Growing Businesses and Shrinking Ones Are Making Opposite Bets on AI

In plain terms: Most small businesses now use AI in some form, up from about a third three years ago. And when you split the survey between businesses that are growing and businesses that are shrinking, the growing ones use it far more. This is about what that does and does not prove.

Here is the finding, and then the honest argument about what it means.

The US Chamber of Commerce's 2026 research found small business AI adoption has climbed to the large majority, up from roughly a third in 2023 (report). That is one of the fastest adoption curves ever recorded for a business technology. Faster than smartphones. Faster than broadband.

Split that by how the business is doing and the gap opens up: businesses describing themselves as growing adopt at a dramatically higher rate than businesses describing themselves as declining.

The part where I argue against my own headline

That gap is a correlation and I am not going to pretend otherwise.

Growing businesses have money to spend, time to experiment, and usually an owner with enough slack to try something new. Declining businesses have none of those. You could tell this entire story backwards, where growth causes adoption instead of the other way around, and the data would look exactly the same.

Anyone who shows you that split and says "so adopt AI and you will grow" is selling something. I do this for a living and I still cannot tell you the arrow direction from a survey.

What it does tell you

Something quieter and more useful: your competitors moved, and they moved fast.

Three years ago, using AI in a small business was unusual enough to be a differentiator. Today it is close to table stakes, and the people who got there early are not talking about it as an advantage anymore. They are just quietly answering calls faster than you are.

That is the real content of the number. Not "AI causes growth," but "the thing that used to make you unusual now makes you normal, and being the last one without it is starting to cost something."

The trap on the other side

The mirror-image mistake is adopting because everyone else did. That is how a business ends up paying for four overlapping subscriptions nobody opens, which I see constantly.

Adoption stats measure whether people bought something. They do not measure whether it worked. A business that signed up for three AI tools and uses none of them counts as adopted in every one of these surveys.

So the useful version of this is not "catch up." It is: pick the one job in your week that costs real money and can be counted, fix that, measure it, and ignore the survey entirely. One working system beats five accounts, and it is not close.

If you want help finding which job that is, that is the whole point of the free game plan.