Summary
At a Microsoft Ignite pre-briefing event, Microsoft walked through how they are extending M365 Copilot, Power Apps, and Copilot Studio into a more complete platform for building and governing agents. The story connected three main threads: natural-language app and workflow creation inside M365, an AI-first rethink of low-code development, and stronger evaluation and security controls for agents at scale.
Underneath it all, Microsoft is preparing for Agent 365, a consolidated control layer for agents across the Microsoft ecosystem and beyond.
Momentum in Low-Code and Agents
The pre-briefing opened with adoption numbers that show how central this stack has become. Copilot Studio is used by 90% of the Fortune 500. Power Platform has 56 million monthly active users. Agent usage is doubling quarter over quarter.
Over the last several months, Microsoft has also added support for Anthropic models (including Sonnet 4.5) into Copilot Studio, delivered SDKs to wrap third-party agents and bring them into M365, and adjusted licensing to make agent usage more predictable. These changes set the stage for Ignite’s new capabilities: making it easier to create agents, connect them to existing business apps, and prove that they behave as intended.
The scale numbers confirm that agent building is no longer a niche experiment. For many enterprises, M365, Power Apps, and Copilot Studio are becoming the default place where AI agents are defined and managed.
M365 Copilot as a Surface for Apps, Workflows, and Agents
On the M365 side, Ignite brings three key areas of expansion: App Builder, Workflows Agent, and a more capable Agent Builder.
App Builder is a new M365 Copilot agent that lets users describe an app in natural language and have it built automatically. It’s designed for non-developers who need simple logic for a task or team workflow. After the initial build, users can refine the app through the same conversational interface. It’s already available to M365 Copilot customers.
Workflows Agent offers the same simplicity for automation. Users describe triggers and actions in plain language (based on Microsoft Graph data) and Copilot turns that intent into a working workflow across Outlook, Teams, and other M365 tools.
Agent Builder inside M365 Copilot is also getting upgrades. Users can now create agents that generate Word, Excel, and PowerPoint files and work across their personal Microsoft Graph, including Outlook and Teams. SharePoint response quality has improved, and agents built here can now be exported to Copilot Studio for more advanced customization and governance.
Microsoft also clarified naming: the former “Copilot Studio Lite” is now just Agent Builder within M365 Copilot. Copilot Studio remains the standalone environment for building and publishing more advanced agents.
This is a clear move to make M365 Copilot not just a consumption surface, but a creation surface. Business users can start with lightweight agents, apps, and workflows, then hand them off to development and platform teams through Copilot Studio when they need more control.
Carrying Low-Code into the Agent Era
Microsoft is redesigning Power Apps with an AI-first, conversational approach. Instead of starting from a drag-and-drop interface, makers will be able to describe the app they want and get support from a set of agents.
A data agent suggests tables and schema, a requirements agent asks clarifying questions, and a coding agent helps assemble the app. The experience becomes more like a guided conversation, while still letting makers refine and extend the output. Microsoft is also adding MCP servers on top of existing Power Apps. This lets apps expose their forms and workflows as tools through MCP, so agents can interact with them, such as filling out forms or updating records for a user.
The end-user experience is evolving as well. Copilot chat is coming directly into Power Apps, replacing the older Power Apps Copilot. Users will see the same sidecar chat they use in Word, Excel, and PowerPoint, with context shared from M365 Copilot. They’ll be able to ask questions about app data or take actions through natural language.
This is a strong bridge between past and future. Organizations that invested heavily in low-code do not have to start over. Those apps become first-class tools in an agentic world via MCP, while the new AI-first maker experience lowers the barrier for building the next wave of applications.
Copilot Studio: From Building Agents to Proving They Work
Copilot Studio is adding more enterprise-grade features focused on reliability, security, and scale, with the biggest update being built-in agent evaluation. Makers can create question sets that simulate real user interactions that are either generated automatically or written for specific scenarios like onboarding or support. They choose evaluation methods (semantic match, strict match, quality scoring), set thresholds, and run the tests across the agent.
Results appear in a dashboard with pass/fail status, scores, and grader feedback. Strong and weak responses are highlighted so makers can see where the agent meets expectations and where it needs improvement. Evaluation becomes part of the normal lifecycle instead of a manual, separate step.
Copilot Studio is also extending its computer-use capabilities by letting agents operate Windows 365 desktops. IT can manage these desktops with standard policies and security controls, allowing agents to perform UI-dependent tasks in a controlled environment.
For security, new webhooks let Copilot Studio agents connect to Microsoft Defender or third-party security tools to mitigate threats like prompt injection. This brings existing security stacks directly into agent workflows.
Agent building is maturing from “can we make it work?” to “can we trust it in production?” The built-in evaluation framework and security hooks are clear steps toward making agent development look more like traditional software engineering, with tests, monitoring, and integration into existing security practices.
Governance, Power Platform Admin, and the Role of Agent 365
Governance ran through the briefing as a recurring theme. Alongside the product changes, Microsoft is improving how admins manage agents and apps in the Power Platform Admin Center, with more detailed views of usage and configuration. These improvements are designed to line up with Agent 365, which will be the broader, cross-Microsoft control plane for agents.
Agent 365, briefly referenced in the briefing and covered in its own session, is described as a consolidated security and observability layer for agents running both inside and outside the Microsoft environment. It is expected to integrate with Entra, Defender, and other first-party governance tools, giving security and platform teams one place to see what agents exist, what they are doing, and how they are configured.
As more agents are built by business users in M365 and Power Apps, central governance becomes non-negotiable. Power Platform Admin Center improvements and the upcoming Agent 365 capabilities are aimed at giving enterprises a way to scale agent usage without losing control.
Looking Ahead to Ignite
This pre-briefing makes it clear how Microsoft views the next phase of enterprise AI: business users create simple agents, apps, and workflows inside M365; low-code teams extend and connect those assets; and platform teams govern them through Copilot Studio, Power Platform tooling, and Agent 365.
At Ignite, we’ll be watching for real-world examples of how customers are:
- Using App Builder and Workflows Agent to push more creation into the hands of business users.
- Turning existing Power Apps into agent-accessible tools via MCP.
- Applying Copilot Studio’s evaluation features to keep agents reliable over time.
- Centralizing agent governance and security as Agent 365 comes into view.
For enterprises, the direction is clear: AI agents, low-code apps, and governance are no longer separate tracks. Microsoft is working to tie them into a single operating model for agentic work across the Microsoft 365 and Power Platform estate.
AI Automation Targets the Growing Linux Security Skills Gap
Axonis and HebronSoft Target Secure AI for Regulated Industries
Digital.ai Embeds AI-Driven Mobile App Protection Into CI/CD
Becoming AI-Lingual May Determine Who Operationalizes AI
Wordly Expands AI Translation for Microsoft Teams Meetings
