Selling the shovels

The pitch is easy to recite. Software now does the work of an employee; small businesses cannot afford employees; so a young person with a laptop and a monthly subscription can rent out the software at a margin and watch the invoices arrive. A course will explain how, for a fee.

We should say at the outset what we could not find. No regulator we can cite has counted the teenagers running “AI agencies”, and no independent survey says what they earn. What exists is a record of the sellers of the dream, which is more instructive than it sounds. The Federal Trade Commission has spent two years establishing that the people most reliably enriched by the promise of AI riches are those who sell the promise.

• After one year
• After five years

Starting a firm is a coin toss over five years. Share of new private-sector establishments still open, by year of opening, %

	After one year	After five years
1994	79.6	49.6
2000	78.4	48.2
2005	80.1	46.8
2010	78.6	51.1
2015	79.6	50.2
2019	79.2	51.5

Source: Bureau of Labor Statistics, Business Employment Dynamics, Table 7⁵

In September 2024 the FTC announced a set of actions it called Operation AI Comply. Three concerned business-opportunity schemes. Ascend Ecom, it alleged, promised that “cutting edge” AI tools would help customers earn thousands of dollars a month in passive income from online stores, charged tens of thousands to set them up and tens of thousands more for inventory, and took at least $25m.^{1Federal Trade CommissionFTC Announces Crackdown on Deceptive AI Claims and Schemes25 Sept 2024} Ecommerce Empire Builders sold training for almost $2,000 and “done for you” storefronts for tens of thousands, and advertised $10,000 a month, a figure the FTC said had no evidence behind it.^{1Federal Trade CommissionFTC Announces Crackdown on Deceptive AI Claims and Schemes25 Sept 2024 3Federal Trade CommissionEmpire Holdings Group LLC, et al., FTC v. (Ecommerce Empire Builders)9 May 2025} FBA Machine promised a “7-figure business” with AI-powered pricing tools and, the FTC said, cost customers more than $15.9m.^{1Federal Trade CommissionFTC Announces Crackdown on Deceptive AI Claims and Schemes25 Sept 2024}

None of these was an automation agency, and we do not suggest otherwise. They sold storefronts, not chatbots for dentists. But the architecture is the same: a technology that sounds inevitable, an earnings claim that cannot be checked, and a price that is paid before the first customer appears. Lina Khan, then the FTC’s chair, put the legal point briefly: “there is no AI exemption from the laws on the books.”^{1Federal Trade CommissionFTC Announces Crackdown on Deceptive AI Claims and Schemes25 Sept 2024}

The cases have since ended as such cases do. In May 2025 Ecommerce Empire Builders and its owner agreed to an order banning them from selling business opportunities and handing over assets for refunds.^{3Federal Trade CommissionEmpire Holdings Group LLC, et al., FTC v. (Ecommerce Empire Builders)9 May 2025} In June 2025 a proposed order was filed against Ascend Ecom on similar terms.^{2Federal Trade CommissionAscend Ecom (FTC matter 242 3023)23 Jun 2025} In July 2025 a stipulated order was filed in the FBA Machine case, with a permanent ban on its operator.^{4Federal Trade CommissionFBA Machine/Passive Scaling, FTC v.30 Jul 2025} Each case page still lists the matter as pending, because court approval is a formality that takes time. The money, meanwhile, had been spent.

There is an honest case for the agency model, and it should be stated properly. A competent 17-year-old who can wire a booking system to a spreadsheet saves a plumber an hour a day. Plumbers have never been well served by software firms, and a local reseller who answers the phone is worth a premium. Nobody needs a licence to be useful. If a few of these young people earn real money, that is a fair reward for doing something older rivals find tedious.

The trouble is the base rate. The customers of such an agency are, by definition, small firms, and small firms are fragile. Bureau of Labor Statistics data on establishments opened between 1994 and 2019 show that 78% to 80% survived a year, and only 47% to 52% survived five.^{5Bureau of Labor StatisticsSurvival of private sector establishments by opening year (Table 7)} An agency that sells a monthly retainer to such clients is selling to a population in which roughly half the accounts vanish within five years, before one counts those that cancel for being unimpressed. The reseller also depends on a tool vendor that can change its pricing or its terms, and on a model that a larger competitor can bundle for nothing. This is a business with thin margins, no moat and a high rate of churn. It is not forbidden. It is also not passive.

Nor do the FTC pages tell us who the buyers were. Ascend’s case page says nothing of their ages, and describes them only as people seeking online storefronts and passive income.^{2Federal Trade CommissionAscend Ecom (FTC matter 242 3023)23 Jun 2025} We will not guess. But anyone selling a get-rich-in-AI course to the young is selling to people with small savings, short histories and good reasons to hope. That is exactly the audience a prudent seller would want to treat gently, and the record suggests that some do not.

The test for any course is simple: ask what its founder earns from clients, not from students. In the last gold rush, the shovels did very nicely.