Microsoft has quietly rebuilt how Copilot picks its brain. Across the last few weeks, Microsoft 365 Copilot’s model picker has moved out of a corner menu and into the prompting box itself, now split into a GPT section and a Claude section, with Claude Opus 5.5 and GPT-6 Sol rolling out together across Word, Excel, PowerPoint, Chat, Cowork and Copilot Studio. There’s also a new reasoning-effort slider letting you choose a faster, cheaper pass or a slower one that thinks harder about the task. For UK SMEs who’ve spent 2026 being told to “just use Copilot,” this is the first time the software itself is asking them to make a decision they’ve never had to think about before.

Most small business users have no idea models even differ, let alone which one suits which task. Left on default settings, that’s fine for casual use. But for a business actually trying to get consistent value out of AI in its daily workflow, understanding what changed and why it matters is worth ten minutes.

Why Microsoft is giving you a choice at all

Different AI models genuinely behave differently on the same task. Historically, Copilot ran on OpenAI’s models only, and Microsoft controlled the entire experience end to end. Now, with Anthropic’s Claude sitting alongside GPT inside the same interface, Microsoft is betting that businesses want the freedom to route work to whichever model handles it best — long, careful documents to one, quick drafts or code assistance to another — rather than being locked to a single vendor’s strengths and blind spots. That’s a meaningful shift in how “your AI assistant” actually works under the bonnet, even though the Word or Excel window in front of you looks identical either way.

What this means for your business right now

If nobody in your organisation has touched the model picker, you’re still on whatever Microsoft sets as default, and that’s a reasonable place to stay for basic email drafting or meeting summaries. But if you’ve noticed Copilot producing oddly generic first drafts, missing nuance in longer documents, or getting confused on multi-step requests, it’s worth having someone actually open the picker and try the same prompt against both GPT and Claude options. The differences can be genuinely useful: some teams are already finding Claude a stronger fit for careful, structured writing and policy documents, while GPT-6 Sol remains a solid all-rounder for quick turnaround work. This isn’t a decision to make once and forget — it’s worth revisiting as both model families keep updating through the rest of the year.

Don’t let this become another unmanaged AI tool

The bigger risk isn’t picking the “wrong” model, it’s nobody in the business deciding anything at all, leaving every employee to quietly develop their own habits and preferences with no consistency, no one checking output quality, and no policy on what’s appropriate to paste into a prompt in the first place. A five-minute team conversation about which model to default to for which task type, and what should never go into either one, closes that gap. If your business is serious about getting real productivity out of Copilot rather than just having access to it, this is exactly the kind of practical AI adoption work firms like BuildApps help UK SMEs get right — the tooling decision is easy, using it well is the actual work.

The takeaway

Microsoft’s new model picker isn’t a headline feature, but it’s a real one: your Copilot subscription can now think in two genuinely different ways, and which one it uses by default may not be the best one for your business. Take ten minutes this week to check what’s set, try the same task both ways, and decide deliberately rather than by accident.