AI isn't an app. It's an operating system for your business.
Why the next decade of small-business software won't look like SaaS — and what that means for the founders who get there first.
There’s a quiet shift happening in how small businesses use software, and most of the people selling AI tools haven’t noticed yet.
Until now, the model was: pick an app for each job. A booking app. A reviews app. A finance app. A marketing app. Each app had a UI, a learning curve, a monthly fee, and a tiny island of context — what it knew about your customer, your tone, your numbers.
The result was an inbox of dashboards. None of which knew about each other. None of which knew you.
The OS metaphor isn’t a marketing line
When iOS launched, “an iPhone app” was a real thing — distinct, contained, on its own grid square. Now? An iPhone app is a thin shell over the OS’s capabilities. The interesting work is in the OS: the shared identity, the shared notifications, the shared file system, the shared intelligence.
The same pattern is repeating for business software. The interesting work is no longer in the app. It’s in the layer underneath that knows your business, has shared context across surfaces, and can act on your behalf when you sleep.
That layer is what we’re building. We’re not asking small businesses to switch to a new app. We’re asking them to install an operating system, and then watch their existing tools become better because the OS handles the coordination between them.
Why the non-technical founder wins this round
If you’re a Cambridge-educated engineer who can stitch ten tools together with custom code, you’ve already had a “personal AI” for two years. You wrote it yourself.
But there are tens of millions of small business owners who don’t write code — and who, until very recently, had no way to access the same productivity boost.
The people who’ll capture the most value from this transition are not the engineers. They’re the non-technical founders who learn to direct the AI rather than build it. Direct in the sense of:
- Tell it what your business is, in your own words.
- Watch it propose a plan; reject the bits that don’t fit.
- Approve drafts until the tone is right; then trust it on autopilot.
- Catch it when it’s wrong, once a week, in five minutes.
That’s the new skill. Not prompting. Not engineering. Directing.
What we believe about the next decade
A few honest predictions, none of them safe:
-
Most “AI for SMB” startups will become features inside the OS layer. The single-purpose AI chatbot is a 2024 product category. By 2027 it’s a checkbox.
-
The winners look like Microsoft, not OpenAI. They sell a platform, not a model. They make money from the coordination, not the inference.
-
Trust beats capability. The model that always responds, never invents, and remembers the boundary is more valuable than the model that’s two points better on a benchmark.
-
Personalisation is the moat. A generic AI is interchangeable. An AI that has spent six months learning your business is not.
-
The owner does less, not more. This is the whole point. The owner runs the business; the OS handles the chasing, the drafting and the reporting. If we make the owner work harder, we’ve built it wrong.
We’re building the OS for that world. We’d love it if you joined us.