If you only watched the Ignite keynotes, you could come away thinking this year was about spectacle: agents everywhere, Copilot doing more, and a wave of “IQ” layers on top of your data.
From a Microsoft 365 admin point of view, the story is sharper than that. Ignite quietly made AI agents first-class citizens inside your tenant, and it made it much easier for those agents to find and act on the content you already have.
At the same time, most of the admins we spoke with in San Francisco were still wrestling with the classics: SharePoint storage limits, OneDrive and Teams sprawl, and permissions that are hard to unwind once they've spread.
If your governance is strong, this is the start of a powerful new phase. If you're still fighting oversharing, ownerless workspaces, ROT, and storage growth, the risk curve just bent upward.
Here are our takeaways after an exciting week at Ignite 2025.
Microsoft introduced Agent 365 as a control plane for AI agents. In their words, it is a way to deploy, organize, and govern agents securely, regardless of whether they are built on Microsoft platforms, open-source frameworks, or third-party tools. It brings together a registry, access control, observability, interoperability, and security in one place.
At the same time, Microsoft is rolling out native agent infrastructure, including agent connectors, an agent workspace, and new ways for agents to run in contained, policy-controlled environments. These pieces let agents connect to apps through Model Context Protocol, operate under clear guardrails, and act on your behalf with built-in governance hooks.
For admins, this means:
That’s a new governance surface. It’s no longer enough to know who your users are and which workspaces they own. You also need to know which agents exist in your tenant, what they can see, and who is accountable for them.
If you don’t have a clean handle on registry, permissions, and lifecycle for human-owned workspaces, adding agents on top will create an agent-shaped version of shadow IT.
Ignite also introduced or expanded a stack of “IQ” layers that give Copilot and agents better context over your data.
Work IQ extends the intelligence behind Microsoft 365 Copilot so it can understand how users work, who they work with, and what content they use. Fabric IQ and Foundry IQ push this idea across analytics and line of business data, tying multiple sources into a unified semantic model and retrieval layer.
In plain terms, these layers make it easier for AI to:
That’s only helpful if the content and access model are in good shape. If your SharePoint and Teams estates are full of overshared sites, stale projects, and sensitive data that should have been archived, you’re giving Copilot and agents a lot of bad inputs.
Better context doesn’t distinguish between “useful and safe” and “risky and unnecessary.” It simply gets you to whatever exists, faster.
This is where basic hygiene work matters more than ever:
If you want to use these new IQ layers confidently, you need to know the data they see is the data you actually want in play.
Microsoft also announced that Security Copilot is now included for Microsoft 365 E5 customers at no additional charge. This is a clear signal that AI-assisted security operations are moving from pilot to mainstream in larger tenants.
Security teams with E5 will be able to adopt AI quickly. They will get richer threat analysis, faster investigation, and better recommendations across Defender, Entra, and Purview.
The problem is that most organizations don’t have collaboration governance at the same maturity level. They still have:
The result is a gap: sophisticated AI-driven security on top of a collaboration estate that is still noisy, loosely governed, and growing in all the wrong places. Security Copilot may highlight risks faster, but it cannot invent ownership, labels, lifecycle, or sane storage policies where none exist.
Closing that gap means aligning security and collaboration work. Security can’t be an E5-only AI project while Teams, SharePoint, and OneDrive hygiene remain an afterthought.
Microsoft showed Copilot generating complete SharePoint pages from a single prompt. The demos included layouts, text, charts, and live components.
From a productivity perspective, this is great. It kills a lot of blank page time and lets teams get to “something reasonable” very quickly.
From a governance and storage point of view, it looks like a rerun of an old movie, just at a faster speed:
If you don’t have clear templates, approvals, and lifecycle rules, you'll see the classic SharePoint mess recreated in a fraction of the time. Copilot will help people publish more, but it won’t decide when something should be archived or deleted.
This is exactly where provisioning guardrails and lifecycle automation pay off. You want Copilot to plug into a model where every new workspace has:
Otherwise, every “instant site” is a future cleanup task.
If you’re an admin or platform owner, you can’t control Microsoft’s product roadmap, but you can control the state of the tenant these agents and Copilots will live in.
Orchestry Health Checks help you monitor workspace provisioning, archivals, storage savings, and more.
Over the next quarter, focus on:
At Orchestry, this is exactly where we’re focused. Our platform already surfaces ownerless and overshared workspaces, highlights risky links, and supports archive workflows for Teams and SharePoint. Our Health Checks and Recommendations features are designed to give you the same kind of “IQ layer” for your governance work, so you can see tenant health and act on it with less guesswork.
Ignite 2025 didn’t change the basics: you still need clear ownership, least privilege, and a lifecycle that includes more than creation. It did change the stakes. AI agents will sit on top of whatever you have today, and they’ll move faster than any human ever could.
Now’s the time to make sure the tenant your agents inherit is one you actually want them to learn from.
You can’t prepare for AI adoption without preparing your governance. When permissions are clean, Copilot and agents become a productivity multiplier. When they’re not, they act as a mirror, showing exactly where your risks live.
See how Orchestry helps organizations restore control and prepare for AI-driven collaboration. Start a 28-day trial to uncover hidden risks in your tenant.