If you haven’t already, check out Arun Ulag’s hero blog “Microsoft Build 2026: Building Agentic Apps with Microsoft Fabric and Microsoft Databases” for a complete look at all of our Microsoft Build announcements across our Fabric and database offerings.
_______________________________________
Operations agent in Microsoft Fabric are autonomous agents that watch your data, reason about what’s happening, and take action when something needs attention. Operations agent moves your team from dashboards-and-alerts to outcomes. You describe the goal; the agent does the monitoring, diagnosis, and with your approval the work.
Capabilities with operations agent
Provide operations agent instructions in plain language, and the agent then:
- Generates the rules it needs to monitor your data, grounded in the semantics of your Fabric IQ Ontology so it understands what orders, pipeline, and SLA mean in your environment.
- Watches continuously using Real-Time Intelligence, so issues are caught as they emerge rather than at the next dashboard refresh.
- Investigate when something looks off, correlating signals across your data to build a picture of what’s going on.
- Acts, either by surfacing a recommendation or, if you’ve authorized it, by running the action itself.
It’s the difference between being told there’s a problem and having the problem already in motion toward resolution by the time you read the notification.
What’s new
Smarter reasoning and a creator that asks before it guesses
Operations agents use LLMs to generate the rules it monitors, and pairs that with a chat-based creator experience that treats setup as a conversation. Together they produce agents that reflect what you meant:
- Better interpretation of vague or compound instructions like “keep latency reasonable across the EU tenants”—fewer false positives from over–eager thresholding, and rules that hold up as your data shape evolves.
- Clarifying questions instead of silent guesses. When your instructions are ambiguous—“monitor revenue” could mean a dozen things—the agent stops and asks rather than picking an interpretation and being quietly wrong.
- Preview before you commit. Preview the agent’s output, adjust it in plain language, then save the rules.
Use chat to refine how the agent playbook and rules are created
Taking action across Fabric
The operations agent now does more than notify. With your authorization, it can:
- Run Fabric pipelines to refresh data, reprocess failed loads, or kick off downstream work.
- Execute notebooks to run diagnostics or remediation logic you’ve already built.
- Invoke user data functions (UDFs) to perform targeted, parameterized actions against your data, and invoke systems outside of Fabric.
- Kick off workflows in Power Automate to take action and communicate across your company.
These actions are scoped, audited, and reversible by design. The agent doesn’t get powers you didn’t grant it.
Bringing the team into the loop with Teams
Operations agent can now post directly into a Microsoft Teams channel. When something fires, the agent shares what it saw, what it thinks is going on, and what it recommends— right where your operations team already works.
- Insights show up in context. Charts and summaries posted to your operations team’s channel.
- Recommended actions, ready to run. The agent proposes next steps; an authorized teammate can approve and trigger them from the message.
- One source of truth for the team. Everyone sees the same signal at the same time—no forwarded screenshots, no “did you see the alert?” pings.
Ask the agent what's happening in your business
Operations agent now answers questions in natural language. You can ask questions like:
- “What rules are you monitoring right now?” Get a plain-language summary of every active rule, what triggers it, and which data it watches.
- “When did this rule last fire?” A recap of what was monitored and actions taken, with links into the underlying detail.
- “What values have increased/decreased or met other conditions?” The agent can query and pull the current state from any data it is monitoring and explains it, so you don’t have to open a separate report or query to check.
It’s the fastest path from an alert to a clear, sourced answer.
Tracing and auditability
Every decision the agent makes is now visible. New tracing and audit screens show you:
- Which rule fired, on which data, at what time.
- What the agent reasoned about the signal.
- What action it recommended or took, and the outcome.
- Who authorized the action path that was taken.
This is the audit trail you need to put an autonomous agent into a governed environment and the debugging surface you need when something doesn’t behave the way you expected.
Activity log screen shows the rules monitored, alerts sent, and actions recommended and approved.
Governed alongside the rest of your AI estate
Operations agent is now integrated with Microsoft Agent 365 (A365) and Microsoft Entra ID—which means:
- Your operations agents show up in the same governance plane as the rest of your organization’s agents.
- Identity, access, and lifecycle are managed centrally—no separate control surface for “the Fabric agents.”
- Tenant admins get a single answer to, “What agents are running, who built them, and what can they do?”
If your organization has an AI governance posture, operations agent fits inside it on day one.
Getting started
Operations agent is available now! You'll see it still marked as Preview until later in June when our 'General Availability' version rolls out to all regions.
- Create your first agent: open any workspace and select New item → Operations agent.
- Refer to the Operations agent overview documentation.Read about other updates to Fabric IQ and Real-time Intelligence: Fabric IQ blog.
What’s next
Reaching generally available is a milestone, not a finish line. The team is already working on:
- Deep root-cause investigation and remediation.
- Multi-agent coordination, so operations agents can hand work off to other agents.
- Deeper Copilot integration so you can communicate with your operations agents like any others.
We’d love to hear what you build. If you have feedback or a scenario you’d like to see supported, connect with us in the Fabric Community forum.