Frontier Transformation | Judson Althoff | Microsoft | USA House

Yeah, it's not the most self-explanatory title, is it? So broadly speaking, you can think of Microsoft being in two businesses. One that we sell to consumers and the other where we sell to businesses around the world. So I'm responsible for our product strategy, marketing, sales, service support for everything that we deliver to businesses around the world, large and small, and governments as well.

Sure. It's been an exciting couple of years. I would say you rewind back three years ago when we first started demoing generative AI in our products. And most of the world thought it was a series of parlor tricks. You know, hey gosh, maybe this can generate my holiday cards, but it will never really amount to anything in business. To, you know, fast forward ahead and everybody thought the AI robots were going to take over the world. And, you know, there's a lot of fear and concern over gosh, Microsoft, how can you help? To today, clients show up at our briefing centers around the world with 400 big ideas of all of the things that they think they can get done with AI. If they've talked to an advisory firm, they have 900 big ideas of all of the things they think they can get done with AI. And you know, frankly, not all of them make a ton of sense. And so we spend a lot of our time helping customers navigate the applied use of AI across the enrichment of the employee experience, customer engagement, the reshaping of business processes and bending the curve on innovation. So deeply rooted in industry work, government work and supporting our clients around the world.

That's right. So I think it's maybe worth breaking that down for everyone because I think a lot of folks are attuned to how AI can be used to write software and we of course do that across our portfolio products. AI is responsible for roughly 35% of all of the products we produce today, how they're documented, how they're tested. But as you state, it goes well beyond the production of software and into the applied use in science and discovery. So all of our work in quantum computing for example, most of that research has been AI-forward research in terms of the molecular discovery, breaking through on new states of matter, all of that has been AI-driven in its pursuit and that same concept takes itself broadly across other industries we serve from healthcare and life sciences to agriculture to industrial design. The applied use of AI connected to human ambition to unlock all kinds of growth is where we've been focused.

I think the biggest breakthrough has been around the work with agents. We spent the first couple years of our journey getting AI assistants into the hands of so many of our customers around the world. Our Copilot product is used very broadly now. But Copilot in and of itself is meant to just be a gateway into broader AI solutions. You can think of it as the assistant that can help you in your daily tasks and what you get done as an individual. The real unlock is connecting agents into the business processes that matter most. Whether that's, you know, running a public institution or running a large business. AI can serve to reshape a process from left to right. If you have agents doing the work first and then involving human expertise in a guided capacity, it can really help dramatically improve the production function of any business both in terms of efficiency as well as net yield and product design. And so for us connecting agents back into the Copilot experience so that you know humans can be at the center of driving that production function with all kinds of new yield is where we've seen a tremendous amount of development this year. So it's kind of been the year of agents. And then the other thing that has really been a big part of our work is the trust foundation for AI. Everyone wants to know the ROI associated with AI and everyone wants to understand broadly what AI is doing in their environment. Is it generating positive outcomes or is there a bunch of random acts of innovation happening inside of an organization? So earlier, well I guess sort of towards the end of last calendar year in November we launched a product called Agent 365 which is designed to provide an observability platform for everyone to understand how AI is being used in your environment whether it's being built on the Microsoft platform or any other platform. You can observe how AI is doing work inside your environment. You can register it and give it a namespace the same way you would to an employee. But more importantly you can provide governance and controls and security posture across any AI asset in your environment so that you can make sure that it is providing the outcomes that you expect and not frankly straying the course of how you've designed it to be implemented.

Yeah, exactly. I mean, one of the fun things we did this past year was we actually created an AI agent of me. So there's a thing called Agent J running around Microsoft.

Yeah, but it's kind of creepy to talk at something that has your own voice talking back at you. But the experiment was quite a positive one both in terms of business impact as well as just the cultural shift and getting people to be immersed in using AI because we basically trained this thing on a decade's worth of information about me. Every email I've sent, every Teams conversation I've had, every business coaching session I've given, every presentation I've given publicly or internal to the company, we trained it on my voice. We had my family and people closest to me poke at it to try to break it and ask it all kinds of questions it shouldn't answer. And then we released it out inside the company. And so one, it serves to train all of my people. It democratizes the experience so that even a new hire who wants to ask a question of the CEO can be given direct information about the strategy of the company, the latest products that we're taking to market, receive coaching sessions on how to address concerns with their customers. And it also I think culturally was a leap forward to us because I think for those who are afraid of AI taking over their jobs or somehow taking their identity and what is most important to them, for me as a leader to basically say, 'Hey, listen, I'm going to be vulnerable here and put this thing out in the wild.' And I'm not at all concerned about the things that it's going to do on my behalf because it actually frees me up to do a lot more inside of the company.

That's right. When we put a business success framework in place around measuring every AI artifact that we deliver inside of Microsoft and how it impacts our business. We measure the enrichment of the employee experience and the KPIs that they look at. So for example in our salesforce we look at the use of Copilot and the agents that are infused into the selling environment and we can tell that the top decile of users have 10% more pipeline, 23% faster close rates, 9 million more revenue per head. So there's quantifiable business metrics that we put in each dimension of the business. On the customer engagement side of things we've used AI in our call center environment to revolutionize how we engage with our customers. Not only have we taken $750 million of cost out of that environment, customer satisfaction is up, employee satisfaction is up because we're solving problems faster and more effectively for our customers. We've reshaped business processes inside of the company. We're a 50-year-old company, which doesn't sound that old, but in the world of tech, that basically makes us ancient. So I have contracts with customers for products that don't exist anymore and all kinds of arcane legacy business processes that we're using AI to now reinvent so that we can take people off those bodies of work, obsolete the mundane and get them focused on forward innovation with our customers. And then on the innovation side, again as I said earlier, we use AI every day in the coding work that we do, in the trials that we do to make sure that the products that we're producing in market are going to meet the expectations of our customers. We use it in basic research and science. And we use it in terms of how we collaborate as well and bring people together across all of our different lines of business. So it's pervasive in how we run the company.

Yeah, it's a good question. So, we've spent the bulk of this last year doing R&D around intelligence and trust. And it's an interesting thing, right? By the way, because if I asked each of you to put in an envelope what you think, you know, really empowers AI today, what has really revolutionized how AI is being used, you'd get a different set of answers from a lot of folks. But I would be willing to bet that the two common things everybody would talk about would be models and silicon. Who has the best AI model, who has the best silicon. That's what's really driving AI. And I would counter that with to me the two most important things enabling artificial intelligence today are intelligence and trust. And you have to be wickedly focused on governing both of those inside of your organization. So the intelligence piece is not inherently built into a model. It's the data inside of your company or your organization that makes it unique or differentiated in its purpose. So we spent a lot of time building an IQ platform so that customers can harness their IQ, protect their IQ and actually use it as a differentiating capability inside their organization and then provide the platform and tools that allow a trusted visualization layer of how all of it comes to reality and produces real bodies of work. Because if you're a believer like we are that AI has to empower human ambition and it's about the outcomes that businesses need to drive and the differentiation that humanity can direct then governing how you manage that intelligence layer with whom you work, how you work, how the content over which you collaborate, your company or your organization's most valuable business flows. That's what we're really focused on in terms of making sure that organizations can harness AI for good.

Yeah, indeed, indeed. So the journey we're on has us helping customers harness their knowledge. It's about our IQ platform enabling our customers' IQ and enabling them to protect and harness that. There's a lot of discussions about AI sovereignty and what that means in terms of interoperating on the global scale. But each organization needs to protect its own digital sovereignty. Because absent of that, if you break it down, what a large language model can do is connect to all of your data through thousands of tiny straws, hoover all of that data up, inference over it, and become smarter itself. What we're guiding our customers to do is deeply govern and protect their own IQ, their own knowledge repositories to empower their most important business flows. That way you know as model diversity grows and the best model for the best scenario at the best time can always be deployed against those artifacts and you can achieve your outcomes and maintain your own ambition.

That's right. Look, at the end of the day, right now the demand is so high to make sure that AI is producing real valuable outcomes for customers across different industries. And the demands of data center buildouts are perhaps the single biggest constraint that we have. And so finding the right energy sources, the right critical infrastructure to support all of that, working closely with governments to unlock all of that, is hugely important to our future.

But simply speaking, physical AI is basically taking AI, large language models, the way in which you think about interacting with a text-based chat environment and infusing that into the physical sense and connecting it to the world of robotics. So there's a ton of things that we can do with physical AI. Everything from revolutionizing the manufacturing sector, you know, so as you talk about wanting to reinvigorate manufacturing, digital factories, there's huge ramifications for physical AI. One of the biggest challenges we've had in helping customers build out digital factories is the gap in the knowledge divide because you can build the most intelligent systems in an environment, but you still have to connect them back to the front lines and that gap and people understanding how to interoperate with an advanced machinery or an advanced cloud simulation model is super hard. Now basically you can talk to these environments using natural language, you know, so the shift operator can come in and say hey what were the anomalies last night, what happened, you know, how should I think about that in terms of areas that I need to inspect in the manufacturing environment. So there's deep ramifications in how we think about industry but there's also areas where if you think about discovery being limited by human safety factors there's places where we would not want to expose humans for the factors of safety where you can put robots into those environments and if they're guided, directed by AI, you can advance a lot of work. So, it's a huge topic. We won't do it justice in the 30 seconds here, but there's a pretty big lock in connecting these smart natural language systems back into the physical realm.

No, look, I think there's a real leadership imperative here though because there's really increasingly very few things that technology cannot do. But there's a whole list of things that technology should not do. And from that comes a prioritization of a real leadership imperative to make sure that we're putting AI to work in the areas that matter most, that matter most for business, that matter most for society. There's still huge gaps in healthcare, huge gaps in education. There's all kinds of things that we can do to democratize that experience using AI. And so it is much of a business and government leadership imperative as it is a technology imperative. But I think in the future, the more we do what is good for the people, the better it's going to be for business, including accessibility.