Having spent last week at Microsoft 365 Conference in Orlando, it was clear that the conversation has shifted. Copilot agents are no longer something we talk about as future potential for our organisations. They’re here, being used, and organisations are already trying to work out what it means in practice.
What stood out wasn’t just the continued push around Copilot. It was the growing focus on agents, and more importantly, how they are going to be governed. The tone has changed. We’ve moved past “what can Copilot do?” and into a much more important conversation around control. Another thing that came through strongly, especially from the keynote from Jeff Teper, was the role of SharePoint. It wasn’t positioned as a headline, but it was obvious where things are heading. SharePoint is becoming the central knowledge layer for AI. That’s where the content lives, and that’s what Copilot and agents are reading from, learning from, and acting on. When you look at it like that, the conversation about AI quickly becomes a conversation about your data, and more importantly, who has access to it.
What Microsoft Agent 365?
At a basic level, Microsoft Agent 365 is Microsoft’s way of managing AI agents across an organisation. It acts as a control layer over the top, giving visibility into what agents exist, what they can access, and what they are doing.
Microsoft first introduced the concept at Microsoft Ignite 2025 on November 18, 2025, positioning it as a control plane for AI agents. At the time, it felt important but still early. Then in March 2026, Microsoft expanded on this as part of its broader Frontier Suite vision. That’s when it became clear this wasn’t just an experiment. Agents are becoming a core part of the platform. With general availability expected from May 1, 2026, this is moving quickly from concept into something organisations will need to deal with properly.
From Copilot to agents
Up until now, most organisations have been getting used to Microsoft 365 Copilot. You ask it something, it responds. It helps with content, summaries, and analysis. It’s useful, but it’s still reactive.
Agents change that. Instead of just responding, they act. They can run tasks across Outlook, Teams, Word, and Excel, pull data together, update content, and automate processes. In simple terms, you are giving AI permission to operate inside your environment. That brings a completely different level of impact, and with that, a different level of risk.
Why Microsoft Agent 365 matters
This is where Microsoft 365 Agent 365 becomes important. It gives organisations a way to stay in control as this starts to scale.
In practical terms, it allows you to:
- See what agents exist across your environment
- Control what those agents can access
- Apply governance, security, and compliance policies
- Monitor behaviour and usage
- Understand ownership and accountability
Without that level of visibility, you effectively have AI operating inside your tenant with very little oversight.
The bigger issue here isn’t what AI can do. It’s what it should be allowed to do. Agents don’t just read data, they act on it. They can create, update, and move information around. If your permissions are wrong, your AI will behave incorrectly. It’s as simple as that. When you combine that with SharePoint’s role as the knowledge layer, the importance of getting your data and permissions right becomes hard to ignore.
The risk: agent sprawl in Microsoft 365
We’ve seen this pattern before across Microsoft 365. SharePoint brought site sprawl, Microsoft Teams brought team sprawl, and the Power Platform introduced app sprawl. Now we are heading towards agent sprawl.
Tools like Microsoft Copilot Studio make it easy for people in the business to create agents. That’s a positive from a productivity point of view, but it also means adoption will be fast and, in many cases, uncontrolled.
It doesn’t take much to see how this plays out. Multiple agents doing similar things, no clear ownership, different levels of access, and very little visibility of what is actually running. This isn’t something to worry about in a few years. It’s going to happen quickly.
A shift in responsibility
What’s also changing is responsibility. Traditionally, IT controlled the systems and the business used them. Now the business can create agents themselves, IT is still expected to govern everything, and AI sits somewhere in the middle, acting on behalf of both.
That raises a fairly obvious question. When something goes wrong, who owns it?
What organisations should be thinking about now
From where I sit, the organisations that get this right won’t be the ones that move fastest. They will be the ones that stay in control.
That starts with asking some simple questions:
- What data do we actually trust AI to act on?
- Who should be allowed to create agents?
- How are we going to monitor what they’re doing?
- Where are the current risks around access and permissions?
These are governance questions, and they need answering early, not after agents are already embedded into day-to-day processes.
Seeing SharePoint’s 25th birthday celebrated at the conference was a bit of a moment. It’s gone from a place to store documents to the backbone of how organisations manage knowledge. Now it’s feeding AI, powering agents, and effectively shaping how decisions get made. Which makes one thing pretty clear. If SharePoint is the foundation, then how well it’s been managed really starts to matter.
