Apple's Siri now runs on Google's Gemini. Here's what it means for your business

What Apple actually announced

Strip away the design changes (a glassier interface, fresh versions of every operating system, macOS now called "Golden Gate") and the AI news came down to four things:

• The new Siri runs on Google's Gemini, a partnership Apple first signalled in January 2026.
• Siri can now see your screen, read a screenshot, open a file and act on what it finds. It does things, rather than just setting timers.
• Apple leaned into agents. Its Passwords app will visit websites on your behalf and quietly fix weak passwords.
• The new Siri ships in English first. It is delayed in the EU under the Digital Markets Act, but it works in the UK.

Why even Apple is renting its AI

Apple has more cash, more talent and more devices than almost anyone alive. If a single company could build a world-beating AI model in-house, it is Apple. It chose not to.

That decision cuts through most of the noise in the market. Building and running a frontier model is brutally expensive, and the field moves too fast to win on the model alone. So Apple did the sensible thing. It rented the best brain on offer and spent its own energy on the parts only Apple can do: the device, the integration, the experience, and the mountain of context it already sits on.

The moat isn't the model. It's your data.

This is the part that matters for your business. Gemini, GPT, Claude, Llama: they are all turning into commodities. They get cheaper and better every few months, and you can swap one for another. Three things do not commoditise:

• Your data. The documents, history and know-how only your business holds.
• Your workflows. The way work actually moves through your company.
• Control. Where the AI runs, who can see your data, and whether it meets your rules.

So "which model is best?" is the wrong question. The businesses that pull ahead are the ones that get their own data and processes into a shape AI can use, and keep control of them. That is the whole idea behind a private, custom AI model: take a strong model, make it yours by training it on your data, and run it where you decide.

Agents just went mainstream

There was a second signal at WWDC that is easy to miss. Apple shipped an agent. Its Passwords app does not just answer a question. It goes off and completes the job. When Apple makes AI that takes action feel normal for hundreds of millions of people, "AI agents" stop being a buzzword and start being an expectation.

For an SME, the useful version is not a password fixer. It is an agent that chases overdue invoices, drafts the weekly client update, or flags the deal going quiet, with a person signing off before anything leaves the building. That is the gap between a tool you have to prompt and a result that lands on your desk, which we got into in Copilot vs ChatGPT vs a managed service.

A privacy footnote that is really about strategy

Notice that the new Siri is delayed in the EU over the Digital Markets Act, yet live in the UK. Even Apple has to bend its AI plans around data and competition rules. For a regulated UK firm in law, accountancy, wealth or care, the same logic runs in reverse. The safest way to use AI on sensitive data is to keep it under your control, in your own environment, under UK rules. Pouring client data into whatever public chatbot is trending is the opposite of that.

What to do about it on Monday

Start properly, not by panic-buying the tool in the headlines:

1. Stop shopping for models. Treat the model as a commodity you can switch later.
2. Make your data readable. Find the systems where your real knowledge lives.
3. Pick one or two workflows where AI would save real hours.
4. Decide what has to stay private, and build those on a model you control.
5. Keep a human in the loop before anything acts on its own.
6. Measure it in hours back, not headlines.

Apple just told the whole market that even it would rather rent the brain and own the experience. For UK businesses the takeaway is the same. Don't try to out-build Google or OpenAI. Own the two things they cannot touch, your data and the way your business runs, and put a model you control on top.

That is exactly what we do at Nerdster: private, custom AI for UK SMEs, built on your data, run under UK rules, with a fractional Head of AI to own it month to month.