Microsoft brings Claude Sonnet 5 into Microsoft 365 Copilot

Microsoft has added a second AI model to Microsoft 365 Copilot. On 2 July 2026, the company announced that Anthropic's Claude Sonnet 5 is now available inside Copilot Cowork and PowerPoint. That sounds like a straightforward upgrade. In practice, it creates a genuine decision point for SME teams that are only just finding their footing with Copilot.

A second model arrives inside Copilot

Claude Sonnet 5 is Anthropic's latest model, designed specifically for complex, multi-step tasks. It handles longer chains of reasoning than a general-purpose assistant, makes plans across several workflow stages, and produces polished, structured documents with fewer manual corrections.

Before this integration, Microsoft 365 Copilot ran on Microsoft's own models. Users had one AI engine and no choices to make. Sonnet 5 changes that. It sits alongside the existing model and is available for specific tasks in Cowork and PowerPoint.

For SME IT administrators, access is managed through your existing organisation policies. No new licences are required. Rollout is progressing across regions and tenant configurations, so availability may vary depending on your setup.

Why this matters for SME IT administrators and office managers

AI adoption inside any organisation is 80% people and culture, 20% technology. Adding a second model to a platform your team is still learning does not automatically make things easier. It adds a layer of judgement that was not there before.

SME IT administrators now need to decide which model is the default, who can switch between them, and whether any governance policies need updating. These are not difficult decisions, but they do require deliberate attention. Without clear guidance, team members will either ignore Sonnet 5 entirely or use it inconsistently.

For office managers overseeing day-to-day document production, the practical upside is real. Sonnet 5's stronger logical reasoning and context handling means complex presentations and structured business documents can be drafted with fewer back-and-forth corrections. Tasks that previously required several manual review steps become more straightforward. That is time returned to the people doing the work.

Three things your team can do this week

1. Check your tenant configuration. Log into the Microsoft 365 admin centre and confirm whether Claude Sonnet 5 has rolled out to your organisation. If it has not appeared yet, note it for your next IT review cycle.

2. Review your Copilot usage policy. Spend 20 minutes updating your internal AI guidance to acknowledge that Copilot now offers model choice. Specify which tasks, if any, should use Sonnet 5 and which should use the default model. This quick step prevents confusion before it starts.

3. Run a short test in PowerPoint or Cowork. Pick one real document your office manager or a team member produces regularly. Ask them to generate a draft using Sonnet 5 and compare it against their usual output. This takes under 30 minutes and gives you a concrete reference point for your own organisation.

Considerations and limitations

Claude Sonnet 5 is rolling out progressively. Not every Microsoft 365 tenant will have access immediately. Availability depends on your region and how your tenancy is configured, so check before communicating the change internally.

The integration sits within the Microsoft environment, which means your existing data handling and security policies apply. That is a genuine reassurance for UK SMEs concerned about where data goes. However, it is worth confirming with your IT administrator that no policy exceptions have been introduced as part of the rollout.

Model choice also introduces a new kind of complexity. A platform with one AI model is simpler to govern than one with two. SME teams without a clear framework for when to use which model risk inconsistency across departments.

Where to take this next

If your organisation is still working out how to govern AI model choices inside Microsoft 365, that is a common position and a sensible one to address now rather than after the inconsistencies compound. The AI Readiness survey takes a few minutes to complete and gives you access to 65 or more free resources alongside a custom AI Readiness report. That report then leads to a free 45-minute AI Readiness call to walk through the results with you directly.

gecco's Training and consultancy programme is built around exactly this kind of decision point: model choice, governance frameworks, and upskilling the people who manage and use these tools day to day. If you want help building that clarity inside your team, that is what we do.

Website · LinkedIn · Case Studies · Newsletter