The Million-Dollar Company That Never Hired a Marketer

Ernie Tedeschi, Stripe’s chief economist, put out a count in June with two colleagues that went nowhere for six weeks and then resurfaced this week. The number of solo operators on Stripe’s platform clearing $1 million in revenue more than doubled between 2023 and 2025. The ones crossing $5 million and $10 million came close to tripling in the same window.

Those are one-person companies at a revenue scale that used to require staff. Tedeschi and his co-authors, Marisa Rama and Chris Cruickshank, name the specific jobs that stopped being necessary:

“Whether it’s how to evaluate or size a market, code an app, price a product, write and execute a marketing campaign, or close a deal, AI (and AI-augmented software) can fill many of the gaps that founders previously turned to another human for.”

Note the fourth item on that list. It’s yours.

The first non-founder hire at a small software company was usually a marketer, because a founder could build the product and sell to friends but couldn’t run demand generation alone. That hire is the one the data says isn’t happening.

The second dataset is the one that matters

Stripe’s number is interesting and self-interested, which the report says out loud. Its proxy for a solo business covers roughly 115 solopreneur-focused platforms plus solo Stripe Atlas companies, and the authors concede the obvious limit:

“Stripe cannot distinguish between solopreneurs and employers among business sign-ups, but we do validate the overall acceleration and the meaningful activity of these businesses.”

When the finding recirculated on TBPN on August 4, the hosts spent several minutes around the 17-minute mark picking at exactly that. How would a payments company know how many people work at its customers? It’s the right question, and it’s why the second study is the load-bearing one.

Liya Palagashvili, a senior research fellow at the Mercatus Center at George Mason University, published a paper on July 17 using government data instead. She took Census Bureau business-formation records and Current Population Survey microdata running from early 2022 through early 2026. Then she split sectors by how exposed they are to AI and ran the comparison as an event study, holding sector, year, and quarter effects fixed.

Between the first quarter of 2024 and the first quarter of 2026, applications for businesses unlikely to ever employ anyone rose 26.8% in AI-exposed sectors, which include professional services, information, education, and finance. In the control group of construction and wholesale trade, they moved −0.4%. Solo self-employment in the 10 most AI-exposed occupations rose about 20% against a 2022-2023 baseline. In the 10 least exposed, it didn’t move.

Two different measurers, two different methods, same direction. Government data agreeing with Stripe’s data is what moves this off the marketing-blog shelf.

The mechanism is a hiring decision that never gets made

Palagashvili’s framing in her own write-up is the clearest statement of what’s going on:

“AI may allow one person to perform tasks that once required a team, support staff, or firm-based infrastructure.”

Think about what a first marketing hire was buying. Not brilliance. Coverage. Somebody who could write the emails, keep the site current, run the ads, and answer the question of what to do next week, because the founder had run out of hours. It was a purchase of capacity, priced at a salary and paid for months before it produced anything.

A founder who can get a serviceable version of that from software doesn’t sit down and decide to eliminate a marketing role. They just never open the requisition. The job doesn’t get cut. It fails to get created, which is why this shows up in business-formation statistics before it shows up in any layoff announcement. It’s the same pattern we traced through the disappearing bottom rung of the marketing career ladder and the experience debt that follows when juniors never get the reps.

What a founder-marketer should take from this

If you run marketing at a 5-to-30-person company, the thing to watch is the shape of your own next hire, not the solopreneur count.

The roles most exposed here are the ones defined by coverage rather than judgment: the generalist who executes across five channels because nobody else has time. That’s precisely the job software now does passably. The roles that hold are the ones where being wrong is expensive and somebody has to own the call.

Two honest caveats before anyone reorganizes around this. First, both datasets measure business formation and self-employment, not marketing headcount. The marketing-hire read is an inference from adjacent evidence rather than a measured fact, and nobody has published the clean number. Second, three years of data spanning a period of cheap capital and cheap tokens is thin ground for a structural claim. Palagashvili is careful to say “may” in a way the coverage of her work usually isn’t.

Still, the direction has now survived contact with two independent methods and one skeptical podcast, which is more than most AI labor claims manage.

So here’s the move. Take the job description for the marketing hire you were planning to make and split it into two lists: tasks where somebody has to be accountable for a judgment, and tasks that are coverage. Then price the second list against the software you already pay for. If the coverage list is most of the role, you weren’t planning to hire a marketer. You were planning to buy hours, and hours got cheap.