Ask a small business owner why they have not done more with AI and you will almost never hear that it is too expensive. Twenty euros a month is not what stops anyone. What stops them is a quieter thing, harder to admit in a room full of people talking about transformation.
They opened the tool. They typed something. They got back something reasonable that they were not sure what to do with. And then a customer called, and the tab stayed open for a week, and eventually it got closed.
That is the actual failure mode, and OpenAI appears to have noticed. On 21 July 2026 it launched a programme aimed specifically at small businesses, alongside a claim of 10 million ChatGPT Work and Codex users. The programme is not a product. It is a stack of teaching, which tells you something about where OpenAI thinks the bottleneck is.
What is actually in the programme
The offering breaks into four pieces, and it is worth being specific because the marketing language around it is vague.
The first is virtual training: webinars covering how to use ChatGPT Work for particular functions, with accounting, marketing, and ecommerce named explicitly. The second is in-person events, branded OpenAI Academy, running in the United States, with guided instruction and hands-on exercises alongside other business owners. The third is a content library of customer stories, interactive guides, and short videos. The fourth, and the most practically interesting, is a set of plugins, skills, and offers from partners including Dropbox, Shopify, Intuit, Slack, Atlassian, and Wix.
The training resources are free. The thing they teach you to use is ChatGPT Business, at roughly $20 per seat per month on annual billing or $25 monthly, with a two-seat minimum. That is the honest shape of the deal, and it is a completely normal shape. Nobody builds free academies out of civic feeling.
The underlying model is GPT-5.6, which is the same frontier model available on every other tier. That is worth stating plainly because it removes a common worry. This is not a cut-down small-business edition of the model. The capability is the same one large enterprises get.
Why OpenAI is doing this
The strategic logic is not hidden, and understanding it helps you judge the offer clearly rather than cynically.
OpenAI has a genuine retention problem in the small business segment, and it is not a pricing problem. Seats get bought, used enthusiastically for two weeks, and then quietly stop being opened. The subscription continues until someone reviews the card statement. That pattern is expensive for a vendor because acquisition cost is spent on a customer who never reaches the point of depending on the product.
Training attacks that directly. A business owner who has sat through a session on using AI for their monthly reconciliation, and who has done it once with help, has crossed from curiosity into workflow. Workflows renew. Curiosity does not. The partner integrations do the same job from the other direction: an AI that reaches into Shopify and Intuit is embedded in the operation rather than sitting in a browser tab waiting to be remembered.
This is not a criticism. It happens to align almost perfectly with what small businesses actually need, which is not more capability but more idea of what to do with the capability they already pay for. When vendor interest and customer interest point the same direction, the offer is usually real. It is worth knowing why it exists all the same.
The gap it is trying to close
The reason this programme is more interesting than a typical vendor launch is that the problem it addresses is the correct one.
Small businesses do not fail at AI because the models are insufficient. GPT-5.6 is drastically more capable than what almost any small business is asking of it. They fail at the translation step: knowing which of the forty repetitive things in their week is the one worth handing over first, and then getting it wired up so it happens without anyone remembering to trigger it.
That translation is genuinely hard, and it is not really a technology problem. It requires looking at your own operation with enough distance to see which parts are mechanical, which is difficult precisely because you have been doing them by hand for so long that they feel like the job rather than overhead. This is the same observation behind why so many first AI automations fail: the tool was fine, the choice of what to point it at was the failure.
A training programme can help with this, particularly the in-person format where you watch someone else describe their business and recognise your own. What a training programme cannot do is the wiring. It can show you that ChatGPT could draft your quote follow-ups. It cannot make the follow-up happen automatically at 8am on the second day after a quote goes out without you present. That gap between understanding and running is where most of the value sits, and it is exactly the part the programme leaves to you.
Who genuinely benefits
This is a good fit for a specific kind of business, and being precise about which kind saves everyone time.
If you are early, meaning you use ChatGPT occasionally and personally but nothing in your business depends on it, the programme is close to ideal. The structured content answers the question you actually have, which is not how does this work but what should I use it for, and the free tier of that answer has real value. Going from ad-hoc use to two or three reliable uses is the largest single jump available, and this is a reasonable way to make it.
It also suits businesses that live inside the named partner tools. If your operation runs on Shopify and Intuit and Slack, the partner integrations are not a marketing bullet, they are the difference between AI that sees your business and AI that you have to describe your business to every time. That distinction compounds daily.
The in-person academies are worth a separate note, with the caveat that they are running in the United States, which limits them for a European reader. Where they are available, the value is less the instruction than the room: watching another owner explain a problem you also have, and seeing what they did about it, transmits something that no webinar reliably does.
Who should skip it
If you already have AI running in production, this is not for you, and there is no diplomatic way to say that which is also useful.
Introductory training aimed at businesses taking their first steps will not tell you anything if you are already past that point. Your problems are now different in kind: routing between models to control cost, evaluating whether outputs are actually good at scale, handling the cases where the AI is confidently wrong, and deciding what happens when a workflow fails at three in the morning. None of that is what a getting-started curriculum covers, and it should not be.
You should also skip it if your real constraint is integration rather than understanding. Plenty of businesses know exactly what they want AI to do and are stuck because their tools do not talk to each other, or because the process only works if a human remembers to start it. Training does not solve that. More training definitely does not solve that. What solves it is building the connection, which is work of a different type entirely.
And there is a version of this worth naming directly, because it is common. Signing up for training can become a way of feeling like progress is happening while nothing changes in the business. If you have already watched several hours of AI content and your operation runs identically to how it ran last year, another course is not the missing piece. Picking one process and finishing it is.
The honest verdict
The programme is a genuinely good offer for people at the start of the path and largely irrelevant to everyone else. That is an unexciting conclusion and it happens to be true.
What makes it worth paying attention to is less the programme itself than what it signals. The largest AI company in the world has concluded that the constraint on small business adoption is not model capability or price, but the translation from capability to operation. It is spending real money on teaching rather than features. That is a fairly strong admission that the tools got ahead of the ability to use them, which is something most business owners already suspected and have been quietly assuming was their own failing.
It is not. The gap is structural, everybody has it, and the honest path across it is narrower than the marketing suggests. Pick one repetitive thing. Make it run without you. Then pick the next one. The training will help you choose. What happens after the choice is where your business actually changes, and no webinar has ever completed that part for anyone.