Microsoft Ignite (2024): day 0 & 1 - carbs and code

Welcome to the Ignite blog series! I’m hoping to stay on-top of the blog this week to give you my thoughts on the sessions that I attend - most of these sessions are going to be about AI and data, so if that’s you’re thing… you’re in luck! Day 0 (Monday) can be shortly summarized but… I don’t want to understate it’s importance. Me and some guys got into town mid-afternoon, so after we got our badges and some swag we moved onto the most important part of the day (dinner). We ended up at the best pizza spot in the world, Lou Malnati’s (I will die on this hill and it’s not up for debate, sorry)

Ok fine… I‘ll move onto things of substance. Day one brought a wave of new insights, tools, and discussions that highlighted how AI continues to transform not just technology but the way we interact with it on a daily basis. From building better conversational agents to refining how developers approach coding, the day showcased Microsoft’s vision of making AI more integrated, autonomous, and accessible. Here’s a recap of my sessions from the day:

Copilot: The UI for AI

The keynote kicked off by re-framing Microsoft’s Copilot as not just a tool but "the UI for AI" itself. Copilot has moved beyond being a simple assistant and now acts as a crucial interface between users and AI capabilities. Microsoft’s vision is clear: they want AI embedded everywhere—in apps, documents, and even emails. The introduction of Copilot Actions, which works like rules for Outlook but for AI, feels like the logical next step for LLMs. Imagine setting custom AI workflows for your business tasks, much like you would set filters for email. It’s automation with an AI boost. Simple concept - AI is starting to do “things” now.

Another development was the new Azure Foundry, described as a one-stop shop for AI development. It feels like a natural evolution of the portal experience, combining AI tools, a model catalog, and governance features—essentially a one-stop platform for AI practitioners. Seems like it has the potential to be pretty cool - hopefully I will have some time this week to play with it.

Demystifying AI Development

One of the sessions that stood out to me (because I love sessions that get into the code) focused on the process of developing AI from start to production. AI workflows begin with the simplest form of providing prompts and evolve to become sophisticated orchestrations across multiple models and agents. The emphasis here was on iterative development—testing, fine-tuning, and getting feedback to optimize results. The session made a great case for why developers need to be flexible and approach AI as a probabilistic problem, rather than as something deterministic like traditional coding. This hybrid approach may be the future for mission-critical applications.

The "AI Development Lifecycle" was broken down into three key stages: Proof of Concept, Development, and Production. It's all about starting small, iterating often, and creating resilient feedback loops. It’s an approach that feels grounded and achievable for developers at all levels, regardless of whether you’re working with foundational models or fine-tuning existing solutions.

Azure AI Foundry & SDK

Azure AI Foundry aims to provide a unified platform for AI projects, integrating development tools directly with GitHub, Visual Studio, and Copilot Studio. The public preview SDK seems pretty helpful for developers—it unifies access to models and makes prototyping incredibly fast. What’s cool is how Azure AI Foundry simplifies building, deploying, and managing AI, reducing much of the friction developers face. It’s clear that Microsoft is trying to make AI as accessible as possible, which in turn makes it easier for everyone to contribute to innovative AI projects.

Elastic Search + AI: Optimizing Vector Databases

Elastic Search had a nice little session showcasing its capabilities with AI. Their vector database solution is nice—combining traditional search methods with vector embeddings to create a hybrid powerhouse. One of the standout features is their Better Binary Quantization (BBQ), which reduces memory by 95% without losing accuracy. Faster queries, smaller models—it’s a nice perk for developers trying to balance efficiency and performance.

Their integration with Azure OpenAI for conversational AI is another strong move. Imagine combining Elastic's data handling with the conversational depth of Azure OpenAI—the potential use cases for industries like customer support and analytics are huge, especially if you need help starting out with vector databases.

Model Distillation in Azure OpenAI

The concept of model distillation was a highlight in another session. Using large teacher models like GPT-4 to train smaller student models is something every AI enthusiast should be excited about. The result? Massive cost savings and lower latency—perfect for applications in resource-constrained environments like edge computing. Microsoft’s tools for automating and iterating through the distillation process make this approach much more approachable for organizations that need efficiency without sacrificing too much power.

What’s New with Copilot Studio

A lot of energy surrounded Copilot Studio. Microsoft is betting big on its new features. Enhanced integrations with Microsoft 365, better security and governance controls, and tools like Generative IVRs (AI-driven voice response systems) are turning Copilot into an even more versatile enterprise tool. Users can automate workflows, trigger events, and have autonomous agents take care of routine tasks—while keeping security and governance in mind.

Securing AI in the Enterprise

Security took center stage during a session on how to govern AI effectively, especially as the proliferation of citizen developers creates new challenges. The threats posed by generative AI—like prompt injections and data leakage—require robust solutions, and Microsoft’s advancements in this area are promising. Purview Data Security and AI-Security Posture Management tools are now available to ensure safe, scalable, and compliant AI deployments, whether you're a developer or a non-technical business user.

How EY Leverages AI for a Competitive Advantage

Another interesting session detailed how EY is using AI to enhance their operations. They have built a private LLM ecosystem called EYQ for their 300,000 employees and have deployed over 100 AI applications globally. The emphasis was on how AI has helped improve productivity, streamline workflows, and foster innovation across the organization. EY’s significant investment in AI highlights the growing importance of these technologies in driving business transformation. Their focus on democratizing AI aligns well with industry trends, emphasizing the need to make AI tools accessible and impactful across all levels of an organization.

New Announcements from Day One

Day one came with some announcements that showcased the company's focus on expanding AI capabilities and ensuring secure integration for businesses. Here are some of the key new announcements:

- Copilot Actions: Introduced as a way to automate workflows similar to Outlook rules, but for AI. This feature allows users to set custom AI-driven processes, making Copilot even more powerful in handling business tasks.

- Azure Foundry: A new platform designed to be a one-stop shop for AI development. It combines tools, governance features, and a model catalog, simplifying the process for AI practitioners.

- Purview Data Security and AI-Security Posture Management: Microsoft announced new tools to secure AI deployments, emphasizing the importance of risk management for both low-code and pro-code AI solutions.

- Generative IVRs: AI-driven voice response systems integrated into Copilot Studio, allowing businesses to enhance customer service capabilities with natural language understanding.

- Azure AI Foundry SDK: The SDK was released in public preview, aimed at making AI model integration simpler and more efficient for developers.

- Better Binary Quantization (BBQ): A new feature from Elastic, which reduces memory usage by 95% without losing accuracy, significantly improving efficiency for AI-based search and vector databases.

- Copilot Studio Enhancements: Microsoft expanded Copilot Studio's capabilities with better integrations, improved security, and advanced automation features, including multi-agent workflows.

Day One, Final Thoughts

Yes… I ended day with with Portillo’s… two Chicago dawgs with a cake shake, you just have to do it when you’re here! As I am sitting here writing this, I am reminded how much I love things like Ignite. It’s a great way for me to benchmark what we’re doing and to check our blind spots on things we may be overlooking. It’s also awesome being around so many people who just love tech - the energy at these conferences is always really good. The main challenges I have are just taking everything I learned and seeing what would work for us within our current systems and architecture - which is kind of the art of all of this given how we’re all working with different landscapes. I’m happy to come away from day one with some cool ideas and new perspectives.

Appreciate you reading and you can come back tomorrow for day 2!

✌️Steven