Microsoft released a GPT-5 add-on for Copilot Chat (via ANM’s Coffee with Chris), presenting it as their most advanced AI model so far.

Key benefits: stronger natural language understanding, more capable in handling complex prompts, presumably better context awareness. Good for organizations that are pushing use of Copilot for high-impact productivity tasks.

Copilot Memory

ANM’s “Coffee with Chris” also talked about Copilot Memory, announced at Microsoft Build 2025.

What this looks like: Copilot will be able to remember past context, user preferences, settings, etc., so that interactions feel more continuous rather than one-off. Think: less need to re-explain things, more personalization. This raises user convenience but also brings up data-governance/security questions. 

Copilot Notebooks

Another new feature discussed: Copilot Notebooks.

The idea is to have a space (notebook-style) to gather notes, documents, probably drafts etc., in one place facilitated by Copilot. Good for users who regularly need to collect and synthesize information. Could reduce friction between jumping between apps.

Build 2025: Bigger Platform & Customization Moves

From ANM’s blog summarizing Microsoft Build 2025:

• Copilot Tuning: Low-code tools in Copilot Studio for tuning AI models using your own data, workflows, terminology. Useful for orgs that have domain-specific needs (legal, engineering, consulting). Helps bridge gap between generic AI and business-specific AI without needing deep data science.
• Multi-Agent Orchestration: Multiple agents working together on tasks. For example, one agent pulling data, another drafting a document, another scheduling. This reflects a trend: AI components not working in isolation, but collaborating. Might complicate architecture / oversight but has big upside.
• Copilot Studio Enhancements: Improved tooling for building, embedding, evaluating agents. More controls / richer analytics and dashboards so you can see how agents are performing.
• Teams as Hub for AI-Collab:
  • Agent-to-Agent Protocol (A2A): agents talking to each other within Teams.
  • Model Context Protocol (MCP): agents can access external services / tools.
  • Export APIs to audit inspect usage / compliance.

What To Watch Out For & What This Means

These are big shifts, but they come with trade-offs. Here are some things to think about:

• Data privacy & governance becomes more critical. Things like memory, tuning with internal data, and or agents that handle user workflows mean exposure risks. Organizations will need to ensure proper controls, auditing, compliance.
• Complexity of management rises. More customization plus multi-agent workflows means more to configure, monitor, maintain. More chance of misalignment or unexpected behavior.
• Adoption curve: Some of these are in preview (private preview) or early stage. Real benefits may lag until tooling stabilizes and best practices emerge.
• User training / change management: More powerful AI tools will demand more from users in terms of understanding what AI can/should do vs what it can’t. Also, how to use features like memory or agent orchestration well (without giving up control or creating risk).

Microsoft’s 2025 announcements are already reshaping how organizations think about AI, productivity, and collaboration, and there’s plenty more to come. To stay ahead of the curve, tune in to Coffee with Chris every other Friday on ANM’s YouTube channel. You’ll get practical insights, real-world context, and expert perspective so you never miss a major Microsoft update.