Far too often we talk about efficiency, productivity, and there is nothing new in these enterprise technology conversations.
AI marks a turning point. It is not about doing things faster. It is about doing things we could not do before.
This is about augmenting human intelligence at scale, unlocking creativity and innovation.
The intelligence revolution is in full flight. AI agents are eliminating that soul-crushing work to amplify human creativity.
This technology will enable the user to talk to our platform and abstract away that gap between how quickly we're incorporating new capabilities and how people need to keep up with them.
KT is accelerating large-scale AI innovation optimized for the Korean market where individuals and companies can work more smartly and creatively.
70% of the world's workforce, they're in the field, every single day, keeping economies running. Today, we have intelligence like we've never had before to enable frontline workers to do significantly more.
We help allocation of capital and allocation of capital in an efficient way. So to be able to do that in a more effective way that creates more jobs, driving better growth for the world, better financial stability, that's beneficial to everyone.
Scientific advancements are converging with artificial intelligence to redefine what is possible, helping our scientists move faster and our patients get the care they need.
We can boost education systems. We can help people get job-ready skills to enter the workforce and we can help workers in their jobs augment their skill sets.
We're accelerating the deployment of AI to build a better future, a future for our people, our clients and society as a whole.
We can solve complex global challenges, unlock new opportunities and empower people to thrive in a rapidly changing world.
The opportunity isn't just for our businesses, it's for our people, our communities and the full force of human potential.
We are shaping a future defined by human ingenuity and technological advancement.
When you use AI against parts of the job that people don't want to do, that unlocks passion and when you have people who love their jobs, you get nothing but magic.
Please welcome Chief Executive Officer Microsoft Commercial Business Judson Althoff.
Well, good morning. Good morning, everyone and welcome to Microsoft Ignite. It's fantastic to be here with more than 20,000 of you live here in San Francisco and more than 200,000 of you watching online around the world. So good morning, good afternoon, good evening, no matter where you may be. We are all very excited here at Microsoft to talk to you about the more than 70 announcements we have for you this week. For today's keynote, we're going to focus on the major ones that empower the Frontier Firm. I want to thank you all from the bottom of my heart very sincerely for spending time with us today. Your time is valuable. Once you spend your time, you never get it back. So we're grateful that you've chosen to spend it with Microsoft today. Let's get started. I want to begin with this notion that frontier transformation is distinctively different from AI transformation. But like any good journey, it's important to reflect upon where you've been so that you can plot the course forward appropriately. So, like for any event, I used Copilot to prepare for this one. I asked Copilot, "Hey, what is the state of the union of AI?" Gives an honest response. Hey, look, there's been no wave of technology ever in the history of the world that has ever been adopted faster than AI. But on balance, the success rates of AI projects are not what we'd like them to be. Depending on the report that you read, there's a range of failures out there. And you have to look into the data to understand why. There's four very common reasons why AI projects fail. First, the alignment between business and IT professionals is very inconsistent. Second, there's a sea of data quality issues out there that need to be rationalized by AI developers in order to get meaningful output from the efforts. Third, governance and requirements in the markets, whether they be regulatory or otherwise, require AI to be kept on the sidelines from real major efforts that can impact business. And finally, there's an overemphasis on experimentation, random acts of innovation, rather than AI being put to use at scale across real business scenarios that can make an impact. But fear not, the ROI in AI is not somehow elusive. At Microsoft, we have the pleasure of working with thousands of customers around the world, across industries, and they are making a difference with AI in their business. Many of them use the success framework in front of you, what I call the "Frontier Success Framework". And there are some common patterns that emerge. These companies are deeply focused on enriching the employee experience, hiring the best talent, nurturing that talent, equipping them with the world's best tools, AI infused into the way they work, tying it back to measurable KPIs on the business so that they can see real impact in the results that they strive to achieve. Second, frontier firms are ruthlessly focused on making sure that they see the results in the customer experience journey. It's about real-time engagement with customers. It's about making sure that they have connections, better relationships with customers, so that at the end of the day, they achieve better results with a better cost structure. Third, you have to reshape business processes. You can't simply apply technology to an existing process and expect revolutionary results. You have to comprehensively look left to right to see how a business process can become AI-first and truly be changed to impact the business. Last but not least, the Frontier Firm focuses deeply on innovation and bending the curve so they can drive real competitive advantage in the industries that they serve. Whether it's advances in material science or drug discovery or simply writing code faster than you've ever written it before, there's no question that putting AI to work for your competitive advantage is the right approach. Look, the more you study it, the difference between successful projects and those that fail are this mindset shift from technology-focused efforts to business-led transformation empowered by AI, and in that order. Look, the Frontier Firm needs to holistically reimagine the business, putting AI deeply aligned with human ambition to achieve the organization's most aspirational goals and growth potential. At the heart of every frontier business flow is this notion of democratizing intelligence. Again, human ambition at the core, coupled by your AI assistants and agents to get real work done. At Microsoft, this is Copilot and an agent ecosystem aligned to your business infused into the way you work. Simply stated, AI transformation needs to do more for humanity. We need to evolve the approach to become frontier. We have to obsolesce the mundane, unlock creativity and innovation. The world's greatest filmmaker, the world's greatest musician, educator, the cure for the uncurable, advances in climate science and basic research, all of these things that can move humanity forward can be unlocked by democratized access to intelligence. So there's no question that aligning to the business aspirations and putting AI to work is the right approach. But beyond that, if you study successful frontier firms, there are three common traits to their approach. First, they put AI in the flow of human ambition. And this is contrary to somehow cutting and pasting your life into a chat window and expecting insights. It's about AI infused in the tools that people love to use each and every day, seamlessly accessible to get real work done. Second, it's about ubiquitous innovation. There's a maker in every one of us. And the frontier firm has a maker in every room of the house. By putting tools into the hands of people that can create AI artifacts right in the flow of work, the person often closest to the problem and the challenge and the opportunity is the best person to solve it. And that's what we do here at Microsoft. Third, the frontier firm needs to govern and manage the AI they deploy. And it requires observability at every layer of the stack. What agents do you have? How many people are accessing them every day? What are the workflows they're assigned to? What results are you getting? Govern, manage, secure, and optimize the AI that you put to work in your business. Beyond these three traits, there are two fundamental building blocks, intelligence and trust. Now you might say, "Hey, Judson, intelligence is kind of obvious", but I will challenge you that we need to put the "I" back into AI. Because if you look at what's happening today, developers have to navigate a sea of data, thousands of connectors, thousands of APIs, all while sipping data through a thousand tiny straws and trying to inference over it, which is just a fancy way of saying "guessing". Folks, we can do better, and today we're going to show you how. The fundamental building blocks of AI need to know how you work, with whom you work, the content over which you reason. They need to provide real intelligence to developers to help them scale and build better solutions than ever before. Finally, trust. We're going to talk a lot about observability today, but it's about aligning AI to your purpose and your principles to get the outcomes that you want to achieve. At the end of the day, it's about your IQ and your differentiation in your business and in your industry. Today, everything we will show you will be model-diverse, open, and heterogeneous at every layer of the stack, because, again, it is about your ambitions, what you aim to achieve, and Microsoft here is to empower all of those things. The rest of our keynote this morning is organized around these three successful traits of the Frontier Firm. AI and the flow of human ambition, ubiquitous innovation, and observability at every layer of the stack. At Microsoft, we're at our best when we put our portfolio of technology and capabilities to work for your business needs. So accordingly, everything you'll see today is an integrated flow of our capabilities aligned to real business outcomes. Let's get started with AI and the flow, and to do that, please join me in welcoming to the stage Ryan Roslansky.
All right. Thanks, Judson. It is an honor to be here today with all of you shaping the future of work. And you might be wondering, why is the CEO of LinkedIn, a company owned by Microsoft, here working on and talking about M365 Copilot? Well, the reason is simple. The future of work and the future of AI are inseparable. LinkedIn is the largest AI-driven professional product and community in the world. We know what it takes to build AI that works for 1.3 billion people. Now we're bringing that expertise together with Microsoft's enterprise scale. We're bringing Copilot to put AI in the flow of human ambition. Over the next 20 minutes, you'll see the future of work unfold. How we take the most powerful AI innovations and make them yours, built into the rhythm of every job, every role, every day. From developers, security analysts, knowledge workers, Copilot is becoming essential to how work gets done. Today, more than 90% of Fortune 500 companies turn to Copilot as they journey to become frontier firms. Why? Because Copilot is built for work. And there are three things that make Copilot special. First is something that we call Work IQ. Second, Copilot is embedded in the Office apps that millions of people use every day at work. And third, it's your window into the world of agents. But let's start with Work IQ. This is the intelligence layer that enables Copilot and agents to know you, your job, and your company. Work IQ isn't just Copilot's brain. It understands your business, workflows, and relationships. Work IQ connects everything that matters inside your company and turns it into an experience that feels effortless, powerful, personal. It's made up of three things. Data, memory, and inference. Every organization is built on data. Think about it, your emails, documents, databases, line of business systems. Data is what codifies how work gets done, and it serves as your biggest differentiator and your most valuable asset. Think about it, so much of that data is untapped, it's tied up in disparate systems and files. But with Work IQ, Copilot enables you to unlock that data, put it to work to deliver real value for you and your business. Memory is what gives Copilot the ability to deliver personalized responses based on your style, your habits, and your preferences. Tracks and understands workflows, recognizes how people, tasks, and tools all fit together. And then last, inference uses all that rich signal in the data, combined with memory, to make valuable connections, to unlock insights, and predict the next best action or even the next best agent for the job. Most importantly, Work IQ is the AI that you can trust your most valuable information with. For five decades, organizations like yours have trusted Microsoft to securely handle that data. Now, you might be wondering, how is Work IQ different than just using connectors? Work IQ works across all of your data in real time and has full context, which allows you to find patterns and insights. While connectors only pull fragments of the data with no real understanding of the context, which means they can miss critical information, they can ignore relationships, and give incomplete answers. But Work IQ is built into your trusted M365 environment, so it respects all your permissions from the start. With Work IQ, Copilot understands the whole picture securely, instantly delivering faster, smarter, and more accurate results in the flow of work in ways that connectors just can't replicate. Now, look, many of you are asked, like your CEOs, your board of directors, "What's your AI strategy?" Here's the good news. Whether you know it or not, inside each of your companies today, you have what you need. The data, the memory, the inference. And it's all wrapped up in Work IQ. Now, we all know that AI works best when it's woven into everyday workflows. So then you ask yourself, where is work happening? It turns out it's happening in these apps, literally. Billions of Word docs are created monthly. Hundreds of millions of PowerPoint presentations and team meetings are created weekly. Apps like Word, Excel, PowerPoint, Outlook, and Teams. That's where work happens. And they have changed the way that people get their jobs done. They've seamlessly integrated Copilot into the Office apps, enabling productivity directly in the flow of work. And while we all know that Copilot can do a lot inside of these apps, we also know that there are just so many specific use cases, processes, tools that are unique to each company, which is why we've built Copilot to be your window into the world of agents, to handle custom and specific use cases. Agents can take on tasks, handle routine workflows, and automate really complex processes. Now, many of you already know this, you already embrace it, but every organization will run under a distributed ecosystem of agents moving forward, and Copilot is your front end for that. You can now tap into that Work IQ intelligence layer and build custom agents tuned for unique workflows and business needs; because Copilot gives you the ability to build, direct, and coordinate that network of agents, whether they come from you, they come from us, they come from any other partner. Together, Copilot, agents, and Work IQ amplify what your people can achieve. So to recap, you've got your intelligence layer with Work IQ, we've got Copilot embedded in the flow of work, and we have an ecosystem of agents where Copilot is the orchestrator putting agents to work for you so they can execute any task or business process. Okay, that's enough talking. Let me show you a little bit more about how we're bringing Copilot and agents to life for every employee in your organization. I start every morning by pulling out my phone, asking Copilot, "What's on my agenda for today?" Copilot said, "You've got the Ignite 2025 opening keynote organized by Judson Althoff starting at 9 a.m. and going until 11 a.m. today." I then asked "Has Judson sent me any emails today about Ignite?" And Copilot said "Yes, you have an email from Judson this morning about Ignite. He's suggesting giving everyone a T-shirt during the keynote to boost excitement and make the event memorable." Yeah, it's easy for you to clap, but I've got to figure this out now. So, Judson wants 20,000 Copilot t-shirts delivered to this audience. We don't have a lot of time, but I'm going to hand this over to my friends at Zava. This is a fictitious company we created just to show you what a frontier firm can look like. We don't have a lot of time. I'm going to give them six minutes to fulfill this order. So Zoe, over to you. Hope you can pull off a little magic.
Thanks, Ryan. I'm the COO of Zava, a fictional retail company that specializes in intelligent athletic apparel. Like a lot of you, my day is busy juggling lots of competing demands. I'm gearing up for the holiday demand. I'm dealing with security concerns and managing urgent customer requests. I use Copilot to stay on top of it all. I'm going to ask it to review all my emails. And with GPT-5, but powered by Work IQ, Copilot knows who I am, what I work on, and who I work with. It's actually why I use Copilot. It has the context it needs to give me real-time and actionable results. Like, I can see here there's a request for 20,000 branded T-shirts from Microsoft, and I heard the crowd is pretty excited about this, and I only have a few minutes to get it done! I can already see the timer going, so I gotta be quick. I'm going to get a quick overview of Microsoft as our customer. That way, I have full understanding. And Copilot is going to, again, tap into that Work IQ, pull my meetings, my emails, my chats, but it's also starting to infer what I'm trying to accomplish. It understands that I'm trying to dig deeper into my customer. And so, while it gives me this great result, it also recommends an agent that's specially built by my company for customer insights. Now, in this case, the Copilot response looks great, and I need to quickly mobilize my team because work doesn't happen in a silo. So I'm going to start a group chat with my analyst, Anne, and my sales lead, Nirja. And Copilot is going to bring this conversation over into Teams in the same place that I already communicate and collaborate with my coworkers today. And they have all of the context that Copilot just showed. So when I reach out and ping them, they have the same context I do. Oh, they're super excited, I can already see they're replying. And now we can continue to use this to work together and with Copilot as we get this amazing opportunity. But I want to quickly check in with my analyst, Anne. Anne, how's inventory looking? Are we going to make it?
Well, this is a pretty big rush order, so we're going to have to get creative. We'll need to break it up into smaller pieces and allocate it across multiple production facilities if we have a chance to get this done in time. As soon as Zoe pinged me, I turned on Agent Mode in Excel and I set it to work asking it to whisk us up a plan of action. And this is what we're looking at in real time. I can follow the step-by-step process of the agent as it works through the problem and we can see how it's building out inside the spreadsheet right in front of us a solution, and it looks like it's just now finishing up. Because the agent is built into Excel, it speaks Excel natively, which means it can do everything from simple to complex. In the plan here, we can see the scope of our problem, the total quantity, the due date, and the question. And I get the detailed breakdown of the proposed plan in table format. And this is not just static fields, this is dynamic formulas. So as I make tweaks to the plan, everything updates in real time. And not just tables, I also got a chart here from the agent which shows me the impact of the proposed plan to overall production capacity. And finally, I get this beautiful conditional formatting. It gives me a really clear green flag that yes, we can meet the order by the due date. And if Zoe wants to know how I got that answer, I have the entire history of the agent's calculations and reasoning that I can share with her. So Zoe, looks like inventory is good to go.
Excellent, Anne. That's great news. We have about two minutes. Nirja, how's the sales quote coming along?
I'm on it, Zoe. All right, so here I have an email with details about the opportunity from the customer. Now, I need to generate a sales quote and I need to do it fast. So I'm going to jump over to our sales quote template where I already have Copilot opened up. Here, I'm going to activate Agent Mode and then give it a bit more direction to get started. Now, this type of work is very time consuming, but the agent helps lift things off my plate. What I want the agent to do is I want it to pull in information from the email, specific information like the address and the number of t-shirts. But, because of Work IQ, it's going to look across my entire work context, pulling in information like the shipping costs and the pricing that didn't exist in my email. And I can see here that it's getting to work. I can see it co-writing with me as it fills out the form. But I am still in the driver's seat. I have the ability to edit as needed or have Copilot take another pass. But this looks great, so I'm going to keep it all and done. All right, Zoe, here's the final quote for you to send to the customer.
Thanks, Nirja. Okay, let me let Ryan know. We have less than a minute. I'm going to have Copilot help me quickly draft a response saying that we can fulfill this order. I'm going to keep that and I'm going to attach the sales quote that Nirja added and then I'm going to press "Send". Now we can stop the clock! We made it just in time with 27 seconds to spare, so everyone gets a T-shirt. And that's the magic. It's the magic that comes from Copilot powered by Work IQ. It goes beyond connectors to truly understand the context of how work gets done at Zava. But of course, Microsoft isn't our only customer. We're getting ready for the holiday rush. We're pulling on our aprons, we're rolling up our sleeves, and we need to be sure our website is ready to deal with the surge. Let me hand it over to Dean, our Cloud Architect. Dean, how's it going?
Hey Zoe, everything's going great over here. So here's the scoop, I just recently scaled our Kubernetes environment and I added a few new SKUs to our website so now I'm going to test our checkout flow with this new water bottle. So, we'll go ahead and add her to cart, head over and proceed to checkout. Hmm, it's saying that checkout is currently unavailable, which is odd because I know we have inventory on hand. So let me head over to the Azure portal to investigate our Kubernetes environment. Azure Copilot Agents makes this sort of thing super simple. So I'm going to prompt Copilot to investigate the health of my Kubernetes cluster. We'll go ahead and tag the cluster and let Copilot get to work. So, Copilot understands the full context of my AKS cluster. It checks pod health, looks for anomalies, and pinpoints which pods are unhealthy. Here we can see that access to the Cosmos database is blocked by the account firewall. So there might be an API issue, but let's use Copilot to help with this. So, we'll go ahead and ask Copilot to help me troubleshoot my Cosmos database and go ahead and tag the Cosmos DB and let Copilot get back to work. So this time, Copilot's checking firewall signals and service health checks and firewall diagnostics here, and it's saying that network access is disabled within the firewall. And actually, here's my favorite part. Rather than integrate the code, which we can do using GitHub's integration, we can approve the change and let Copilot do it for me. Okay, that was really, really quick, but go ahead and head back to the website. We'll double-check the work by trying our checkout again. So we'll proceed to checkout. Okay, great. Looks like we've got a fully operational checkout page, and wow, that used to take my infrastructure team hours. We just did that in just a couple of minutes. Hey, Zoe, the website is good to go for the holiday rush.
Awesome. Thanks, Dean. For a retailer like us, the volume of holiday business can also expose us to heightened security threats. Our security team is working around the clock, so let's check in with Scott, our Security Analyst, to see how things are going. Scott?
Thanks, Zoe. Now that many orders are coming in, my job as the Security Analyst is to keep Zoe and team safe, and that's why I use Security Copilot directly inside Defender to stay ahead of constantly evolving threats, like phishing attempts. Now, every day, users submit hundreds of suspicious emails to me, and most of these turn out to be harmless spam. The Phishing Triage Agent helps me cut through the noise at-scale and identify malicious emails. Let me show you. It sits right here in my flow of work, and it autonomously filters out the false alarms, so I can focus on what truly matters and what really requires my attention, and that is, of course, the true or real active alerts, such as these. Let's drill into this one. You can see that the agent explains what it's thinking in simple, natural language, which is awesome. It reasons over content, headers, URLs, and attachments, and then visualizes its entire process in this intuitive decision graph. And if I open one of these cards, I can see what it's thinking at any step in the evidence it used. As you can see here, this looks like it is a real order from Zava but unlike your T-shirt order, this one is actually a phishing attempt targeting Zoe. And if I continue to follow the flow from the agent to the end, it does, in fact, confirm that this is a true phishing email. With Security Copilot, I'm able to keep Zoe and Zava safe, faster than ever before. Back to you, Zoe.
Thanks, Scott. As a Frontier Firm, every one of our employees, including our most specialized teams, they rely on Copilot to get work done. We can't imagine working without it. Back to you, Ryan.
Awesome. Well done, Zoe and Zava. Round of applause for them. So, that was a whirlwind tour, and what you just saw was M365 Copilot, Azure Copilot Agents, and Security Copilot Agents bringing AI into the flow of work for every knowledge worker and every domain expert. M365 Copilot is AI built for work. It delivers our richest, most powerful experience. And Copilot Chat is secure AI chat for every employee in your organization. It has enterprise data protection, it gives you access to agents in the flow of work, and it's available to every M365 subscriber at no additional cost. Today, we're announcing two new updates to Copilot tied to some of the innovation that you just saw. First, we're excited to announce we're bringing Word, Excel, and PowerPoint Agents into Copilot Chat. So whether you start in Chat or start in the Office apps, Copilot can actually do work for you, and you stay in control the whole time, like you saw, reviewing, editing, or guiding Copilot to rework as needed without ever having to copy and paste from Chat. Also, Copilot Chat can now reason over and make sense of your entire inbox. It can prioritize and summarize your emails, help you prepare for upcoming meetings, and even highlight action items from previous meetings. All of this is coming to every employee in your organization. Of course, security is the foundation of everything we do at Microsoft. I'm excited to announce we're bringing Security Copilot into M365 E5, putting AI-powered protection and response right in the flow of work. And if you only take one thing away from the past 20 minutes, let it be this: Work IQ is the intelligence layer that knows you, your job, and your organization in a way that's simply not possible with connectors. This is the new era of work where every employee across the organization has an AI thought partner, and there's just no other solution on the market that comes close to the value that we're able to deliver with M365 Copilot.
Excuse me, excuse me, I hope I'm not interrupting anything important. I have a delivery for Ryan, 20,000 T-shirts.
Excellent. That's right on time.
Would you like me to leave them on the counters by the exits, or do you want me to hand them out?
Why don't we go ahead and put these at every exit so that when the keynote is over, everyone can grab one on the way out. But everyone -- please don't get up yet. We've got a lot more to show you because we're not the only ones bringing AI into the flow of work. It's happening across all industries. Teachers, farmers, factory workers, even doctors are transforming their day-to-day with AI. To talk more about that, please join me in welcoming Seth Hain from Epic.
Thanks, Ryan. The real value of healthcare tech should be measured in the human possibilities it unlocks for both patients and their clinicians. In the context of medical care, embedding AI into the flow of work requires disciplined acceleration, combining the immediate assistance of AI and workflow with the increasing familiarity and confidence needed to quickly adopt the next wave. The importance of getting this right has never been greater given the workforce shortages and aging population needs facing our society. At Epic, we create technology to help people get and stay well. Our clients call it their operating system, informing mission-critical decisions from the emergency room to the ICU and, importantly, connecting with you, their patients, through MyChart. In March of 2023, just as generative AI was gaining momentum, doctors started using it in our software to respond to questions from patients faster. Our first AI agent was released in the fall of that year to help study patterns in medical data. Fast forward to today, and AI is embedded throughout our applications, from the clinic to the bedside to the back office, and worldwide, hundreds of thousands of clinicians are using it to improve care. Central to this rapid global rollout was our established AI foundation on Azure and the secure, reliable environment it provides for sensitive healthcare data. AI in the flow of work both saves time and improves quality. In AI-generated end-of-shift notes, discharge summaries, those things help nurses wrap up their day quickly. And with Epic integrated with Microsoft DAX Copilot, doctors' notes are prepared for them based on the conversation in the exam room. Health systems are finding a 30-40% reduction in the time it takes to generate medication prior authorizations. Let me translate that for you: It means you get treated faster. And AI also helps make sure no handoffs in your care are missed. Last month alone, using our software, 6 million summaries of medical records were generated for clinicians. Doctors reported discovering information 28% of the time that they would have missed with manual review. AI also assists in identifying patients needing follow-up imaging. Using our AI capabilities, the Christ Hospital reached a 69% rate of early-stage cancer diagnosis compared to the national average of 46%. Approaching AI with disciplined acceleration helps physicians care for more people today while building toward a future where AI unlocks new possibilities in drug discovery, precision medicine, and potentially understanding disease itself. My colleague, Dr. Jackie Gerhart, is here to share what this looks like in practice. Over to you, Jackie.
Thanks, Seth. As a family physician, AI is transforming the way I practice, and today I'd love to show you how. I'm ready to see my first patient, Oscar. So, before I enter the exam room, I'll pull up his chart. On the right side of the screen, I can see that generative AI created me a summary. This provides me with the most relevant details of his health history. Since I last saw him, he's had genetic testing done because of his family history of high cholesterol. I can scroll down to that section of the summary and click to see his test results in the progress note. Now that I'm up to speed, let's head in. The notes panel on the right side of Oscar's chart is where I would normally type my observations during the visit, but instead, I'll use Epic's AI charting to capture the conversation. I can place a mobile device between the patient and me that will securely record our interaction and turn that transcription into a clinical note. It's sent right back to me here in the chart, and in fact, I can see it right here on the right-side. I love that this feature enables me to focus on face-to-face care with my patients, but the AI didn't just produce a note. It did so much more. The AI picked up on my care plan for Oscar and already teed up his orders and follow-ups like a referral for physical therapy. This makes it so easy for me to review, edit as needed, and sign. Now, one of my patients' biggest frustrations is when their prescriptions are delayed, often because insurance still needs to gather more information. Luckily, AI can help me there, too. It's noticed that one of the medications I prescribed Oscar might not be covered by insurance, but the prior authorization feature can search through his chart and fill out the answers to his questions from insurance company to make sure he gets approved faster. But the story doesn't end when I close his chart. My care plan becomes actionable for Oscar right on his phone in the MyChart app. AI can provide him with a set of to-dos post-visit. In the app, Oscar can see suggested tasks based on my reminders from the visit, like this one to ice his shoulder. And what's really cool is I can check off when he's completed that task so I can follow along with his progress. AI lets me focus on what matters most, my patients, deepening the purpose and joy that brought me to medicine in the first place. Back to you, Judson.
Thanks, Jackie. How about a round of applause for Jackie and Seth? Look, the work that Jackie and Seth do is deeply personal to me. My father was a surgeon for over 50 years and even as a child he would lecture me that there was a mantra about operating at the top of your license as a doctor. The work that Epic is doing to infuse AI into the flow of how healthcare providers deliver care allows physicians to go one step beyond -- not just performing at the top of their license but performing at the top of their potential. We're really privileged at Microsoft and honored to be able to provide the platform and tools that Epic uses to build their AI solutions. We're actually very excited to share more with all of you today about how to leverage the same. Like I said in my opening, there's a maker in every one of us, empowering citizen developers and professional developers and enabling them all to collaborate with IT to create amazing AI solutions and then allowing them to build upon new layers, new fundamental building blocks that embody context and intelligence is what it's all about. Here to share more about ubiquitous innovation, please welcome to the stage Asha Sharma.
Every person in every role now holds the power to create right inside of the tools they already know and love. From the soil of a Nebraska farm to the code running a global bank, innovation is no longer the domain of a few. It has become the superpower of many. Work will be defined by human creativity and agents that help it scale in this era. And at the center of this shift across individuals, domain experts and developers is Microsoft. Let's start with the individual maker inside of all of us. Every one of us has had an idea for an app that would make our lives just a little bit easier. But today, turning that idea into something real takes someone technical and often months or years of work. Not anymore. Today, we're announcing App Builder, so anyone can create an app in minutes right inside of M365 Copilot. All you do is start with your data you already have, your charts, your files, your meetings, all powered by Work IQ. You design the experience you want in actual language, and you share it with your coworkers just like you would a doc. This is the beginning of something new: Software evolving from being made for people to being made by people. Now, innovation doesn't just stop at the individual builder. Some ideas require more domain expertise, more data, and deeper connection across your business. Copilot Studio is the only low-code agent builder that understands your business right out of the box. So it uses Work IQ, workflow, security, identity, to take real actions and not just generate basic prototypes. Today, more than 230,000 organizations use Copilot Studio, including 90% of the Fortune 500. Now, for decades, developers have lived at the frontier, and they have been the ones to solve the world's hardest problems. And that doesn't change in this era. What changes is the leverage. We're bringing them new AI-powered tools to continue to build the future of what's possible, and they do this on GitHub. This year, someone new joined GitHub every single second, which is the fastest growth we've ever seen. GitHub Copilot started as a pair programmer, and now it's become an entire fleet of coding agents. And now, it is also the single largest contributor to GitHub's code base. But we know that developers want choice, enterprises want control, and everyone wants fewer subscriptions. That's why, at GitHub Universe, we announced Agent HQ, a single place to pick agents from GitHub, OpenAI, Anthropic, Google, Cognition, xAI, and more; all with consistent governance and observability as part of Copilot subscriptions. Now, when innovation becomes ubiquitous, everyone gets a new set of powerful tools to drive progress forward. And in a retailer like Zava, that means the frontline associates in the aisle, the regional managers are running performance across the stores, and developers are writing the systems to code every single line. That's exactly what we're going to show you. Please welcome Miti, Lydia, and Kyle to the stage.
Today at Zava, the store is packed and sales are strong; but as the store manager, I am scrambling behind the scenes. Every shift change means tracking down who's free, having multiple forks to negotiate swaps, and calling in favors just to keep things running. I really need some help. So I reached out to Lydia over in the store Operations team, who says there's no bandwidth for this right now. There has to be a better way. Lydia says I should check out Microsoft 365 Copilot, where I can build apps, agents, and workflows. Now, I want to build an app to solve my staff's scheduling challenges, so I'm going to go ahead and ask it to build me one. I'm going to ask it to create an app with a dashboard and rich visuals so I can easily manage staff scheduling. Now, App Builder understands what I'm asking for, and it immediately begins generating the app right here inside M365 Copilot. I'm really hoping it uses all the documents, spreadsheets, notes, and emails I use today to manage schedules. And here is the app it built. And I can see here a dashboard for shifts across time, a manager view just for me, a visual for store coverage, and even a table for our team members and their shift. Most importantly, App Builder built on top of Zava's Microsoft 365 data, and it understood the business the way I do. I did not need to learn any new tools, I just described what I needed. No more frantic calls, no more guessing who's free, because I've got an app to do it for me. Now, I'm so excited with what I've built here that I'm going to share it with Lydia to see where it can go.
And the second I saw it, I knew this wasn't just going to help one store. All of my stores could use this. Plus, I'm going to address another challenge. When someone can't make a shift and chaos ensues, enter Microsoft Copilot Studio to help me build an agent to fix that. So I say to Copilot Studio, "When someone emails that they're sick, I want an agent to secure a replacement team member." Instantly, Copilot Studio jumps into action and creates the full agent framework. Copilot Studio has native access to Work IQ, including Microsoft SharePoint and the Microsoft Graph, so it's grounded in rich context. First, I need to determine when I want the agent to be triggered. When a team member emails our scheduling inbox to say they're sick, I want the agent to jump into action. So we'll go ahead and choose that, go ahead and give it an appropriate title. And just like that, my trigger's added. Next, I want to go ahead and I want the agent to interact with a staff member to confirm their availability. To do that, I'm going to go ahead and add a Copilot Studio flow. This one's called "Contact Replacement Team Member", so we click on that. And just like that, that's added to my agent. Finally, I want the store manager in the physical store to be updated about which staff member will be covering the shift. I'm going to use an existing agent that my colleagues built in Microsoft Foundry, which drastically speeds up my efforts. All we need to do to integrate Copilot Studio with a Microsoft Foundry agent is give it a name, add the description, and the ID. So let's play this out. It's 7:45 a.m. on a Saturday, and Wim, a team member, says they're sick. Before anyone sees the email, the agent jumps into action. It starts by suggesting a replacement team member based on skills and availability, then contacts them to confirm they can cover the shift. Finally, it goes ahead and contacts the store manager and lets them know that in this case, Luca has confirmed he's going to be able to cover the shift. And now, my Zava pop-up stores have an app and an agent working together to ensure every shift is covered inside Microsoft 365 Copilot and Copilot Studio. And now, on to the next task on my to-do list, I want to handle this message that Ryan sent. It looks like the new e-commerce site isn't currently optimized to work for mobile devices. So let me go ahead, file that as an issue with GitHub, and let Kyle from our development team figure that one out.
Thanks, Lydia. I'm a developer at Zava, and I work on our online retail store. We use GitHub for our entire software development lifecycle. While I'd normally be making my coffee right now, I'm going to start by burning down issues in our backlog. And I've asked Copilot to tell me what's on my plate. GitHub Copilot gives me a ton of choice with models from all the frontier labs. And as of earlier this morning, that includes Gemini 3 Pro, and we even have a custom fine-tuned model deployed to Azure Foundry. So here's the issue from Lydia, improving the website to make it more mobile-friendly. Okay, honestly, this is important, but it's kind of boring to do. So, I'm going to ask Copilot to start work on this. And because this is a UI task, I'm going to use the custom coding agent that our team built, focused specifically on front-end changes based on our frameworks and styles. Let's also ask Copilot to include screenshots so I can easily see the difference without having to run the app. Copilot gets to work right away, planning and writing code, tests, and opening a pull request for my team to review and merge later. I can manage this task in all of my agent sessions in one place: Agent HQ. Okay, Copilot has that task in-hand so let's go be a team player and work on some security remediation for our project. Our Security team uses Microsoft Defender for Cloud to monitor our security posture and find risks in production. Today, we are announcing a new code-to-cloud integration between Defender and GitHub Advanced Security. So, we can filter for vulnerabilities using runtime data like sensitive data runtime issues. Our security team has created a campaign to tackle these vulnerabilities. They choose which vulnerabilities to focus on and a deadline to get it done so we have an achievable target. I'm going to select all of these vulnerabilities and assign them to Copilot. Copilot will generate these fixes all in a single pull request for our team to review. Now, I also spend a bunch of time in VS Code. And of course, I use coding agents to help with my work here, too. Here, I can see the agent sessions related to the code I'm working on across local chats, CLI sessions, and even third-party agents like Codex. I can even start new tasks from right here, and they'll run in the cloud. Now, I realized I haven't written any tests for this code, so before my team calls me out in a pull request, let's add some tests so we can make sure that we're using auth on all of our pages. At Zava, we have developers using Copilot in a lot of different ways. And now, all enterprises can see how Copilot is being used within their organization with Copilot metrics. I can see usage statistics, what models my team is using or not using, and even the programming languages that Copilot is assisting with most. Okay, let's go back to where we started with Agent HQ. We kicked off a bunch of coding tasks, and I can see all of that work here. Now, I want to take a look at the mobile task I created earlier to resolve Lydia's issue. I want to give it a little more context, and I can just do that while it's running and let it know that I only care about phone sizes, not tablets right now. So in only a few minutes, I got Copilot to remediate security issues, start making our online retail store more mobile-friendly, and improve our testing so I stay off the "naughty" list this year, all while keeping me in control. What we just saw was Miti using App Builder to solve a scheduling problem, Lydia using Copilot Studio to scale it with an agent workflow across their region, and I used GitHub Agent HQ to tackle three tasks in two minutes. That's the power of ubiquitous innovation for each of us. Back to you, Asha.
Thanks, Team. Incredible work. Now, with this new work and this new world of creation, people that build AI-driven products need a very different foundation, one built from models, tools, and intelligence. That's why we built Microsoft Foundry, an open and modular app server for the AI era. Foundry has powered nearly every single product that you've seen on this stage today, and as we mark our first year, the momentum is unmistakable. Over 80,000 organizations are building on Foundry, and today, we are the number one AI platform in the world. Now, we find ourselves in a multi-model world. Foundry has become the home for the leading models like OpenAI, Mistral, DeepSeek, Llama that your teams use every day. But there's been one frontier provider that we've been missing. This morning, we changed that. Anthropic models are now live on Foundry. Please join me in welcoming Mike, the Chief Product Officer of Anthropic.
Thanks for being here. What a year it's been together.
It's been an amazing year together. I am so excited to be here and so excited for the news.
Amazing. Do you want to -- let's tell everybody what we launched this morning.
Absolutely. So, our partnership together is really about enabling developers and enterprises to build intelligent agents with Claude no matter where they are. And that's why today we announced the availability of Claude Sonnet 4.5, Haiku 4.5, and Opus 4.1 models, all our frontier models, in public preview on Microsoft Foundry. Claude on Foundry enables customers to build, scale, and secure production-ready applications and enterprise-ready agents in one trusted platform.
Now, we've been finding ways to work together over the last year, but we've been cooking up some exciting things. Do you want to show everybody what we're going to be rolling out over the next few months?
Absolutely. So we've already partnered together to bring the Claude family of models to GitHub Copilot's 26 million users. We've been collaborating closely on the MCP standard for tool calling, and we've been offering Claude as part of M365 Copilot Chat for billions of users. Now, we're taking the next big step by bringing Claude models to all of Microsoft Foundry. In this example, we're using Sonnet 4.5 and Skills to build a product plan for our favorite line of socks for Zava. Claude builds the plan by using Work IQ to find context about market details and potential pricing, and then it uses the Zava branding skill to create potential taglines that feel on-brand. Next, we use Claude to create a PDF by using the PDF Skill, and this quickly creates a beautiful, fully designed executive summary to share with the team. And then finally -- and this is cool -- we'll use M365 to send the PDF back to the two of us. So this is Microsoft's enterprise knowledge combined with Claude's agentic capabilities and Skills.
The thing I love the most about this is Claude Skills working together with Work IQ. I feel like we're the only two companies in the world that can bring that type of use case together for enterprises.
Absolutely. And this goes beyond knowledge work to coding as well. So, thanks to our partnership with GitHub, Claude has become a first-class agent, which means your developers can collaborate with Claude as part of GitHub's Agent HQ. So example, you can start by assigning Claude an issue just like you would a teammate. Claude instantly picks it up, creates a branch, opens a pull request. And then from there, you can interact directly with that PR. You can ask questions, request changes, iterate together. When it's ready, you can merge, and that's it. So this is Claude working alongside you like a coworker directly in your team's workflow. It'll write code, learn from your feedback, and help you move faster.
Amazing. You know, you have been so thoughtful about your model advancement, safety, everything you're doing in coding and enterprising, and everyone you partner with. I'm sure everybody here and everybody at home would love to know, why is our partnership different than anyone else's?
From the very beginning, there just felt like there's been a lot of shared DNA and trust across both companies, and I'm really excited about what happens when you combine the power of our trusted models with Microsoft Foundry and M365. So we're making Claude available to enterprises that have already invested in Fabric and Foundry and M365, and removing barriers like navigating separate vendor contracts and billing systems so developers can get started with Claude right away. I can't wait to see what you all build. Stay tuned for much more to come.
Now, what's unique about this moment is that now Azure is the only a hyperscaler to offer both OpenAI and Anthropic models to fully empower our customers to build the best AI applications on the planet. So with all this choice, picking the right model for the right job to be done becomes even more critical. That's why today we're also announcing the general availability of our new Model Router. This Model Router automatically selects the best model based on accuracy, performance, cost, or balance. It enforces US and EU data boundaries by default, and customers are already seeing 50% lower latency and 15% higher quality. So with models multiplying, tools everywhere, we've never had so much raw power. But raw power is not enough. Models don't understand your margins, tools don't understand your customers, and most solutions only understand the "what" of the work, not the "why". The "why" lives in documents, lives in chats and meetings and web pages and files. And in every organization, this sits in two oceans of data, one that's structured and one that's unstructured; and between them lies this intelligence gap. Agents should not sit on top of these systems. They need to live in between them. They need to be able to see across silos, reason through context, and act on behalf of people. That's why today we're also announcing a new intelligence layer, one that knows how your company works, understands your business logic, and can take real actions in the real world. It is made up of three connected planes, Work IQ, Fabric IQ, and Foundry IQ. Now, Ryan talked about Work IQ, so let's start with Fabric IQ. Fabric IQ builds on 20 million semantic models in Power BI that encodes your business logic. Now those models are going to extend beyond analytics and into operations so humans and AI can share the same understanding of your organization in real time. And because Fabric IQ is integrated with Microsoft 365, every single user benefits. In Power BI, in Excel, in Teams, in Copilot, all with security flowing automatically. To take us deeper, please welcome Zia.
Thank you, Asha. Imagine this: You need to understand if Zava should open a new pop-up store. Our store operations dashboard has an agent built on Foundry taking advantage of our data in Microsoft Fabric. You start by asking for markets with high demand signals, but the agent didn't understand what demand signals meant or where the more relevant demand data is. Results by just dropping an agent on top of your data can be mixed at best. AI needs an understanding of what the data means and how things relate to take further action. With Fabric IQ, you can adopt the language of your business, defining entities, and connect them across all your data in OneLake. From real-time data, semantic models, and even mirror data from sources like Snowflake, Oracle, or Google BigQuery. You're not just analyzing data, you're giving AI and your people a shared understanding of what it means so they can act on it in real time. Similar to what semantic models did for business intelligence, Fabric IQ is doing for AI, making results consistent, accurate, and transparent. With the model built, you can now see a live unified view of your organization and visualize how everything connects, from relationship graphs to supporting reports and other business data. You can also trigger rules to take action and leverage agents to help run your business through automated decision making. Using Fabric's geospatial capabilities, you can instantly see how shipment delays impacts demand across regions and where opportunities are emerging. After adding Fabric IQ as a knowledge for our agent in Foundry, it looks like Boston is the strongest candidate for the next pop-up store. We now have confidence in the results provided by our AI solution. It has a deeper understanding of our business and provides much more relevant guidance. And look at that jump in quality. Same question, same agent, but now it understands us. Adding our business knowledge turns a generic response into something actionable and tailored. Now that we've published Fabric IQ, it's ready to fuel other solutions and any agent, including Foundry. From unified data to unified intelligence, this is IQ in Microsoft Fabric.
Thank you so much, Zia. Now, if Fabric IQ is shared business logic, Foundry IQ is the intelligent connection point across all of your structured and unstructured knowledge that your agents rely on and take action on. This is the first large-scale implementation of context engineering, extending RAG built for agents. And unlike traditional RAG, Foundry IQ doesn't just retrieve; it plans, it reasons, and it iterates across Work IQ, Fabric IQ, Blob storage, the web, and more. So to put it to the real test, let's build a multi-agent system, end-to-end, live on stage. Please welcome Elijah to bring it all together.
As a developer at Zava, my job is to make sure that our store ambassadors can focus on customers while our AI agents watch the store. Let's dive in. Here we can see our workflow with three different agents. We're going to talk about all of them today, but first I want to talk about our ZavaStoreOperations agent. This is the agent that knows everything a store manager could possibly need. Let me show you how I built it. Every agent in Foundry starts with a model from the Foundry catalog. Foundry gives us same-day access to the latest open-source and frontier model releases, so we're always working with the best the moment it's available. The Foundry catalog even lets me compare two models against each other to see which one is best for my use case. Today, I'm going to be using DeepSeek-V3, which gives me the perfect blend of speed and accuracy for our low-margin retail environment. Now, let's go ahead and create our agent. I'm going to name it "ZavaStoreOperations", and I already have it ready, so let's go check it out. So you can see here, first I gave the agent some instructions, and then I connected a knowledge base using Foundry IQ. We're going to talk a little bit more about Foundry IQ here in a second. But first, let's give it a real task and help plan a sale for the upcoming CIM marathon in Sacramento. I'm going to use the starter prompt and kick that off. While that's running, I'm going to use Foundy IQ-- our knowledge based in Foundry IQ contains a bunch of different pieces of information, including our product catalog, store inventory, employee handbook, and local market intelligence. Foundry IQ, it's possible to do this using a wide variety of data sources, including Azure Search Index, Blob Storage, Bing, SharePoint, OneLake, and Fabric IQ. Fabric IQ, which Zia just showed us, gives our agents real-time business context on our store operations. Now, if we go back to our agent, we can see here that it created a full plan for our marathon sale, pulling in info from previous events using Foundry IQ. But let's take it a step further. I'm going to ask, "What should I do if there's an item on the floor in the store?" And what's going to happen is our agent will then pull from Foundry IQ to pull out information such as our safety policy documents and other information to give us the answer. Foundry IQ supercharges our agents to be able to make critical business decisions quickly. So not only does this agent help us plan, but also react to what's happening in the store. Speaking of reacting, let's go to our next agent, which is our video processing agent powered by the Foundry Video Indexer. This agent watches our store in real time and emits inventory events, like if there's low inventory or if -- whoops -- somebody knocked something on the ground. I don't know who knocked that shoebox over, but we'll come back to that. Zooming in on this agent, you'll see that Foundry's open ecosystem lets me run Microsoft Agent Framework, LangGraph -- which I'm actually using here today -- or other frameworks on Agent Service. This is all possible through our new hosted agents, which handles the deployment, lifecycle management, and auto-scaling for you. Foundry's open ecosystem also gives me access to over 1,400 enterprise-ready tools and MCP servers. And because it's all A-to-A compatible, it interoperates seamlessly with agents built in Foundry, Copilot Studio, or any other compliance system. Let's go ahead and check out our store camera feed, and if we see here, there is a shoebox on the ground. And it says right here, and our agent picked it up, saying, "Hazard, shoebox on the floor." Pretty cool. So now, finally, I'll add our "PolicyAndNotification" agent, which uses our new agent memory to remember previous conversations and personalized messages to store managers. And so here we can see our agent memory. Pretty great. And so, returning to our workflow, if the video agent deems an event requires action -- and I just got a Teams ping, we're going to get back to that here in a second. So first, our video processing agent will monitor the situation, and if it has an event that it feels like it requires action, it'll pass that to the Zava store operations agent, which will, in using Foundry IQ, will add context around store layout, safety policy, and risk level. And finally, the PolicyAndNotification agent notifies the store manager. And Asha, I don't think you sent me that ping, so let's go see who it is. And if we flip over to Microsoft Teams, there it is. Cleanup on aisle six. Our workflow just notified us that there's a white cardboard box currently lying on the floor. Fully automated, real-time, end-to-end. We just saw how easy it is to build multi-agent workflows with the all-new capabilities of Microsoft Foundry. Only Foundry brings together multi-agent fleets, hosted agents, and live enterprise knowledge with Work IQ, Fabric IQ, and Foundry IQ across any store, any region, and at any scale. And it works with any model, any framework, and any tool. And with that, back to you, Asha.
Thanks so much, Elijah. With every platform shift, it redefines how we create. Computing made it personal, Cloud made it connected, and AI makes it ubiquitous. And that is changing how every company needs to operate. What you saw today expands what work can be. Every person can now shape intelligence to fit their job, their team, their ambitions. And they can stop scaling by effort and start scaling by capability. And they can do that with the tools that they already know across M365, Copilot Studio, GitHub, Fabric, and Foundry. Judson, back to you.
Thanks, Asha. Well, I'm excited to talk about ubiquitous innovation in real life, and to do it with one of my favorite brands, Mercedes-Benz. And I'm pleased to be joined by Daniela, the CIO of Mercedes-Benz North America. Daniela, welcome to the Ignite stage.
Thank you, Judson. I'm really happy to be on the stage with you together.
So Mercedes-Benz has truly been one of our best partners on this AI journey. And we chose to highlight our partnership here because you really have represented ubiquitous innovation across the firm.
Maybe tell us a little bit about your journey with Copilot, GitHub, and the other assets that you've deployed across Mercedes-Benz.
You know, next year, we will celebrate 140 years since the brilliant engineer Karl Benz invented the automobile. But it was his wife who took it on a long ride and made it really popular. So, innovation is deeply rooted in our DNA and fast forward to today, when we talk about AI, AI for us is our best digital colleague. Talking about Copilot, we started with 10,000 licenses. We quickly increased to 15,000. Now we are heading towards all employees have their own Copilot. I extended my workspace with another monitor, so I have my Copilot working for me 24-7. With Copilot Studio, we have 52 areas in our company that are currently automating workflow and creating their own agents. My favorite example is the finance team that, without IT, without any coding, then automated the invoicing workflow. You mentioned GitHub. GitHub is the place where the best minds behind the millions of lines of code in our company are meeting together. And since the introduction of the GitHub Copilot, we see 70% increase of the engagement. So the developers nowadays just focus on creative tasks, on more logic-related topics, and the AI does the routine coding. So, with the Microsoft AI tools, we actually give a lot of power in the hands of our employees, and they can do this really faster.
That's awesome. So you've also done a lot of work with AI on the front lines in the factories. Maybe you want to share a little bit about that innovation and the unlock that's provided for your employees there.
Yeah. We are bringing AI also on the factory floor, and we have MO360 Data Platform that is connecting 30 passenger plants across the world. So, we have real-time data, historic data, so we can analyze this data and improve the efficiency across the factory. With the Digital Factory Chatbot ecosystem, now we have an intelligent multi-agent AI system, so we can get all the actionable insights in just an instant. Maybe we can have a look at the demo that we provided.
