shifts. The question isn’t “are people using AI?” It’s “is the way they’re using AI building or depleting their cognitive fitness?”

The 7% sweet spot, in Newport’s terms, is probably the zone where AI augments expert thought without crowding it out. Below that band, the tool is underused. Above it, the tool starts doing the thinking, and the muscle atrophies.

This is why I think vanity metrics, on their own, can quietly mislead us. A user running 200 prompts a day might look like a power user on a dashboard, and a person whose attention has been shredded into 200 pieces in real life. The dashboard can’t tell those two apart. Cognitive fitness can.

Two strategies that don’t require a dashboard

I’m not going to pretend I have the org-level redesign answer figured out. That’s the slower, more patient work. But there are two things you can do this week, as an individual, that line up with where the data is pointing.

1. Use AI to support your expertise, not replace your thinking. Before you prompt, indicate what you actually think. Draft the rough version yourself, then bring AI in to pressure-test or sharpen. The order matters. When AI runs first, you’re editing someone else’s thinking. When you run first, AI becomes a real thought partner.

2. Defend your focus harder, not less. Time-blocking is straight out of Cal Newport’s Deep Work. The instinct when AI gets faster is to fill the saved time with more inputs: more tabs, more tools, more small tasks. Resist. AI generates more shallow-work noise, not less, so block the time and close the tabs.

Stepping back

What I’m starting to believe, sitting with this data, is that we measure what’s easy to measure, and there’s no shame in that. License counts are clean. Cognitive fitness is messy. But the messy stuff is what actually tells you whether the muscle is getting stronger or wearing out.

If only 3% of users are in the sweet spot, maybe the better question isn’t how to push more people into more tools. It’s how we redesign the work, and our individual habits, so the sweet spot grows over time. This is where cognitive muscle gets built.

I’m still figuring out what that looks like in practice. Probably you are too.

What’s one thing you’ve noticed in your own work that tells you AI is sharpening you, or quietly depleting you?

Let’s figure it out together. 💚

—

_{¹ ActivTrak. (2026). Workforce productivity behavioral data report.}

_{² Deloitte AI Institute. (2026). State of generative AI in the enterprise.}

^{Newport, C. Deep Work and related writing on cognitive fitness as a trainable, deplorable capacity.}