Microsoft has detailed its Microsoft 365 Copilot Wave 2 Spring release, outlining significant updates designed to integrate AI more deeply into daily workflows. The release introduces sophisticated AI agents like Researcher and Analyst, new user features including AI-powered search and content creation tools, and a central Agent Store.
Alongside these capabilities, Microsoft is substantially bolstering its Copilot Control System (CCS) to provide IT departments with enhanced tools for managing the security, cost, and deployment of this expanding AI ecosystem.
New Copilot Capabilities Emerge
The core Microsoft 365 Copilot application receives a refreshed design and acts as the hub for several new features. A new Copilot Search feature leverages semantic indexing—AI that understands the meaning and context of data—to provide richer answers from across company information stored in Microsoft 365 and connected third-party applications like ServiceNow or Google Drive. For web searches, it integrates results from Bing without directly crawling sites.
The new Create experience incorporates OpenAI’s GPT-4o model for generating and customizing images that align with corporate brand guidelines. Copilot Notebooks offer an expanded workspace (up to 18,000 characters versus the standard chat limit) for users to iterate on complex prompts while keeping previous versions visible, grounding the AI in specific documents or notes.
Copilot also adds a Memory capability, allowing personalization based on user-provided details and instructions, with controls for managing remembered information.
New Agents Enter the Workplace
Central to the Wave 2 release are specialized AI agents, building on capabilities first announced in March. Rolling out via Microsoft’s Frontier early access program starting April 23rd are the Researcher and Analyst agents.
Powered by advanced OpenAI reasoning models (specifically including the o3-mini model for Analyst), Researcher handles complex, multi-step information synthesis, while Analyst brings data science workflows into tools like Excel. A third agent, Skills, is planned for a June debut, using the People Skills data layer to help leaders with talent discovery and team formation.
These agents, alongside others from partners like Jira, Monday.com, and Miro, and custom agents built with Copilot Studio (which itself gained GUI interaction capabilities in preview in April), are discoverable in a new central Agent Store within Copilot.
Microsoft Corporate Vice President Charles Lamanna previously stated about the GUI agents, “If a person can use the app, the agent can too.”
Strengthening IT Governance for an Agent-Filled Future
Deploying potentially autonomous AI tools necessitates strong oversight, a challenge Microsoft addresses with significant updates to the Copilot Control System (CCS).
Key is the upcoming “Apps and agents in Data Security Posture Management for AI” within Microsoft Purview (public preview planned for June 2025). This aims to provide a unified dashboard for managing AI risks across M365 Copilot, other Microsoft copilots, and third-party LLM apps, allowing monitoring of sensitive data access and policy application. For usage tracking, Copilot Analytics gains a “Message consumption report” (preview expected May 2025) and an “Agent usage report” (preview expected June 2025).
The Copilot Studio Agent Report in Viva Insights (in preview) will add customizable business impact reporting in June 2025 to correlate agent activity with business metrics. Agent lifecycle management within the Microsoft 365 admin center sees updates to inventory views (May 2025), options to block/deploy agents and export lists to CSV or manage via PowerShell (Shared Agent Inventory Management, May 2025), and more granular user permissions controls planned for Q3 CY25, building on existing capabilities in Integrated apps.
Contextualizing Human-Agent Collaboration
Microsoft positions these advancements as a response to workplace demands identified in its 2025 Work Trend Index, which highlights a “Capacity Gap”—with 68% of employees struggling with workload pace and 46% reporting burnout—and suggests AI agents can help. The report introduces concepts like “Frontier Firms” reorganizing around AI and employees becoming “agent bosses.”
This coincides with industry warnings, such as Anthropic CISO Jason Clinton’s recent prediction of “virtual employees” arriving within a year, posing challenges in areas like Non-Human Identity Management (NHIM)—where automated accounts already vastly outnumber human ones—and accountability. Clinton highlighted the difficulty, saying, “In an old world, that’s a punishable offense… But in this new world, who’s responsible for an agent that was running for a couple of weeks and got to that point?”
While Microsoft’s CCS provides administrative tools, the practicalities of securely managing highly capable agents remain a developing area across the industry. Foundational standards like Anthropic’s Model Context Protocol (MCP), now adopted by Microsoft, OpenAI, AWS, and Google, are facilitating the necessary interoperability. Most features from Microsoft’s Wave 2 release are expected to start rolling out in late May 2025.
