Two years ago the AI conversation in most organisations was about whether to roll out Copilot. Today the question is a different one: how many agents do we actually have running, who built them, what data can they touch — and what are they costing us? In most places I talk to, nobody can answer that with confidence.
That’s the gap AvePoint is aiming at with AgentPulse, and it’s why I think it deserves a place on the shortlist for anyone running Microsoft 365 or Google Cloud at scale.
What AgentPulse is
AgentPulse Command Center is a module in the AvePoint Confidence Platform. At its core it’s an agent registry: a single inventory of the AI agents in your environment, regardless of how or where they were created. It was announced in November 2025, went through a private preview, and became generally available in March 2026 — at the same time gaining support for Google Cloud alongside Microsoft 365.
Coverage today includes the agent sources most organisations are actually using:
- Microsoft Copilot Studio (including the “lite” agent builder in Copilot)
- SharePoint agents
- Microsoft Foundry (formerly Azure AI Foundry) and the Microsoft 365 Agents Toolkit
- Third-party agents from the Microsoft agent store
- Google Vertex AI
For each agent you get visibility into who created it, where it’s published, what it has access to, how much it’s being used, and how the inventory is trending over time. The October 2026 release added an “origin” view that separates first-party, third-party and other publishers — useful when you’re trying to figure out which agents your own people built and which ones came in through the side door.
Why it’s interesting
1. Shadow AI is the new shadow IT
Building an agent in Copilot Studio or SharePoint takes minutes and requires no involvement from IT. That’s the whole point — and also the problem. Business users will (rightly) build agents to solve their own problems, and those agents inherit whatever data access their creator or their connectors have. If you can’t see them, you can’t govern them. An inventory is step one, and it’s a step most organisations haven’t taken yet.
2. Agents make oversharing worse — fast
Every Microsoft 365 tenant I’ve seen has some degree of permission sprawl: old sites shared with “Everyone”, Teams that were never cleaned up, sensitive files in the wrong library. Before AI, that sprawl was mostly hidden by the fact that nobody went looking. An agent grounded on that data will go looking, every time someone asks it a question. AgentPulse ties agent activity to the data being accessed, flags agents touching sensitive content, and — because it sits on the same platform as AvePoint’s access and governance tools — lets you fix the underlying permissions from the same place.
3. Pay-per-use AI needs cost visibility
This is the part I think is most underrated. A lot of agent consumption is metered, and the bill shows up after the fact. AgentPulse surfaces the high-activity and high-cost agents so you can decide whether they’re earning their keep — or whether it’s time to retire them. That turns AI spend from a surprise into something you can actually manage and report on. Finance will thank you.
4. Most organisations are multi-cloud whether they like it or not
Microsoft and Google are both building governance tooling into their own platforms, and you should use it. But each vendor naturally governs its own house first. If you have agents in Microsoft 365 and Vertex AI, a neutral layer that spans both — and works across tenant boundaries — has real value. AvePoint also positions AgentPulse as a way to make use of the controls Microsoft already provides, rather than requiring additional Microsoft licensing to get there.
What to keep in mind
No tool solves governance on its own, and AgentPulse is no exception. A few things worth considering before you buy:
- It’s a platform play. The value is highest if you use (or plan to use) other parts of the AvePoint Confidence Platform — that’s where the remediation lives. As a standalone inventory, compare it carefully against what you get natively.
- Coverage follows the big platforms. Microsoft and Google agent sources are covered; if your agents are mostly built on other frameworks or custom code, check the roadmap.
- Visibility without ownership is just a nicer report. You still need someone who decides what an acceptable agent looks like, who approves it, and when it gets switched off.
- Vendor statistics are vendor statistics. AvePoint’s own research says three out of four organisations using AI had a data breach in the past year. The direction is probably right; treat the exact number with the usual skepticism.
Bottom line
We’re moving from AI experiments to AI in production, and agents are where that shift is happening fastest. The organisations that handle it well will be the ones that treat agents like any other governed asset: inventoried, owned, reviewed and costed. AgentPulse is one of the more complete attempts I’ve seen at giving IT and security a single place to do that across Microsoft and Google — and if you haven’t got a clear answer to “how many agents do we have?”, it’s a good place to start the conversation.
If you’d like a second opinion on how to approach agent governance — or on whether a tool like this fits your landscape — feel free to get in touch.
Sources: AvePoint launch announcement (Nov 2025), AvePoint GA announcement (Mar 2026), AvePoint October 2026 updates, ChannelE2E.
Leave a Reply