Microsoft is making a number of announcements around its Copilot Studio and Copilot Agents during its Build event in Seattle this week. Not to be topped by ServiceNow’s recent announcement of its AI Agent orchestration tools, the highlights of this week’s releases focus on enabling developers and line of business to better unite around building, maintaining and optimizing AI for both productivity and even cost effectiveness.
- Larry Dignan: Microsoft Wants to be Your Agentic AI Developer Stack
- Agentic AI protocols: MCP and A2A today, many more tomorrow
- ServiceNow Launces Workflow data Network, Aquires Data.word
Additional Copilot Studio capabilities announced which include enhancements to AI Agent support include:
- Multi-agent orchestration - Rather than relying on a single agent to do everything—or managing disconnected agents in silos—organizations can now build multi-agent systems in Copilot Studio (preview), where agents delegate tasks to one another. This includes those built with the Microsoft 365 agent builder, Microsoft Azure AI Agents Service, and Microsoft Fabric.
- Computer Use in Copilot Studio agents - Agents can now interact with desktop apps and websites like a person would clicking buttons, navigating menus, typing in fields, and adapting automatically as the interface changes. This opens the door to automating complex, UI-based tasks. Bring your own model and model fine-tuning - Makers can access more than 1,900 models in Azure AI Foundry, including the latest models available in OpenAI GPT-4.5, Llama, DeepSeek, and custom models, and fine-tune them using enterprise data to generate domain-specific, high-value responses.
- Model Context Protocol makes it easier to connect Copilot Studio to your enterprise knowledge systems.Microsoft Entra Agent ID, for agents created through Copilot Studio or Azure AI Foundry, automatically assigns agents an identity, giving security admins visibility and control in the same tool they use to manage organizational identity and access. The Agent feed hub will allow end users to oversee teams of agents in Power Apps, showing task status and flagging where an agent is stuck and needs help. And Microsoft Purview Information Protection will be extended to Copilot Studio agents that use Microsoft Dataverse, allowing organizations to automatically classify and protect sensitive data at scale.
In addition, the company is also providing some interesting new tools that can help developers and project leaders on the line of business side better account for the costs of consumption-based tools like AI agents. These updates include:
- Enhancements to Billing and Usage - To support flexible deployment, CCS introduces pay-as-you-go (PAYG) group-level billing for agents in M365 Copilot Chat. This model ensures that organizations only pay for what they use, while maintaining full visibility and control over agent usage expenses by departments and user groups.
- Message consumption report - The new Message consumption report supports agent management decisions by enabling AI admins to monitor billed messages, identify high-usage scenarios, and gain visibility into message consumption by agent, user, billing policy, and user-agent pair.
Microsoft is also looking to improve the effectiveness of agents by expanding the types of data and content they can consume out of the box. This is supported by two new capabilities:
- New agent publishing channels - Copilot Studio can now publish agents to Microsoft 365 Copilot, and coming soon, will be able to publish agents to SharePoint and WhatsApp. We’re adding new categories to ground and tune your agents, including “Responses,” “Moderation,” “User Feedback,” “User Input” and “Knowledge.”
- Knowledge controls - Copilot Studio now supports more input sources including OneDrive files, SharePoint lists, Teams and external sources such as ServiceNow, Zendesk and SAP.
The interoperability between ServiceNow and SAP is notable. As most of Microsoft’s applications customers can be considered more mid market in size and scope, but Microsoft’s AI ambitions are clearly in the enterprise, a strong multi-agent approach that incorporates common enterprise applications is table stakes.
All of these announcements point to one increasingly obvious fact. The future of business will be “multi-agent” - both in terms of multiple agents inside single applications working together, as well as needing to orchestrate agentic flows between multiple systems to automate even the most common tasks. This is less a race for dominance and more a race for sensible interoperability given the obvious implications among legit enterprise AI providers like Microsoft, Oracle, ServiceNow, Salesforce, etc.
For growth leaders, these advancements continue to offer up opportunities to reevaluate the go-to-market tech stack, and find areas to “re-balance” between human and digital labor. This is both from an overall cost and budget perspective, but also from a perspective of reducing complexity and friction in GTM motions. Take time to evaluate these new announcements (even as they come with increasing rapidity), and strategize how they fit into your overarching AI plans. For Microsoft customers, these tools are easy to consume and test. But regardless of either how reliant you are on Microsoft technology, or if you’ve chosen (or looking to chose) another vendor as your “anchor” agentic AI provider - the truth is that in this multi-agent future, we are going to have to understand, consume and integrate with multiple agentic AI platforms on a constant basis.
