Microsoft AI Tour, Bengaluru ｜ Keynote session by Satya Nadella

Name: Microsoft AI Tour, Bengaluru ｜ Keynote session by Satya Nadella
Uploaded: 2026-04-15T13:26:43.814Z
Duration: 55 min 29 s
Channel: Satya Nadella
Description: Relive Satya Nadella’s exclusive keynote on at the Microsoft AI Tour in Bengaluru. In this session, Microsoft’s Chairman and CEO shares how AI transformation and human ambition are together shaping the next era of growth.

Satya Nadella

3 Video Views·Apr 15, 2026 #BecomingFrontier

Relive Satya Nadella’s exclusive keynote on #BecomingFrontier at the Microsoft AI Tour in Bengaluru. In this session, Microsoft’s Chairman and CEO shares how AI transformation and human ambition are together shaping the next era of growth.

We find ourselves in a new era. One where access to information is access to expertise. From the farm, to the lab, from the boardroom, to the classroom. This new generation of AI is for everyone, everywhere. Now anyone can save time with a personal assistant. You're able to have an agent that is able to detect the patterns, able to make sense of it, and have it speak to the direction of the investigation itself. Anyone can access a personal tutor to learn new skills. This is a new way to analyze with a personal coach. AgriPilot.ai allows farmers to avoid guesswork and use science to make them successful. AI is unlocking creativity for us all. Shiksha Co-Pilot is a generative AI solution where we want to empower teachers to create engaging content for their classroom. AI is supporting us to provide opportunities for the individuals and youth of the country. And we are just getting started. All right. Good morning. Good morning. It's wonderful to be back here in Bangalore yet again and see the energy and the place. And not only is everybody talking about what this platform shift is all about, using all the tools that are available, but most importantly, building. What's always exciting as a platform company for us is to see the builders who go about taking any platform shift and making it their own, right? Bringing their creativity, their entrepreneurial energy to it is just super to see. So I want to talk about how, as builders, we're moving the frontier. At the end of the day, when you think about it, for us, it always starts with empowering every person, every organization on the planet to achieve more. That's why we build. And that means we're moving the frontier of outcomes for those individuals or those organizations, right? We're not building for building's sake. We're building to have impact. Ultimately, it's about the customer experience that's changed, the employee experience that's changed, the operational efficiency that we are able to bring to any institution, the innovation ramp itself, right? So any place where you're able to move the needle on how fast can you innovate, that's ultimately the measure of any technology. And to us, to go about doing it, one of the most exciting things, quite frankly, is the change in the approach to how we build these systems. Perhaps unlike any previous period of software, this one requires a rethink in even the life cycle of how we build, right? I mean, think about it. We start not with sort of the spec itself, right? You start on the other end by thinking about the eval loop, right? So in some sense, you start with the test, not the spec is how I sometimes think about it. It inverts the entire process and that requires your mindset shift, right? So you kind of have to be a lot more grounded in the real world outcome that you want to change. What's the world, the rubric, the job, the task, the eval? That's kind of where you first go, right? And then you back into it the rest of the workflow. So to me, mindset shift, even for developers, is going to be important. Obviously, we need to pick up all these skills that we need in order to be able to impact that mindset shift, right? We need a tool chain that works with us in order to go then build these solutions. And data, it becomes, when we talk about context engineering, it's a little bit of good old data engineering, but it's a different type of data engineering where you want to get your data into the shape that allows you to then feed it to the intelligence and then really orchestrate around it. So that's a bit of how I think we need even a new way to think about the frontier of the classic SDLC has to change to a new form of AI-driven SDLC. And I think that that's going to be pretty critical. Now, to that end, what we're doing at Microsoft is just building out a full stack to help get leverage on it. The first thing is, at the experience layer, can we put this stateful co-pilot experience across all of what you use every day, right? Whether you're using the co-pilot app itself or you're in Excel or you're in Word or you're in Outlook or Teams or what have you, can we bring essentially this ability for you to interact with an agentic system, starting with conversation, request, response, but going fully agentic, in other words, assign tasks and come back and inspect the output and discover other agents, not just one agent, but many agents can be discovered in any one of these experiences, right? So building out this experience layer. And that's what we have done. And it's great to see, like, if I look at even my everyday experience, right, it's no longer just a chat-centric experience. When I look at something like Researcher and Researcher with Cua, Researcher with Cua, it's like having a researcher that you assign tasks to and letting them go off and really do deep research for me and use computer use as just another tool, right? So that's what Researcher is. Analyst is like having a data analyst with you all the time, so you can upload or point it to a OneDrive full of Excel spreadsheets and say, hey, just give me insight, right? I mean, that's my favorite query, is to just dump a bunch of spreadsheets and just ask it questions, starting with the blank question saying, please give me insight, right? I mean, it's just liberating that you can now do that. And then, of course, the agent mode in Excel. I think the next big favorite thing that all of us are going to be vibing with is this agent. It's crazy. I did this digital challenge at the Excel Championship. I mean, to watch as someone who has worked on Excel for all my life, this was so cool to be able to see the agent mode understand natively all of OfficeJS, every object of Excel, to be able to do that file grade manipulation. So anyway, so I'm really excited about these agents that are now very much part of the co-pilot experience. Today, I mean, there's great momentum in this country in terms of these applications getting deployed deeply. I'm very, very excited, in fact, of the announcement this week and today with all of Cognizant and Infosys and TCS and Wipro where they're deploying this at scale over 50,000, yeah? So the thing that I love about it is not only are they just building out this and deploying co-pilot everywhere, but they're building deep agent factories, AI factories, which then integrate into this experience, right? So it's not about, think about any one of them being able to transform our collective clients. They're able to then take the experience layer, have it meet the agents that get built for every task, every job, every role, every business process. That's the opportunity that we have, and I'm really excited to see all of this. So just to show you a little bit of what we've been up to on the co-pilot side, let me introduce my colleague, Zoe. Zoe, come on up. Thanks, Satya. Hi, everyone. I'm Zoe from the Microsoft 365 co-pilot team, and like all of you, although with a little less traffic, I start my day commuting in and preparing for the day ahead. For that, I use co-pilot on my mobile. Hey, co-pilot, what's on my agenda for today? You've got a few things lined up today. First, there's an AI tour India meeting that started earlier this morning. Then at 11 a.m., there's a Contoso sales call. Later in the afternoon, you've got a team sync at 2 p.m. Thanks, co-pilot. And do I have any emails from my manager? Yes, you've got a few emails from Nicole. Just yesterday, she reminded the team about a Friday mixer happening in the common area. She also mentioned that finalizing the monthly business review should be a high priority. She also shared news about it. Thanks, co-pilot. Now, co-pilot, it doesn't just have my work data. It understands who I am, what I work on, and who I work for. It's why I use co-pilot. It's my AI assistant. Now, I want to dig deeper into the ask from my manager to get an update on our monthly business review. I'm going to ask co-pilot to pull that together for me. Because co-pilot has all the right context, it can provide me with a real-time and very easy understanding of what I need to accomplish so I can quickly get caught up, like the current status and key decisions I need to make. Now that I feel caught up on what I need to do, I want to go and update my monthly business review template and ask it to update this existing doc with our November data while still maintaining the exact framework. For that, I'm going to use agent mode. Agent mode in Word allows the agent to edit, format, and draft directly onto the left-hand side. Now, it still keeps me in complete control, but it takes a lot of tediousness off my plate. I can start to see it's looking across all of my work context, my emails, my meetings, my team chats, my files. Now, that sort of thing used to take me hours to accomplish. I have to find the information, make sure it's the right thing, copy and paste it into this template. Well, I'm going to show you how the agent can take all of that tediousness off my plate and actually bring it onto the left-hand side myself while still respecting the format on the left. And I can see it's already starting to do that, highlighting it in blue all of the changes it is making as it runs through. I could also do things like bold or change all the font to green, but I can see it's quickly working its way through my whole document. Again, maintaining the same format, just like if it was a scorecard, but updating and adjusting it, but also maintaining some things the same like the mission and objective. It skipped that. There's no change month to month. Next, I want to show you how I can deal with my team budget and understand my marketing activities for the next fiscal year. Again, I'm going to use agent mode, and I'm going to ask it to help me better understand our fiscal year 25 marketing spend, what's committed, what's planned, what's at risk, and to build it in a visual dashboard so it's easier for me to understand. Now, just like agent mode in Word, agent mode in Excel will natively speak Excel on the left-hand side and start pulling together not just numbers, but actual formulas, native charts and tables, things so I can actually understand what agent is doing inside of my worksheet. And the whole time, it explains on the right-hand side what it's accomplishing, like setting up the dashboard and what it needs to do. That way, if my manager, Nicole, comes to me and says, how did you get to this answer? I'm able to walk step-by-step of what the agent did. Now, again, this used to take me hours to accomplish, especially just trying to get an overall insight, as Satya said, on some of my data. But here, agent mode lets me stay in the same place my data already is, but continue to work with it while building it and pulling all of this kind of messy data into a simple way that I can further understand. Agent mode in Excel, it also can do simple things like conditional formatting. So if I just want to highlight in red all the at-risk metrics, it can do that. And right before our eyes, in just a minute or so, it's built me an entire marketing dashboard. If I click in, I can see the complex formulas it's used and that conditional formatting to highlight things that are at risk or committed. It even shows that it's created a table, which, again, is completely native to Excel. I can edit and adjust it as needed. Now, Copilot, it's not just made me faster at my job, but it's actually helped me be more impactful. And I can't imagine working without it. Back to you, Satya. Thank you, Zoe. Thank you so much. You know, it's exciting to see the progress. I believe there's something I was using last night that I'm supposedly not supposed to talk about it, but tomorrow we'll talk more about it. There'll be a new model drop as well, which will take that entire experience to the next level. So now that you have a bit of this experience layer, the access to agent mode, and not only with the agents we have built, I mean, the thing that I'm most excited about is any one of your agents will be discovered in the context of those canvases, right? That's, to me, what's going to be very, very exciting and game-changing. The next layer down from it is what I think of as the IQ factory layer, right? When we say context engineering, which, by the way, is one of the most important engineering considerations, which is if you have the experience, you have your agents, then feeding it the actual contextual data, which is relevant, is going to be very, very important. And in that context, the best bootstrap is perhaps the most important database that is there in any organization, which is the database underneath something like Microsoft 365, right? You can bootstrap the context with the people, their relationships with other people, their relationships to their work artifacts, whether they're in OneDrive or in SharePoint, their projects, their teams, meetings, and conversations, all email, everything brought together as work IQ. Then take the other piece, which is everybody is using Power BI. And in fact, if you look at Fabric, Fabric not only has all of the Power BI data, it has, in fact, operational data today, it has analytic data, it has real-time data, it has time series data. So Fabric has all of this, but most importantly, Power BI has bootstrapped the semantic layer, right? So it's just not the data, but it has the semantics around the data. That's what we mean by Fabric IQ. So you can bring all of that. So you have work IQ, you have Fabric IQ, and then this Foundry IQ takes anything that's unstructured and brings that also. Essentially, then you can rag over all of it, right? So if you think about, and not just simple rag, you can do graph rag, you can do multi-hop rag. So you can take work IQ plus Fabric IQ plus Foundry IQ and have Foundry just build the right context that you feed in to your model layer, right? So that's what this IQ layer does, and we're very, very excited to help, really take the agent development to the next level with this. So you kind of now have the experience, you've got all of this data and the IQ layer. The next thing is starting to build agents. And so this is where the tool chain matters. So for us, the agent factory is, and we have many, many entry points, right? There's just agent builder built right into Copilot. We then have this app builder, right? So this is prompt to app. In fact, Bill's all time thing from the day I joined in 92 to now has always been, why the heck do I have these differences between what is a document and a website and an application, right? What's the difference? Guess what? There is no difference. You can, in fact, now create a document or an app or a website just with something like an app builder. It's pretty exciting to see that. And of course, Copilot Studio brings the same concept to agents. Can I start with a natural language prompt and just prompt my way to building an agent by grounding it in a set of knowledge and what have you, right? So Copilot Studio is the next sort of piece of the tool chain. And then there's no cliff, right? So you can go straight from here to Foundry. And Foundry brings all of essentially the services that you need in order to be able to build not only an agent but multi-agent systems. That's essentially what the Microsoft Foundry does. Of course, you have the model layer and we have 11,000 plus models, right? We have all the usual suspects, keeps on, there's model drops every day and you have access to all of it so that you can then go build all those multi-agent and multi-model systems. And you could say, man, why the heck do I need 11,000 models? Yesterday, I had a chance to sort of read that cell paper on GigaTime which Microsoft published because of the work we did with Providence and University of Washington. And at some level, even for me when I read that and sometimes you're all playing the game and you forget sometimes why you're playing the game. It was a great reminder of why are we playing the game. So what GigaTime does in simple terms for me is today if you really wanna figure out whether immunotherapy will work for a particular kind of cancerous tumor, you have to have this very sophisticated test called IMF which costs a lot of money, which takes a lot of time. And so if you said, so what this team did is they said, okay, can we simulate the test using Gen AI? So just imagine you take a simple pathology slide but simulate the test from that slide. So a microscopic simple image but sophisticated spatial proteomics in essentially simulation. And then you can sort of start figuring out whether the immune system is working or not and then sort of precision oncology can become better. And then you put something like this together with say Ayushman Bharat, right? So some startup here will figure out saying, you know what, we're gonna make this possible for any sort of hospital in any city in this country to be able to do tests using these models that'll allow us to do precision oncology at scale. Like that to me is why I think all of us should just double down and why we are playing the game, not just playing it and forgetting the context in some sense. So it's just fantastic to see that. And to me, building all of this out as a first class tool chain in Microsoft Foundry and then wrapping all of this with the developer tools, right? For us, that identity as a developer tools company with GitHub, with Visual Studio and everything is just a very huge priority. And so to me, seeing GitHub's progress, in fact, India is slated to become the number one community in the world with GitHub by 2030. And it's just, it's fantastic to see the engagement and the type of projects that are coming out of this place, the contributions from India. And so we're building out GitHub as essentially this agent HQ, right? So in fact, one of the fun things everybody does nowadays is you go to your GitHub repo in the morning and you hit the copilot and then issue five agents to go off and work. And then you keep coming back on your phone and checking, right? That's kind of like my daily habit. Now, more than my inbox is just go on the number of tokens. In fact, my wife took away my credit cards. And so it's just fun to be able to play with GitHub and just be constantly modifying. In some sense, it's sort of software on demand. Now, the thing that I wanted to have you, oh, by the way, the other thing that we're doing is really ensuring that connective tissue from GitHub primitives to VS Code or Visual Studio to all the various form factors, right? I mean, like when I look at it, people ask me, what are you using? I'm using everything. I'm using the GitHub CLI. I'm using the agent mode. I'm using the autonomous sort of agent. So it's not any one form factor. In fact, all the form factors matter, quite frankly. It's not about any one thing. And so building that all up into one sort of cohesive tool chain system is so important. And so to show you all of this, as usual, let me introduce Karan from our GitHub team. Karan, take it away. Thanks, Satya. A few weeks ago at GitHub Universe, we announced Agent HQ, which is an agentic platform that unites every agent, making it native to the GitHub flow you are already familiar with. So let me show you a demo of what that looks like. So I have a demo app here that's a storefront for selling some nostalgic toys, which are probably common in many of our households. However, you can see there's no search or filter options, and I want to add that functionality. So here is my code repo on GitHub. With the new agents panel, I can now kick off an agentic task from anywhere on GitHub by just describing the task and delegating it to the Copilot coding agent. Copilot starts working on it in the background. So I can assign a task to Copilot from anywhere I work, from GitHub.com, in VS Code, on the CLI, or even on my mobile. All right, while I was working across multiple repositories and tasks, et cetera, I have delegated many tasks to Copilot. And now I can see all of my agent sessions ongoing as well as completed ones in a single mission control view. I can also kick off an agent task right from here while also using custom agents that I have created or have been built by partners. These custom agents are domain experts which extend the Copilot coding agent across all of your tools and workflows. All right, let's dive into the agent session that we just kicked off and see what Copilot is working on. So I can see Copilot's progress here on the left and also the draft PR on the right. As a developer, I'm still in control, which means I can even steer an ongoing session midway to have Copilot work in a slightly different direction. All right, this is gonna take a few minutes, so I'll let Copilot continue working on that. Meanwhile, here is one of the agent sessions for the same feature I had kicked off earlier. So it gives me a description and, oh, nice, screenshots, thanks to the Playwright MCP. I can also take a look at all the files it has changed and also ask it to make any change further on if needed. All right, now, amidst all of this, if there is anything that I want to ask Copilot, I can jump into chat right from here. Copilot gives me a choice of various different models and various providers as well, but that's not it. I can even bring in my own models from Microsoft Foundry. So I have deployed a few models here on Foundry, including one that I have fine-tuned myself. So I have brought in all of these models to Copilot by connecting them to my Foundry deployments. So that means I can use these models directly in Copilot, either on GitHub.com or even in VS Code. So let's take a look at one of these models that I fine-tuned. Just for fun, I wanted that model to almost talk like a developer in Bangalore. So I'm going to ask it to explain me something which is very simple, what is find an internal server error, and then it says, I don't know on the response for today. So that is something which is coming from my own model that I deployed on Foundry in Copilot. All right, speaking of errors and apps, et cetera, I had also created a PR to implement a new product detail page in my app. So I can see that Copilot has also reviewed it and left a few comments. So let's take a look at some of the comments it has given, and I can see there is one over here which mentions about DB connection pool exhaustion. There is one more here which talks about type safety as well. So this is the new agentic Copilot code review which combines LLM detections, tool calling, and also deterministic scans, helping you with smarter reviews, and not just that. I can ask Copilot to implement all these suggestions for me in a single PR which will kick off an agentic task for me. I'll let Copilot continue working on that in the background. Now, we all know how important security is, right? Thankfully, I'm also a security admin for this account on GitHub. With the new integration with Microsoft Defender for Cloud and GitHub Advanced Security, I can now specifically filter for runtime risk issues in my production environment that have been identified by Defender. So you can see that. So a campaign to address all of these have been created, and oops, I see some of them in my repo as well. But not to worry because Copilot can also help me fix these security issues. So I can just select all of this and then say assign to Copilot, which will address all of these alerts again in a single PR for me. All right, while all of these different agents are working on different tasks in the background, I'll write some code myself as well. So here is the agent session for the search and filtering function that we were looking at earlier, and I wanna make some changes to it. I can continue iterating on this by opening this directly in my VS Code. So that brings up my VS Code where I can either check out the branch or even directly apply the changes for me and continue with it. So I will open up Copilot and choose agent mode along with one of my custom models from Foundry as well where I can ask it for another small change as well. So I will let that continue working locally in agent mode. All right, just like on github.com, we have a mission control view in VS Code as well where I can see all of my agent sessions running in locally agent mode, in the cloud, or even in the CLI as well. All right, so we've been talking of a few features, but there's one other feature that I really want to build, which is bundle all of these toys and then gift them as a play box. I'm sure many of you would love that as well, right? So, but I need to plan how to go about implementing this. With the new plan agent in VS Code, I can ask Copilot to create a plan to implement the feature and keep iterating on it with clarifications until I am confident. So I have already a plan that I created over here. So once I'm ready, I can start coding this or I can hand it off to an agent to help me implement this one. So in this case, what I'll do is I'll hand it off to my Copilot CLI agent, which will get to work in the background. All right, now all of these different tasks running will take some time, but once all of them would complete, this is how something my app would look like with that nice filter bar and some really nice games that we can play out here with. So in the past few minutes, we went from ideas, coded it with Copilot agents on github.com, in VS Code, even in the CLI, had that code reviewed with the agentic Copilot code review, implemented fixes for the security issues with Copilot Autofix agent, and all the while having the ability to see all of my agents and steer them in a single mission control view on GitHub as well as VS Code. And that is AgentHQ. Thank you. Back to Satya. Thanks so much, Karan. And you know, whenever you see these demos and you say, wow, I should be able to do that too. So, you know, we had a Thanksgiving weekend in the U.S. a few weeks ago, and so I had a chance to say, like, what else can one do over Thanksgiving other than build? And so I built an app of my own using all of the stuff that Karan was showing. So this is my Azure environment. I mean, by the way, this is my regular PC that I travel with, so hopefully nothing happens. And so this is, in fact, I have this app deployed, I think, in South Central Canada. This is my GitHub repo. And it's fun, right? So I kind of have my typical setup, in fact, is Windows 365, which travels with me, essentially, everywhere. And then in there, obviously, I have my GitHub. And then it's code spaces. So it's kind of like turtles all the way, right? So you have your code spaces running on GitHub in a Windows 365 instance. And then the idea that now you can go in, and my favorite thing, of course, to do is you come in in the morning and just start issuing whatever coding tasks. And so this is where I go and just, I usually fire off five, six things. It creates five, six draft branches. And then ultimately, at the end of the day, I go back and mostly delete the branches. But there is a PR or two I'll sort of accept and go on and work with it, right? So it's fun. And then the fact that you have all these models, in fact, I think now I'm mostly using a lot more Codex MAX. It's fantastic. It's fast. I'm using, obviously, Cloud Opus 4.5 as well. But the thing that I've now gotten used to is I have enough trust to just say auto. And it picks. And so if I'm just really, I want to really be efficient with my token limits. And so therefore, picking auto seems to be really, really a good way to go about it. So anyway, so great. I did all this. So what the heck did I build? So what I said is, OK, what's my dream? My dream is to figure out how to get a job in this copilot team. So I said, man, I like your deep research stuff. But I want to add a lot more to deep research. So I built my own deep research. And so with all these models available, I said, OK, what if I could start putting new decision frameworks? So one decision framework, Andrij Karpathy recently talked about this LLM council, which I love a lot. So I implemented this idea that you can now have all the models available to you. So GPT, Opus, Gemini, Kimi, K2, Grok, what have you, all of these models. And then you can select a chairman. So you have council members or the selection committee. And then you have a chair. And then you can go issue any query and have it come back and tell you what it thinks. Then another decision framework I implemented was this thing called DxO. We did this actually in health care first. So this was done. There you go. Now let's see. Let's see how all my, oh, so fantastic. So DxO is a thing that, as I said, we implemented for health care. And you have specific roles. So you have a lead researcher. And the lead researcher in this case is Opus. It sort of does the breadth-first research. Then you have another role, which is a critical reviewer. In this case, I selected Phi 1, GPT Phi 1. And their role is to find any method errors, especially bias and recency bias, what have you. Then we have a data analyst. So I picked a domain expert, Gemini. I then said data analyst. I picked Kimi K2. So this, when we published the DxO paper, it performed better than any front-frontier model. So this was in the context of very high-stakes health outcomes. And I said, hey, I want same thing for any decision I want to make. So I implemented that. So I implemented another one as well called Ensembl. So this is basically use all models and as just essentially a set of MCP servers. Anonymize their responses. So take out even who is responding with what. Give them alpha, beta, gamma. And then synthesize into one response. So these are three decision frameworks. In fact, I even extended it, by the way. I built a shopping thing. I built even a finance thing. But basically, decision frameworks. And then, of course, like a good sort of crazy South Asian cricket fanatic, what do you use it for? To select the all-time best Indian test cricket team, especially in a time like this. After what happened in the last test series, I think it's time to get to work. So what I did is I'll show you the history side. If I go in, in fact, you'll see that mostly that's what I've been using it for, which the MLB lineup was also just crazy. It was fantastic. But the test team. So let me go show you some of the stuff. So this is what happened. So this is the chairman synthesis. So this is the AI council. So it came back and it says, you know what? I figured out all of the Sunil opens. Sivag goes opens with him. Dravid makes it, obviously, and what have you. But look at this. Areas of complete consensus was Gavaskar, Sivag, Dravid, Tendulkar, Kohli, Kapil Dev, Ashwin, and Bumrah. Key debates. Oh man, VVS. Do you have him or you don't? And look at the way it sort of made the decision. 5-1 basically said, and Claude's inclusion of Lakshman was heavily weighted because of the crisis management. And as a good Hyderabadi, I love GPT 5-1 and Claude. And of course, Kumbhle versus Zaheer. This is pretty cool. You know, do you really need left arm swing or you need whatever you call Kumbhle bowling? Which is, obviously, the stats won out. 619 wickets and so they selected Kumbhle. And oh, and then captaincy debate. Kohli versus Dhoni and they selected Kohli. And so it goes off and then annotates What I love, by the way, is I implemented it as even a streaming thing. If it was not deployed in South Central Canada, I would sort of show it to you. But it's just nice to see. It's essentially like a chain of debate, not a chain of thought, right? So I can see the morals debate and then synthesize. So anyway, so that's one example. DXO is another one. This is very neat because what it does is you can see it by role, right? So I see the first, the exhaustive search of the lead researcher. It gives me, again, what are all the things. The critical reviewer will find like, okay, what are the method problems and gaps and weaknesses? So for example, error bias, right? Classic thing that happens, right? When you sort of compare across generations, you make all kinds of mistakes because you don't adjust for any of the stats and the difficulty of the wickets. I mean, walking out in a West Indies or an English wicket that was not covered. I mean, how does one play even? And anyway, so to be able to take all of that into account and then see the debate between the various models to resolve. So anyway, so I've had, and then the same thing with Ensembl as well. So the point I want to make was, I built this over maybe a couple of hours and now I'm constantly refining it. In fact, one of my hopes, as I said, is to be able to, by the way, this is all going to come to Copilot and I'm convincing my friends and they're saying, yeah, you can apply as a junior product manager if you are competent enough. So I'm still in the process of interviewing. But now that said, I think all of this will come because this is to me the next generation of metacognition, right? So if you think about these decision frameworks, you have all these agents, you're working with agents, but the metacognition is still us. And this is tools for metacognition is how I think about it. And so to me, building out these type of frameworks, building out these type of agents is interesting to select the Indian cricket team. But think about whether it's a supply chain decision, whether it's a healthcare decision, whether it is a finance decision. These type of chain of debates with multiple agents participating is going to be a lot of what we are going to all build in our systems, in our agentic systems. So I'm very, very excited about this. And then to see the kind of momentum here, right? So there's fantastic momentum. Oh, in fact, by the way, before I move to the momentum, I should tell you about agent 365. So I built all of this. I even, I said, I want to even build this as essentially a declarative agent in Copilot. So I went and built this out as a connector with a great API and did all the hard work. And I said, fantastic, now let me go deploy it. And then out comes DLP and says, nope, you can't. And that's thanks to agent 365. Because guess what? Agent 365 is the runtime that protects your tenant all the time from any malicious agent. And so the idea of agent 365 was all about taking Entra, taking Defender, taking Purview, and extending it to all agents. And so the fact that I built an agent and I wanted to deploy it in the Microsoft tenant, it said, hey, you got to be approved. First of all, you can't have, because it was deployed in my private tenant, right? It was on my own private subscription in Azure. It said, you got to first put it in the Microsoft tenant. In order to put it in the Microsoft tenant, you got to go to your admin. And so I got caught there and I couldn't proceed. And thanks to this agent 365. Because by the way, this agent 365 runtime is going to be the key, key, key organizing layer in a multi-agent and an agentic world. And so we're very, very excited about the progress. You'll have full observability, compliance, governance of agents. And so we're seeing tremendous momentum on all of this. And so I had a chance this morning to meet with many folks building. In fact, I had a chance to see the developers from Cognizant build their own multi-agent framework, deploy it in building what is a very high volume sort of application that uses that framework today inside their company. I saw Wipro use, in fact, data agents in fabric and completely change how they think about sales and finance and the entire function being transformed. I had a chance to meet with the collective arts network folks who showed me their tool chain for building essentially these animated feature films. So they're building an AI studio. In fact, Hanuman is, I believe, going to come out. They obviously built Mahabharata before. And so it's fantastic to see. They built a full pipeline with their models and their orchestrator that now is available for anyone to use. Had a chance to meet with Persistent who's done similar things like Cognizant with their own work to be able to then really deploy across the board. PhonePay showed me a fantastic application of Foundry Local. So they, first of all, implemented great search in the app. But they also built these models that are deployed on device using Foundry Local. It was fantastic to see that. And then I was thrilled to meet with the developers from Swiggy who obviously are taking all of the quick commerce to the next level with their use of fabric and real-time data in fabric and fabric agents. Let's roll the video. Swiggy is one of India's leading food delivery and quick commerce companies. We are present in over 700 cities, have hundreds of thousands of drivers and serve over 100 million customers. Our customers really care about speed selection and price. It's very difficult for me to shop for groceries because of the traffic. And I have very busy schedules to work in. I really can't be waiting for an app to deliver after 30 minutes. Instamart is our quick commerce platform where customers can order items across grocery and home essentials from over 1,000 dark stores. And all of this needs to be delivered in 10 minutes. Delivering in 10 minutes is challenging. To do this, there's a lot of data behind every decision we make. We need to understand traffic signals where our riders are and gather all of this data and analyze them in very quick time. In the past, we have to take data from different sources, transform it, work on it to make it meaningful for the business. And it used to take more than a few hours to get there. Microsoft Fabric with real-time intelligence has helped us reduce data latency from minutes to seconds. It has also helped our analytics team to get their query running from hours to within minutes. Fabric RTI has helped us track millions of orders per day across their lifecycle to provide our customers with accurate order status. It has improved demand prediction and inventory management. Now we have better visibility at our dark stores, which help us serve our customers satisfactorily. If they're showing eight minutes or 10 minutes, they're going to keep that. And that's important for me because I plan my day around it. I would rather spend some time with my people rather than running around doing it. And I love that experience. And it's fantastic to see that. In fact, they had this use of Fabric all through the dark stores as well as the central command center. And then now they're able to build even additional data agents that can act on obviously the real-time data, right? So that's sort of the type of agentic systems that are getting deployed at Swiggy and it's exciting to see. So the last piece. So if you have now the agent factory, you're building all these agents, you now need the most efficient token factory, right? At the end of the day, you need infrastructure to run all of this. And so it all comes down to this one simple yet very important equation, which is all about performance per rupee per watt. And to us, we are really driving that equation, that performance forward with sort of everything that we are building out right here in India and around the world, right? So we're building out Azure as the world's computer and we have 70 plus data center regions all over the world. In India, the footprint continues to grow. We now have Central India, West India, South India, and we have a partnership with Jio as well. And now we are very excited that in 26, we will have a new DC in South Central that will open. And then the most exciting part for me is that it's 100% sustainable. So we are not taking the eye off the ball. It's 100% renewable power. And so we are building out this infrastructure in a rapid way. And by the way, we are not just building out the infrastructure, but we're taking sovereignty and making sure that there's a portfolio of options for sovereignty. So you have the public cloud, you have sovereign controls in public cloud. So for example, you can have key management with you. You can encrypt at rest, encrypt in flight, and in use through confidential computing. So therefore you will have full sovereignty even in public cloud. On top of that, we have the sovereign private cloud. We have many local partners who are going to be able to deploy these sovereign private clouds. And as I said, we even have a Jio region of Azure. So there's an entire portfolio of options for any company to be able to make risk-based decisions so that they can use the public cloud. They can use the public cloud with sovereign services, private cloud, sovereign, Jio, and exercise judgment on what's the best way to use all of that. But one important consideration I'd also say that needs to be thought of in sovereignty is cyber resilience, right? At the end of the day, cybersecurity is a signals game. It's an intelligence game. So you can be sovereign and not have global intelligence, right? That's sort of, you know, essentially means you'll be vulnerable, in fact. So therefore, really thinking about that in the risk equation becomes very important because if you're not really seeing what the threat actors all around the world are doing to protect yourself, you will miss the signal. And so that's one of the things that I think needs to be thought about together as opposed to two different considerations. You have to think about your cyber resilience and sovereignty and the combination of those two as you think about the risk profile. And so to us, we see a lot of the threat actions across the globe. We take all those trillions of daily signals, turn that into everyday defense, right? Whether it's on Azure or on M365 or on Windows. So that to us is the core of what our responsibility is. And obviously that comes through to all our products. And with all that in mind, one of the things that we're very, very excited about is investing in India to make sure that we are able to bring the best-in-class infrastructure here. The $17.5 billion investment, I had a chance to meet the Prime Minister a couple of days ago and talk through all of what we're trying to get done in India. And he was very encouraging of this investment in particular to drive this society, this economy, and the growth. And starting with having world access to world-class infrastructure. This is the largest investment that we have made in Asia. And this is something, this is on top of even the $3 billion I announced the last time I was here. And so we are very, very excited about bringing all of this stack to India. But I'll close where I started. To me, at the end of the day, this is all about empowering every person and every organization in India to be able to achieve more. And in that context, it starts with the human capital of this country being skilled, being confident that they can do things to achieve more, drive their organization, their communities, their cities, their states forward. And that's why we are doubling down even on the skilling commitment. And we are now going to skill 20 million people across India in AI skills. And one of the things that was fantastic to hear about yesterday, I had a chance to meet with the Minister of Labor, is the E-Shrum program. And taking an unorganized labor and giving them this AI boost so that they can get credential, they can get skills, and then to be able to have employment opportunity is sort of perhaps one of the biggest benefits of this AI-driven economy of ours. And so it's fantastic to see the government take interest in taking this technology, applying it to the unorganized sector. So I'll leave you with the video. Thank you all very, very much. And thanks, and let's keep building. India has more than 400 million unorganized workers. They're earning on the day-to-day wages and not getting any benefits. We need to shift the people from the informal to the formal sector. That would help us to develop our economy. Through the initiative of E-Shrum, the National Unorganized Workers Database was developed. And we are able to provide those workers with social security benefits, also giving them access to the National Career Service Portal. NCS is a one-stop solution for job seekers, employers, training providers, and counselors. We have AI-based matching between the job seeker's skills and the employer's. So any informal worker can go to the NCS portal and match with the right employer. NCS has an AI resume builder. And if a person is lacking certain skills, the AI tool helps them to develop those skills by giving them recommendations. We were expecting millions of unorganized workers to register in the system. So we wanted to have scalability, reliability, and security. Microsoft Azure met all these qualifications. In addition to that, Azure OpenAI has become a game changer for unorganized workers. Around 150 million workers are now on the NCS portal. We are also now trying to bring in more and more schemes to benefit these workers. AI is supporting us to provide opportunities for the individuals and youth of the country.