Abram Jackson

I just published a new video for developers showing how to build a Foundry Hosted Agent running inside Microsoft 365 Copilot & Teams using a fun Blender 3D scenario. In this end-to-end walkthrough, I cover the full developer flow: ✅ Using the VS Code Foundry extension and the Hotel Python template ✅ Exploring the local Playground experience ✅ Debugging locally ✅ Switching to the Blender agent running locally to test the scenario ✅ Deploying the agent ✅ Testing it in the remote Playground in VS Code ✅ Validating it in the Foundry portal ✅ Reviewing telemetry and logs with Application Insights The video is about 20 minutes long and is meant to be practical, concrete, and hands-on for developers who want to understand the full build → test → deploy → monitor workflow. 💻 Sample code: https://lnkd.in/ehmmKe4D 🎥 Video: https://lnkd.in/eV3u6pXV I hope this helps developers get a clearer picture of how Hosted Agents work in Microsoft 365 Copilot & Teams and how to build their own. #Microsoft365Copilot #Copilot #AI #Agents #HostedAgents #AzureAI #Foundry #VSCode #Python #Blender #Developers #AppInsights

37

These last 2-3 weeks a lot of folks have been working hard preparing for Microsoft Build. In this journey, due to time constraints, I turned to AI to see how we can get things done faster. There have been many spots where it's saved me time and I'm starting to look at Copilot more as my teammate than a technology. - Finding something: This has been a big help to me. looking back how I used to find things. I'd recall where abouts something was and then my pride would kick in and I'd spend 3-5 minutes going through emails, teams, docs until I found it. Copilot is better at that than me. Sometimes it's a miss, but I'm starting there first now. - Summaries and Actions: Really liking it in meeting contexts. Teams AI meeting generation is pretty good, BUT I caught something and started doing it. I saw people in meetings adding things to chat with actions and items. This helps! just a small tweak, but your future self will thank you. - Demos: WOW. I've gone down a rabbit hole of the concept of a "master prompt" to provide context. I do wish Copilot had some sort of memory capability for me to supply that master prompt. Either I'm not aware, or hopefully it will come. Couple of things I've found on this front. 1. my first prompt is to ask it to generate a prompt for me with a given context. This was HUGE! The detail in just the prompt, and then the outcome that comes from that generated prompt is way more fruitful than just going with a basic prompt yourself! 2. With that generated prompt, AI is actually not bad at creating data to use for demos. (still need to validate you won't get into a problem with the data, but shaves the time down a lot). I'll keep trying things. I'm always working on how to improve productivity and these last two weeks have been a game changer for me.

86

7 Comments

Most AI agent samples teach you to call a model. They rarely show what happens when five agents need to coordinate, tools need to ship independently, and the whole stack needs to deploy to the cloud. Justin Yoo's Interview Coach sample fills that gap. It puts Microsoft Agent Framework, Foundry, MCP, and Aspire into a working application — an interview simulator where specialized agents hand off control to each other through a defined workflow. The patterns are worth studying: handoff orchestration where each agent fully owns its turn, MCP tool servers that deploy independently of agent code (and can be in a different language), and Aspire tying the service graph together with discovery, health checks, and tracing. The part that stuck with me — the principle of least privilege applied to agent tooling. Each agent only gets the tools it needs. The triage agent gets none. If you're moving multi-agent work toward production, what's the gap you're trying to close? https://lnkd.in/eamzATfP

15

There’s been a lot of buzz post-Build around Microsoft’s new Bring Your Own Model (BYOM) support in Copilot Studio and Azure AI Foundry…and for good reason! BYOM is huge! Let’s break it down with a real-world scenario. Suppose you want to spin up an agentic team to help you design, code, market, and ship a product. Ideally, each agent specializes in its domain and works together autonomously to drive the project forward. Without BYOM, you’d need to pick one model provider to power every agent — developer, designer, comms, project manager, etc. But the reality is no single model is best at everything. We want peak performance here. With BYOM + Semantic Kernel + Agent Runtime + Copilot Studio, you can build a fit-for-purpose team, here’s how I’d do it: ✅ Developer Agent — powered by Anthropic’s Claude 3.7 Sonnet (best-in-class code and reasoning, announced in Feb 2025) ✅ UX, Marketing, Comms, and Copy Agents — powered by OpenAI’s GPT-4o (an omni-modal creative powerhouse) ✅ Project Planning Agent — powered by Google’s Gemini 2.5 Pro (massive 1M token context window for long-horizon reasoning, announced in March 2025) These agents collaborate natively inside Microsoft’s orchestration layer, thanks to Copilot Runtime (MCP), Azure AI Agent Runtime, and Semantic Kernel. SK would be the key in this scenario to “wrap” Claude and Gemini as skills for agents to call. Tool use, memory sharing, and multi-agent planning? Fully supported. Boom 💥 This is why BYOM is a big deal: it unlocks best-in-class agents, working together, inside a Microsoft-native Copilot experience. What agentic teams can you envision, and what personas and models would you use? #MicrosoftBuild #Copilot #BYOM #SemanticKernel #AzureAI #AIAgents #AgenticAI #GPT4o #MCP #CopilotStudio #AgentRuntime

11

If you’re using the SQL Analytics Endpoint over Lakehouse data, file health and maintenance really matters. Christopher  Finlan's latest blog does a great job walking through how to ensure your data is in the right shape for fast SQL querying. Check it out https://lnkd.in/gqt4CE4x #MicrosoftFabric #Data #Analytics #SQL #DataEngineering

39

1 Comment

AI isn’t replacing your operations, it’s making them better. Forrester Consulting shows how a composite org achieved a three-year 207% ROI and $38M net present value with Google AI in Google Cloud Customer Engagement Suite.

1

The majority of companies interviewing developers are asking questions relevant 5-10 years ago - the tectonic plates have shifted, learn an AI-Native development workflow or become obsolete. Algorithms are secondary, structure and process are now king. 👑

3

10 Comments