Blog

Microsoft Build 2025 Highlights

Posted on by Team BA Insight
Microsoft Build 2025

Microsoft Build 2025 brought together developers, IT leaders, and business innovators to showcase the future of AI in enterprise productivity.

This year, the focus was on the evolution of AI agents, groundbreaking integrations, and advancements that redefine how organizations work. Key members of the BA Insight team attended, and here we’ve summarized the most significant announcements and sessions that will shape the future of enterprise AI.

AI Advancements Shaping Enterprise Transformation

Microsoft Build 2025 underscored one key theme: AI agents are no longer futuristic; they are integral to modern enterprises, driving productivity, automation, and innovation. Below, we explore highlights from seven key sessions.

 

Session 1 | Copilot Studio Agents & Azure AI Foundry: Better Together

This session spotlighted how Copilot Studio integrates seamlessly with Azure AI Foundry, offering flexible, scalable solutions across enterprise use cases. Key takeaways include:
 
  • Over 1,900 models available to suit diverse business needs.
  • Deep integration with Azure AI Search, Microsoft 365 (M365), and Fabric, making it easier to build multi-functional AI agents.
  • Bring-your-own-model capabilities, allowing organizations to customize AI workflows.
Highlighted Demo: Azure AI Search showcased how SharePoint documents can be vectorized and connected within Copilot Studio, allowing seamless retrieval and custom automation through multi-model retrieval-augmented generation (RAG).
 
This session made it clear how organizations can unlock valuable insights by combining existing knowledge sources with scalable AI tooling.
 

Session 2 | Building autonomous agents in Copilot Studio

Copilot Studio continues to evolve and is making it easier than ever to create complex reasoning to solve real business problems, both independently and with a human touch. Organizations can now build solutions for complex problems with increased ease, thanks to Copilot Studio’s updates.
 
These 10 components aim to minimize complexity while maximizing impact:
 
  1. Multi-Agent Support: Processes can’t reliably run in one agent, but Copilot Studio now supports multi-agents for streamlined collaboration across tasks
  2. Deep Reasoning: Complex reasoning has been a problem historically, but Copilot Studio now supports “Deep Reasoning”
  3. Model Context Protocol: LLMs need help with raw APIs and connectors, and Copilot Studio now supports Model Context Protocol
  4. Agent Flows: Not every process should be reasoned, which is why Copilot Studio now supports Agent Flows
  5. Activity and Debug: Tools and RAG require debugging to work properly, and Copilot Studio now supports Activity and Debug tools to resolve issues quickly
  6. Data Governance: With the support of multi-agents, Copilot Studio now better supports governance and security
  7. Robust Analytics: ROI is critical for any business, and Copilot Studio now offers more comprehensive ROI-driven analytics
  8. Tech Flexibility: Extensibility to LLMs is critical, and Copilot Studio now supports BYOM and SDK
  9. Enhanced Admin UI: Administration in Copilot Studio has been complex in the past, but the new updates to Admin UI make it easier to use than ever
  10. Computer Use Agents: Copilot Studio can now perform tasks and interact with any system without APIs, via Copilot Studio Computer Use Agents
Copilot Studio is transforming into a tool not just for developers, but also for non-technical teams, making it easier than ever to create ROI-driven workflows.
 

Session 3 | More Knowledge to M365 Copilot with Connectors

Microsoft showcased how organizations can connect third-party data sources to M365 Copilot using Copilot Connectors (formerly Graph Connectors). A few highlights:
 
  • 65 pre-built connectors for platforms, including ServiceNow, Gong, GitHub, and more. One live demo showed how users can set up a connector with ServiceNow’s knowledge base in under five minutes.
  • The power of connectivity: Another highlight was a walk-through from Gong’s EVP of Product, who showcased how their connector allows users to surface Gong insights directly with Copilot while enabling full respect for data permissions and user access.
  • Microsoft Connector & Agents Builder Toolkit: This new resource makes it much easier for developers to build custom connectors (whether it’s a fortified BA Insight connector or a standard one) and take action on data, not just surface it.
Why This Matters: This opens immense possibilities for BA Insight to extend capabilities to legacy systems and tailor solutions for unique enterprise environments, ensuring that data is actionable and permissioned.
 

Session 4 | Copilot Studio: Use Agents & Automation to transform business processes

Copilot Studio 2025 empowers organizations to leverage AI agents for enhanced productivity, streamlined operations, and efficient business process automation. During this session, we learned about:
 
  • Multi-Agent Collaboration: Copilot Studio now supports multi-agent systems, allowing AI agents to delegate tasks among themselves. For instance, an agent can extract sales data from a CRM, pass it to another agent to draft a proposal in Word, and then schedule follow-ups via Outlook.
  • Low-Code Agent Development: Users can create custom AI agents using natural language prompts and low-code tools, enabling rapid deployment without extensive programming knowledge.
  • Enhanced Automation with Agent Flows: The introduction of “agent flows” allows for structured automation of complex business processes, combining deterministic workflows with AI-driven decision-making.
  • Governance and Security: Copilot Studio incorporates robust governance features, including activity auditing via Microsoft Purview and integration with Sentinel, ensuring secure and compliant AI agent operations.

Session 5 | What’s new in Copilot Studio

This session highlighted Copilot’s evolution into a powerful platform for building, customizing, and deploying AI agents that transform both individual productivity and enterprise business processes.
 
Here are a few key highlights from this Copilot Studio session:
 
  • Autonomous agent creation in Copilot Studio, enabling automation of business processes like customer service, sales, and HR.
  • Integration of external content into Copilot agents, allowing richer, context-aware solutions beyond Microsoft 365.
  • Enhanced agent-building features: knowledge picker, natural language tools, and full-screen authoring for an improved user experience.
  • Multi-agent orchestration, enabling specialized and even third-party agents to collaborate on complex, end-to-end workflows.
  • New publishing options for agents, including SharePoint and WhatsApp, and improved analytics and testing tools for continuous enhancement.
  • UI automation without APIs, allowing agents to perform tasks like generating property tax statements from public websites using plain English instructions.
  • Real-world demos, such as T-Mobile’s use of Copilot Studio in Teams to summarize and send emails, showcasing a real success story for practical business automation.

Session 6 | Azure AI Foundry: The Agent Factory

We took a deep dive into the Azure AI Foundry, Microsoft’s comprehensive platform for building, customizing, and deploying AI agents at scale. It supports a vast array of open-source and proprietary models, integration with the Microsoft ecosystem, and enterprise-grade tools for secure, scalable deployment, which positions it as a central player in the emerging “agentic web.”
 
  • Unified AI development platform supporting over 10,000 models, including those from Grok, Hugging Face, Meta, and Mistral.
  • Robust tools for model selection, fine-tuning, agent creation (via low-code/no-code or SDKs), evaluation, and deployment across cloud, edge, or local environments.
  • Deep integrations with Microsoft 365, GitHub, and Windows AI Foundry, plus support for open protocols like Model Connectivity Protocol (MCP).
  • Multi-agent orchestration for collaborative, complex workflows across business functions.
  • Key use cases include sales automation (e.g., Fujitsu), healthcare workflow optimization (e.g., Stanford Health Care), customer support, document processing, and software development.
  • Enterprise-ready features: security, compliance, scalability, and built-in evaluation tools for the entire AI lifecycle.
  • A look at the competition re: Google Vertex AI, Amazon Bedrock, IBM watsonx, Hugging Face Infinity, and APIs from Anthropic Claude and OpenAI gave a lens into what makes Azure AI Foundry different, notably: scale and model diversity, deep integrations with Microsoft’s ecosystem, multi-agent orchestration, scalability, and open protocol support.

Why This Matters: Agents are easier than ever to build. It’s never been easier to extend agentic capabilities either!

Session 7 | The power within: Running Large Language Models on your local machine

Curious about running powerful AI models right on your own device? Recent advancements have made it possible to deploy large language models (LLMs) locally, unlocking new levels of privacy, speed, and control. Here are the key takeaways from the latest session on harnessing LLMs:
 
  • Local LLMs are now practical: Thanks to advancements in hardware acceleration (like NPUs) and optimized run times, it’s now possible to run powerful LLMs directly on personal devices.
  • Privacy and control: Running models locally ensures sensitive data never leaves the device, offering a significant privacy advantage over cloud-based solutions.
  • Performance gains: Local inference reduces latency, enabling real-time interactions without relying on internet connectivity.
  • Developer empowerment is huge: The session demonstrated how developers can use tools like ONNX Runtime, Olive, and Windows AI Studio to convert, optimize, and deploy models locally.
  • Integration with Windows: Microsoft is making it easier to integrate LLMs into Windows apps using native APIs and support for models like Phi-3 and LLaMA.

Why These Advancements Matter for Enterprises

Microsoft Build 2025 solidifies the importance of AI in optimizing enterprise operations. From automating processes to enhancing decision-making, these tools offer:
 
  1. Enhanced Productivity: Automate repetitive tasks so teams can focus on high-value work.
  2. Data-Driven Insights: AI tools like Copilot unlock powerful insights from large, complex data sets.
  3. Future-Proofed Solutions: Seamless integrations with external platforms ensure businesses stay adaptable in a dynamic landscape.
Yet, its true impact lies in creating a competitive advantage. Organizations equipped with AI today will thrive in tomorrow’s AI-powered world.
 

Wrap up

Microsoft Build 2025 made it clear: AI agents are no longer a futuristic concept. They’re here, transforming how we work, collaborate, and innovate. With Microsoft’s latest news revealed at the event, organizations of all sizes are better positioned to harness the power of AI to drive real, measurable results – accelerating productivity, automating complex workflows, and unlocking new business value. We saw real AI solutions in action, and so many advancements that set the stage for exciting developments just over the horizon.
 
This event showed us that organizations worldwide are reimagining what’s possible with Microsoft’s AI ecosystem, and we’re so jazzed to be HERE for it. If you’re excited too, reach out to our expert team to talk about possibilities and solutions for your business.

 

-->