Count businesses and AI use is 18 percent. Count employees and it's 32.

The Census Bureau publishes both numbers out of the same survey. The gap between them is the clearest picture anyone has of who is actually using AI, and it isn't you.

Count businesses and AI use is 18 percent. Count employees and it's 32.

Every two weeks the Census Bureau asks American businesses whether they used AI. Over this past winter, 18 percent said yes.

Then the researchers ran the same responses weighted by employment. That number came back at 32 percent.

Same survey. Same window. Same answers. The only thing that changed was whether you count companies or count people.

Nobody writes about this because it reads like a methodology footnote. It isn't. It's the clearest picture anyone has of who is actually using this stuff.

Here's the plain version. Count every business as one vote and you get 18. Weight it by headcount and you get 32. The number climbs because the companies with the most people on payroll are the ones already using AI. Weighting drags the average toward them.

The sector detail makes it harder to miss. Census found very large companies in professional services, finance and information running at 50 to 60 percent. The national rate is under 20.

So when you read that the AI race is on and everybody's running, understand what that sentence actually describes. The companies out front have a budget line for this. They have someone whose job title includes it. A few of them have an entire team whose only assignment is working out where AI fits.

You have a Tuesday afternoon and forty other things.

That's the real gap, and I want to be precise about it, because most of the coverage gets this wrong. It isn't that they have better tools. The tools are the same ones you can open in a browser right now for twenty dollars a month, or nothing at all. It isn't that their people are sharper than yours. I've sat in enough of those rooms to tell you they aren't.

Somebody over there was told to own it. That's the advantage. The whole thing.

Which is better news than it sounds, because you can't buy their budget but you can do the one thing their budget pays for. Pick a single job in your week that eats real time and produces something predictable. Client intake. The proposal you rewrite from scratch every time. The follow-up that keeps slipping past Thursday. Give it two hours and a decision about who owns it.

That isn't a strategy. It's a Tuesday. But it's the same move the companies sitting at 50 percent made, minus the org chart.

The number worth caring about isn't 18 or 32. It's one. One workflow, actually running, that wasn't running last month.