> ## Content Index > Fetch the complete content index at: https://varops.com/llms.txt > Use this file to discover other available public pages before exploring further. # The adoption number in your board deck is wrong - URL: https://varops.com/the-adoption-number-in-your-board-deck-is-wrong/ - Published: 2026-06-30T10:24:16.000Z - Updated: 2026-06-30T10:24:16.000Z - Description: Everyone is using AI for everything? Microsoft sells Copilot and still can only find about a third of us. The adoption number in your board deck is a bubble reading - here's the one the data supports. - Author: Nix Nullty - Tags: Overhyped *Every AI strategy this year rests on a number almost nobody checked: that everyone, more or less, is already using this stuff.* [***Nix Nullty***](https://varops.com/columnist/nix/) *went looking for the receipt and found the opposite hiding in plain sight - in Microsoft's own telemetry, of all places. The company selling Copilot can locate about a third of us. The other two thirds are occasional users and outright holdouts. Before you price another renewal or greenlight another "AI-first" reorg on the assumption of a wave, read what the measured data actually says the room looks like. —* [*Muximus*](https://varops.com/columnist/muximus/) Here is the number under most of this year's AI strategy: everyone is using AI for everything, and the stragglers are about to be roadkill. Here is the number the data actually supports: roughly a third of US adults use AI regularly, a third dabble, and a third don't touch it. One of those numbers is in your board deck. It's the wrong one, and you may be repricing seats and reshuffling org charts around it. ## The claim, stated like its believers mean it Give the claim its best day in court. In June 2025 the New York Times Magazine ran an AI issue whose intro, lifted from the Hard Fork podcast, was headlined "Everyone Is Using A.I. for Everything." ([Gabriel Weinberg quotes and links it here](https://gabrielweinberg.com/p/people-are-consuming-ai-like-they?ref=varops.com).) Two assumptions ride along: that trying AI converts you into using it for everything, and that adoption is now basically universal. Inside a tech company, this feels obviously true. AI is in the editor, in the helpdesk, in every vendor deck's up-and-to-the-right slide. If your entire working world runs on the stuff, "everyone is using it" isn't hype, it's just Tuesday. The catch is that the people saying it and the people nodding along are standing in the same small room. ## The receipt: four sources, one shape This isn't one contrarian poll you can wave off. It's four independent sources - two that measure what people actually do, two that ask them - all landing in the same place. Start with the witness least motivated to lie in this direction. Microsoft sells Copilot; it has every incentive to make AI adoption look enormous. Its own "United States AI Diffusion" report, out in May 2026 and built on anonymized Microsoft telemetry, still can only find that [more than 30% of the US working-age population uses AI, up three points from the end of 2025](https://blogs.microsoft.com/on-the-issues/2026/05/28/united-states-ai-adoption-shows-steady-growth-but-distribution-remains-uneven/?ref=varops.com). When the company with the strongest reason to inflate the figure tops out at "more than 30%," that leaves close to 70% who don't. The same report has it lumpy - about 32.9% in metro areas versus 16.2% rural - and ranks the US 21st in the world. And this is counted usage, not a vibe: the bar is a usage-time threshold the report sets at 90 minutes a month across the major AI tools. A second meter reads the same. Working from Datos desktop data, SparkToro found [about 21% of users had turned into heavy AI users, 10-plus sessions a month, with growth already slowing](https://sparktoro.com/blog/new-research-20-of-americans-use-ai-tools-10x-month-but-growth-is-slowing-and-traditional-search-hasnt-dipped/?ref=varops.com). That one's 2025 data, a little older than Microsoft's, but it's a different method arriving at the same block: heavy, regular use is a minority habit. Now the people who ask instead of count, and they don't rescue the claim either. The Searchlight Institute (survey by Tavern Research, 2,301 adults, plus or minus 3%) found [58% of Americans have used or tried AI - split almost evenly between regulars at about 30%, a few times a month, and infrequent users who've touched it once a month or less](https://www.searchlightinstitute.org/research/americans-have-mixed-views-of-ai-and-an-appetite-for-regulation/?ref=varops.com). "Has tried it" and "uses it" are different populations, and the splashy headline number quietly welds them together. Even the cohort most fluent in this stuff is stalling. Gallup's Gen Z tracking shows [adoption flat year over year while the mood curdles: anger about AI up from 22% to 31%, excitement down to 22%](https://news.gallup.com/poll/708224/gen-adoption-steady-skepticism-climbs.aspx?ref=varops.com). Reading the same Gallup data, Weinberg clocks roughly a fifth of Gen Z who never use AI and about a third who use it monthly at most. When your most AI-native generation has plateaued and started souring, the universal-adoption story is losing at its home game. Stack the four and you get the synthesis Weinberg published on June 13, 2026: [about a third of US adults use AI actively, a third occasionally, a third not at all](https://gabrielweinberg.com/p/people-are-consuming-ai-like-they?ref=varops.com). His analogy is meat: most people eat some, plenty are cutting back, a real minority abstain. Name the interest while we're here - Weinberg runs DuckDuckGo, which ships an optional, privacy-first AI, so "not everyone wants AI" is a tune he's happy to whistle. Which is exactly why the weight sits on the independent numbers he's stacking, not on his reading of them. And to be fair to the figure: one-third/one-third/one-third is an approximation across studies that define "use" differently, not a single clean measurement. Every one of them still points the same way. ## Two distinctions that do all the work First, trial is not adoption. The person who opened ChatGPT once in March counts in a dozen "have used AI" headlines and is nobody a buyer has in mind when they assume the market runs on this. Second, the line is flat. Microsoft's three points and SparkToro's slowing curve say adoption has barely moved in six months, while the sentiment numbers head the other way. That's a plateau with a bad mood, not a hockey stick. ## Credit where it's due None of this says AI doesn't work. For the third who adopted it and stayed, it's genuinely useful - which is precisely why they show up as heavy, regular users in the telemetry. The thing getting punctured here isn't "AI is useful." It's "everyone is using AI for everything." The first is true for a minority. The second isn't true for the population, and saying it louder won't change the count. ## The Hype-o-Meter "Everyone is using AI for everything": four out of five sirens. The tools are real and the minority using them is real. The word doing the lying is "everyone." ## What the real number costs you Treat any "universal adoption" figure in a board deck or a vendor's ROI model as a bubble reading until proven otherwise. The measured floor is a minority, unevenly spread, roughly flat for half a year, and under no obligation to grow on your vendor's timeline. That's the math on a [per-seat renewal](https://varops.com/what-are-you-actually-paying-for/) priced as if every seat becomes an active user. And that's the urgency behind an "AI-first" reorg justified by a wave that, the second you step outside the building, is about a third of the room.