Gallup's Q2 2026 tracking says the divide is usage variety: 45% of employees using AI for one or two purposes report productivity gains, rising to 66%, 78%, and finally 90% among those using it for seven or more. That matches Davenport and Mittal's All-in On AI thesis — dabbling doesn't transform anything; gains compound when AI touches many specific tasks.

I keep meeting two kinds of operators, and the gap between them is widening. The first kind opened a ChatGPT tab sometime in 2024 and still uses it the same way — draft an email, summarize a document, ask a question. The second kind has AI threaded through scheduling, code review, meeting prep, data pulls, and a dozen small workflows nobody else sees. Both call themselves AI users. Gallup just published numbers on how differently that is paying off.

In its Q2 2026 workforce tracking, released in August 2026, Gallup found that 47% of U.S. employees now say their organization has integrated AI tools, up six points from 41% the previous quarter. More than half of U.S. workers (52%) use AI in their role, 30% use it a few times a week or more, and 15% use it daily. The most common uses are exactly what you would guess: writing and editing (51% of AI users), search or research (49%), and general assistance or problem-solving (39%).

Here is the part worth your attention: the most common uses are the least productive ones. Among people who use AI for coding assistance or process automation, 77% say it has had a positive effect on their productivity. Presentation building comes in at 76%, data analytics at 75%. Writing — the thing half of us lean on it for — scores 68%, and search scores 65%. The variety gradient is steeper still: 45% of employees using AI for one or two purposes report productivity gains, 66% at three or four purposes, 78% at five or six, and 90% among those using it for seven or more.

I read that gradient and immediately thought of All-in On AI, where Thomas Davenport and Nitin Mittal found that fewer than 1% of large companies are genuinely AI-fueled — the rest sit on a maturity ladder from Underachievers (experiments, nothing deployed) through Starters (a plan, little in production). Their point was that transformation never comes from a pilot; it comes from AI touching many functions until the organization becomes a learning machine. That ladder was written for enterprises, but it maps uncomfortably well onto individuals. Most operators I coach are personal Starters: one assistant, two habits, plateaued gains — and they wonder why the productivity revolution feels like someone else's story.

Two honest caveats before you rewire your week. These are self-reported productivity ratings, and Gallup itself flags that the variety correlation does not prove causation — people who get value from AI go looking for more places to use it, and some jobs simply offer more surface area. And individual gains are not organizational gains: in Gallup's companion culture research, 99% of 102 surveyed CHROs called AI important to strategy while 50% said they are not confident their managers can guide employees' AI use, and employees at AI-adopting workplaces split almost evenly on whether culture improved (24%) or worsened (25%) over the past year. A lot of 2026's AI productivity is freelancing — individually real, organizationally invisible.

What I actually run — with myself and with coaching clients — is a monthly thirty-minute exercise I call the Variety Audit:

  1. Pull last week's calendar and task list and write down the ten recurring tasks that genuinely consume your hours — from records, not memory, because memory over-reports deep work and under-reports admin.
  2. Mark each task where AI already helps. Most people discover they are at two: writing and search, Gallup's 51% and 49% — the low-payoff end of the curve.
  3. Choose the two most task-specific, unaided items — the coding-and-automation end of the spectrum where 77% report gains: data pulls, meeting prep, code review, invoice chasing, scheduling.
  4. Run a two-week trial on each with a written kill criterion (minutes saved per week, or errors introduced), and drop anything that fails it without sentimentality.

The 90% group is not more enthusiastic about AI; they wired it into more specific places and kept what survived contact with real work. When I ran my own audit in July, one trial failed outright — AI-generated meeting prep against my own notes was slower than doing it by hand — and I dropped it. That is the point. Not using AI more, and not paying subscription prices for a glorified writing assistant, but moving your usage toward the unglamorous, task-specific corners of the week where the measured gains actually live.

Sources: Gallup — AI Use at Work: Organizational Adoption Jumps Six Points (Q2 2026 data); Gallup — AI's Effect on Workplace Culture; TNND wire coverage — AI at work puts new pressure on managers (August 2026); Thomas H. Davenport and Nitin Mittal, All-in On AI (Harvard Business Review Press, 2023).

Related: How to Find Your Passion · Best Self-Improvement Books · How to Make Better Decisions · What University Will Not Teach You