Karl Keirstead: Let's get started. I'm Karl Keirstead. I couldn't be happier to have Microsoft here. You may know that it's very much in the public domain, but UBS is an enormous customer of Microsoft. And I think we're so enormous that when I asked Brett if you could attend, he said yes, and he brought out the big guns in Judson for us.

Judson Althoff: Thanks, Karl.

Karl Keirstead: Thank you for coming by and sharing your thoughts.

Judson Althoff: My pleasure. Good afternoon, everyone.

Karl Keirstead: Judson, maybe just given your role as the Chief Commercial Officer of Microsoft and running an enormous sales organization and being obviously in your role extraordinarily client facing, maybe I'll just start with a couple of questions on sort of the pulse of those conversations these days. I think we're all acutely aware that we went through a COVID boom and then we went through a post-COVID lull. When you're talking to customers, Judson, are you starting to sense that that sort of focus on cost cutting -- are we through that now and now the animal spirits are unleashed? How does it feel to you?

Judson Althoff: I think it's genuinely pretty balanced out there in the market. And there's a high degree of standard deviation depending on where you travel in the world. We in general see a lot of confidence and certainly taking bets on AI both on the bottom-line savings side as well as frankly topline growth. But sort of as you travel the world, you certainly see differences. I would say the U.S. consumer economy, and we obviously have just gotten through Black Friday and Cyber Monday, consumer confidence is relatively high. You travel to other parts of the world, in Europe, and you obviously still have the pressure of 2 more wars than the world needs right now, and that is definitely weighing on a lot of European customers. I think you can sort of look at the dynamics between the U.S. and China as well, and it puts a little bit of pressure on European thinking post our election. But generally speaking, customers are still seeing technology as the way out of either one of those sides of the equation, whether it's a cost-saving exercise or a top line growth, which I think puts us in a pretty strong position to be able to help them.

Karl Keirstead: Got it. And maybe just because you did mention the election, I know it's probably too soon for you to comment, but any change maybe in the demand environment since? Or maybe the way to approach this question is, given that Microsoft does have a fairly significant government exposure, would you anticipate any changes in federal government demand as a result of this new administration?

Judson Althoff: I mean I think you said it upfront. It is a little too early to tell definitively. But we're actually pretty bullish about the government's investment in technology. I think the new administration sees technology as one of the stronger U.S. market sectors, and so I would expect supporting legislation to help back more U.S. technology growth. But yeah, it's early days.

Karl Keirstead: All right, we'll ask you in 6 or 12 months again. Another venue that you had to interface with your largest clients was the big Ignite conference that was in Chicago. I was there. Our team was there. I'd love to get your perspective. When you were interfacing with clients at Ignite, what did you take away from that event in terms of hot topics? I think we know at least one of them, but how would you characterize the mood from Ignite?

Judson Althoff: Yeah. Look, it was great to have customers out in mass in Chicago. We took a little bit of a rolling of the dice to do a conference in Chicago in November, but the weather gods did smile upon us. Look, overall customer sentiment was really, really strong. A couple of things to pick up on was just sort of the pace of AI literacy that we experienced at the event. And the reason why I mention that is, you sort of attended the same event a year prior, we were doing a lot of sessions on like, hey, this is what generative AI is. Here's a large language model. Here are some demos. And you had a lot of customers kind of bobbing their heads and saying well, okay, that's kind of nice. Maybe AI will help me write my holiday cards this year, but we're not so sure it's going to take off in business.

To this year, folks dialed in on like, well, which model for which scenario for which application set? And gosh, we have 100 AI scenarios we want to get started on. To gosh, we paid an advisory firm $25 million and now we have 400 big ideas that we want to get after. And so a lot of the discussion was around the pragmatism of getting real value out of AI and honing in on those scenarios.

But we had 3 big announcements. We had hundreds of announcements, but we had 3 really big ones that I think is really good for the audience to understand. The first was Copilot coming of age as a real AI platform. Most of you I think are familiar with Copilot embedded in Microsoft 365, helping you with everything from your emails to content creation in Word, PowerPoint and Excel, to Teams meeting summaries. And there were some enhancements that we announced in that regard to bring up, dial up the quality.

Since you're all in the finance world, we announced things like Copilot helping you do Python and Excel and stuff like that. But the big piece of it was announcing Copilot Studio, which is the asset that you can use in low code or no code form to create agents, augment Copilot to create a unique AI differentiated business process. We announced that now as an Azure meter, which basically means there's zero barrier to entry to begin creating agents and Copilot extensions and really think about Copilot as the UI for AI. And Copilot standing for AI empowering human achievement. Anytime there's a human in the loop, Copilot being the platform to utilize.

The second big announcement we had was Azure AI Foundry. Which you can think of as the application server for the cloud native era, so what BEA WebLogic and IBM WebSphere, if you've followed the industry long enough or been in it as long as I have, what they did for web applications, Azure AI Foundry will do for AI native application development. And so you can think of this as one common tool chain for whatever AI model you want to use. We have over 1,800 models that we support of. Of course, all of the models from OpenAI, but from Llama to Mistral to open-source models to some of our own small models. And that one common tool chain allows you to have everything from your groundedness work, your content safety work, your data governance, and your observability. You can sit on top of all of these applications being built and say, well, am I using the right model at the right cost structure, the right kind of token yield to build these AI and modern application sets?

And then the final piece was announcing SQL coming into Fabric as the unified data platform for the AI data platform of the future, creating a really cool obfuscation layer so that no matter where data lives, if you have it in Azure or AWS or GCP, you can reason over that data using one common layer for accessing the data without actually moving it. You can think about how much this opens things up. Copilot Studio and Copilot now as a platform with zero barrier to entry. That middleware layer of Azure AI Foundry now being the app server. And the data layer allowing you to connect to data wherever it may be, really kind of unleashing more growth in AI development.

Karl Keirstead: Let's hit on a bunch of those, because you touched on several topics I'd like to go one layer deeper. Earlier you mentioned enterprises searching for the scenarios that might have high ROI. Are you sensing a pattern in the types of use cases that enterprises are pointing AI at? I don't know how you would want to describe it, but are they more internal workflow automation? Are they customer facing? Are there any common threads?

Judson Althoff: There are definitely common threads. In fact, there are 4 common categories that we guide customers on that kind of puts us actually in a far more consultative form of engagement with the client than we've ever been in before. Because of the fact that so many are coming to us with this list of things they think they should go and explore. Category 1 of course being enriching the employee experience. How can you get more productivity out of your workforce? How can you measure that? Can you extend it into the right datasets, and can you improve ROI in the employee space?

Second is around customer engagement. Whether that's your call center or your field facing salespeople, how do you redefine what customer engagement is like? And how do you think about an AI first approach towards customer engagement by leveraging Agents and Copilots and your human expertise to create a really unique engagement paradigm that's personalized?

The third being around reshaping business processes altogether. The AI First business process being one that sort of reinvents the workflows as we know them today. Because if you think about it, all business processes are some artifact of human workflow, either physically moving an object from point A to point B, or moving a piece of intellectual property from point A to point B. If you step back, pull the humans out of the equation, and say, well, how can I think about this with an AI First mindset and then add human capability where and only where I need it, we get tremendous amount of efficiency gains here.

And then the final one being bend the curve on innovation. For us, it's all about writing software with a higher degree of quality, more effectively, more efficiently. GitHub Copilot has a role there. We use GitHub Copilot to write more than 35% of the code for all the rest of the copilots. It's why we've been able to release so much product so much faster. But for other industries, there's the equivalent paradigm. Like how do you use AI for drug discovery and more effective yield and faster time to market in the manufacturing sector? How do you shorten manufacturing cycles?

Those 4 categories are great testbeds for how we work through scenarios with clients. And they tend to, back to your original question, unlock that are you driving for savings? Are you driving for top line growth? We can do both. You can have a methodical approach towards addressing the opportunities in the market.

Karl Keirstead: Maybe the one that struck me when I was at Ignite was the focus on Agents and how Copilot is becoming connected to the Agents you're deploying to automate certain functions. But everybody in this room is also listening to others throw their hat in the ring, Judson. The most prominent is probably Salesforce banging the drums on their Agentforce platform, but Google is in there, even OpenAI is talking about launching agents. What is Microsoft's pitch to enterprises? What's unique about your story to differentiate from these others that are looking to capitalize on what seems like an amazing opportunity?

Judson Althoff: For sure. I think the most important thing to take away from all of this is, first of all, it's great to actually see more folks entering into the fray here. Because, honestly, it stimulates more innovation. But the big thing to take away is this is not going to be an either/or, unlike agents versus copilots. It's the right balance of both. Like I sort of mentioned before in the previous answer, this idea of an AI First business process is agents plus copilots plus your unique, differentiated human talent that can really create a unique scenario for your business.

What's our differentiation? Our differentiation is that, first of all, anytime a human is involved, we have a great sort of embedded base of value that we convey to clients today through our modern work stack. People tend to start their day in Outlook and Teams and leverage Microsoft assets for productivity to begin with. And so if you can stimulate that productivity within that environment with copilot capabilities, then you're off to a great start. If you can then augment that with agents and connect it to data sources that may be unique to your environment and do it in a heterogeneous way, like you mentioned Salesforce, but also ServiceNow, Workday, etc., being able to use Copilot Studio to say, ah, well all my CRM data isn't in Salesforce and so I want to be able to access that as I prepare an e-mail response back to a client. And using that Copilot Studio as the really easy low code/no code way to access it and create the unique value back.

We're well positioned because of Copilot, Copilot Studio as being the low code/no code way to get after all of this. But then also for ProDevs, given the staying power we have with GitHub and GitHub Copilot, the unique relationship we have with OpenAI, Azure AI Foundry, can really help customers build their unique solutions that then sort of embrace and extend what we can do with Copilot. You kind of can come at this from both dimensions and we really don't see anybody else out in the market that can do those things.

Karl Keirstead: Got it, okay. Let's ask you a few more AI questions because it's a rich subject. I'd like to ask you a couple at the infrastructure layer and then the model layer. At the infrastructure layer, Judson, Microsoft has been very clear, Amy has on the call, that you've got enormous demand that you can see in your cloud backlog. But your ability to service that demand is somewhat constrained by the fact that even Microsoft at your scale can't build enough datacenters and lease enough capacity to service that demand. Can you get the group comfortable with your visibility into that capacity coming online? And your best guess as to when the supply demand imbalance evens out?

Judson Althoff: Sure. Well, look, I mean the good news is that we do have this demand.

Karl Keirstead: Exactly.

Judson Althoff: It's a good problem to have. The other piece of good news is that that demand has come on extraordinarily fast. As Amy has commented, we expect to see that supply and demand get matched up a bit better in our second half, sort of the next 6 months, and would expect to see some growth acceleration as a result of that. Even on our enormous base. And I would just, maybe to add a little bit more color on like how this is materializing, I think the thing that has been a bit more unique in how AI has blossomed in terms of this supply/demand scenario versus how cloud came onto our scene, was first of all just the pace at which everybody's developing and adopting. But also, where everyone is choosing to do it. And I would say in this case, the U.S. market as a whole in terms of datacenter locations has been the destination for so many of these entities.

If you think about what we've just been through with Black Friday and Cyber Monday, not only do we support the big U.S. retailers, but also the Sheins and the Temus of the world that are Chinese based entities, but really target the U.S. market. That's all U.S. demand that needs to be served. We've been able to show, in my view, a great deal of agility in being able to get all these new AI capabilities to market. At the same time, address that retail demand here for the holidays. And so we got some breathing room coming out of the holidays. We'll get some added capacity coming in the second half, and we do believe we'll see that growth acceleration just as Amy had said in her remarks.

Karl Keirstead: And at the model layer, Judson, obviously Microsoft has made an extraordinarily successful bet on Open AI. Congratulations on their success because it reflects well on you guys.

Judson Althoff: Thank you.

Karl Keirstead: But we have been listening to Satya and Jared and Mustafa use this term that LLM's are commoditizing. What does that mean to Microsoft? Does that just mean that the feature gaps between the different models are beginning to narrow? And if that's true, how does that trickle down into Microsoft's future?

Judson Althoff: Yeah. Actually, we're pretty excited about this, and it's one of the biggest reasons why we felt like it was time to launch Azure AI Foundry to give people model choice and embrace model diversity. Because we do see OpenAI's models as still being the best in the business. But there's a premium for the best in the business, and if frankly all you need to do is summarize a document or, gosh, if you're in the insurance industry, prepare a claim response for a client, you can do that a lot more cost effectively with better what we call token yield and produce the token or the character string at a lower cost than using say an OpenAI 4.0 or 4.1 mini, etc. You wouldn't want to spend like high calories on a simple scenario like that.

And what Foundry will allow you to do is basically pull up a console and say, what are my developers doing? I joke, I'm from Chicago, so my personal joke on this is it's like asking Michael Jordan to mow your lawn. It might be fun, but pretty costly. And so like, again, if you have a simple scenario, use the right model at the right cost structure for the right scenario.

If you have some really advanced, multimodal vision or voice scenario, yeah, then use the premium model. And the beauty is, you don't have to make that choice just once. You can actually make it multiple times even within the same application, so that you're getting the right balance and yield and then have this common tool chain to do all of your testing and groundedness work.

We think it's great. The OpenAI partnership has been great for us. We've learned a lot. We'd like to think that they've learned a lot from us as well. And it's been a big leap forward in innovation.

Karl Keirstead: Good. Well, as you all know, we're excited to have Sarah Friar, the CFO of OpenAI, in this chair tomorrow morning, so we're going to ask her as well about the Microsoft relationship. And I'm sure she'll be singing your praises. But let's talk a little bit about that, because that's actually, Judson, another important topic for investors, how that relationship is likely to evolve. Microsoft has sort of helped to nurture OpenAI by being its exclusive compute infrastructure provider. But as OpenAI scales and needs ever more compute, could a moment come when maybe OpenAI could look elsewhere for compute? Or how are you thinking about the durability of that exclusivity, if I can put it that way?

Judson Althoff: Yeah. I mean, look, we see a great foundation in the partnership agreement that we have with OpenAI and we would expect that to continue. I think like any good technology company, you will see OpenAI expand into new categories. They may expand into categories that compete with Microsoft, and that's okay. You have to remember that we're pretty good at managing a competitive or cooptative ecosystem out there. One of our largest customers on the planet is SAP. I have a multi-billion dollar Dynamics business that's growing at a healthy pace and yet I enjoy leading market share for Azure underneath SAP's platform. You could expect that same skill and partnering effectively and nimbly to play out with OpenAI as well. And again, it's healthy for the market, it's healthy for innovation.

Karl Keirstead: Okay. Obviously Azure AI services are not the only component of the $10 billion run rate AI revenues that Satya called out on the last call. M365 Copilot, GitHub Copilot are critical. Let's talk about that second piece on Copilot. Can you give us an update, Judson, on how you're feeling about adoption? And I think what people would love to hear is, is there a catalyst that might serve to trigger ever faster adoption with enterprises? For instance, is there a potential model upgrade that might make Copilot a little bit smarter and the value prop improve? Are there integration points that would help organizations that are currently undecided get over the hump? What are those catalysts? And by the way, I should say at the onset, many of you know that UBS signed a 50,000-seat Copilot deal, so apparently we got over the hump.

Judson Althoff: Thank you for that. Look, I think there's 3 things to keep in mind on these Copilot implementations that are really the keys to success. And that's immersion, extendibility and ROI. In every single successful deployment we've seen with Copilot, those 3 things have been well in check. On the immersion front, let's face it, we don't all wake up every day as human beings thinking about, well, what could I layoff on my AI assistant this morning while I'm having my coffee? And so you have to get people used to prompting and having the AI assistant in their daily lives and scenarios. We have a set of services that we've launched now that really help people get it going in their daily lives more effectively and in the environment that's specific to the customer in question.

On the extendibility front, whilst Copilot out-of-the-box reasons over your entire Microsoft estate, after a couple of days, everybody wants it to reason over something else, some other unique dataset that they had. And that's where Copilot Studio really, really shines. Like I said, we're leasing that as a meter now, and so like you just pay for what you use. And if you have 1 user community that does 5 prompts a week, you're not going to pay much for that. If you have your superusers, well, the solution will scale.

And on the ROI front, we released a new module of Copilot we call Copilot Analytics. This is all sort of predicated on some of our power user customers where we've seen the real need to map the utilization of Copilot back to OKRs or KPIs that they may use. I'll give you an example. We've rolled it out across our environment in my organization at Microsoft. You all might say, well please, dude, that's obvious, of course you would. And I would say, yes, but it wasn't free because of the cost and scarcity of GPUs. We had to make the investment and decide it was also going to be an allocation of GPUs that we wouldn't allocate to customers. Because you can sell as much of it as you'd like out there in the market. We rolled it out to my 65,000-person organization. We did the immersion work and like even had the competitions like who can create the best prompts? We have a daily mail that goes out called Copilot Yoga, so people kind of stretch their way into using AI every day.

But the magic came with the extendibility in the ROI. We connected back to our CRM environment and also used Copilot Analytics or what was the precursor to the GA that we announced at Ignite to measure the utilization against my actual sales results. If you take a quartile of users of M365 Copilot even inside of my organization, against their own prior baseline of performance, so take top performers versus top performing data, high utilizing Copilot. That cohort now generates about 10% more pipeline, has more than 20% faster close rates, and I'm driving a little over 9% more revenue per head out of that organization. And so, it pops right away when you see it in those terms because as the business leader, look, you're still measured by the same level of accountability. And if you can show the yield, it's pretty fantastic. It's a big reason why a material amount of my current pipeline in this quarter for Copilot are actually repeat buyers. People coming back for the second tranche.

Karl Keirstead: Upsizing.

Judson Althoff: And it is, let's be super clear, it is the fastest growing Microsoft 365 cloud service we have ever had, Copilot, and so we're excited about it. We think Studio as a meter, lowering the barrier to entry in that regard is going to accelerate utilization and we think this map back to ROI and the way that customers measure it most effectively in their organizations, also going to be a pretty big unlock.

Karl Keirstead: Okay. Judson, in our last few minutes, let's talk about some other areas that might not be so directly AI focused, but are big parts of the Microsoft suite that would be nice to hear your views on. In the core Office 365 or M365 seat growth, I think in this last quarter we saw some slowdown in seats. I'm assuming that that's just scale of the business maturity, maybe a little bit of macro. But then the offset to that that enabled Microsoft to still put up terrific growth in that area was that you're still executing really well on the M365 bundles, the SKU shifts up to E5. Can you offer your thoughts a little bit on the seat trends, but then how much room is left on the bundling and SKU shift?

Judson Althoff: Yep, yep. A couple of things in terms of the upsell there. There's still a lot of headroom in terms of the penetration rates we have. I mean, we're fairly penetrated on the high end of the enterprise, but in what we call our major segment and our corporate segments underneath, there's still a lot of headroom there. There's a couple of big drivers. Most of our security stack hangs off of our modern workstack and security continues to be a very, very healthy business for us. That accrues to both an upsell to the suite, so for those who are less literate of Microsoft SKUs, there's an ME3 SKU, which is a base SKU, but the top of the line SKU is the ME5 SKU. And if you land enough of the identity protection, endpoint protection and security, it makes sense for people to graduate into that ME5 SKU. We see continued health there. But also, discrete security bundles for the mid-market where they may not buy off on the larger more enterprise prone SKU. A lot of upsell on security.

The other thing that's growing remarkably well for us in this space is Windows 365, the virtual desktop. And there's a lot of headroom for growth for us there. And so those 2 things plus Copilot, adding on that ARPU growth on the per seat basis is really helping. And I think you pegged the seat volatility. There's some macro in all of that, but we're still quite bullish in seat expansion opportunities and frankly the differentiation now that we have with Copilot, Windows 365, E5 makes it even into places where we have some competitive entrenchment to go back and take more share and increase seat growth.

Karl Keirstead: Got it. Judson, I want to finish by letting you talk a little bit about what you described as your 65,000 individual sales teams and maybe offer some insights into any changes afoot. How are you restructuring? Little tweaks at this point I'm sure. Anything interesting to call out that might be indicative of changes in the market that you're appropriately reacting to?

Judson Althoff: Yeah. Look, I would say first of all, we're drinking our own champagne on AI transformation.

Karl Keirstead: Yeah, you gave some amazing success milestones.

Judson Althoff: Yeah, absolutely. There's that piece. Also, in our call center we've rolled out all of these capabilities. We'll save hundreds of millions of dollars in the call center. But more importantly, customer sat is up, employee sat is up, case resolution is improved by more than 12%. It's alive and well in all 4 of these categories, I hired a Chief AI Transformation Officer to use Copilot Studio to look across all of my business processes and engineer out overhead where we can. Of course, we use GitHub Copilot in all of our engagements, and that's helping us deliver faster on the back end, so those 4 pillars are alive and well. But effectively, I have 5 key priorities for my sales team.

Karl Keirstead: Yep, what are they?

Judson Althoff: That will be consistent, frankly. We're going to put a Copilot on every device and across every role. And we talked a lot about that and how. Two, we want to secure these really unique and differentiated AI design wins with customers and that will go across all 3 of those vectors from Copilot to Foundry to Fabric. And we've amassed hundreds of these already and the way we built Foundry was to look at the patterns and the services that were being used by most of those customers and packaged it in a way now that it's more repeatable and scalable. We're going to run hard at that.

Third is cyber. We do genuinely believe that while AI and digital capabilities will harden customers moving forward, it does introduce vulnerabilities that make cyber strategies an imperative. We want to secure the cyber foundation of every customer we have.

Fourth, you picked up on it, M365 is still growing outside of all of this AI stuff. Pull it out and we still have headroom for growth, and so we're focused on our core execution there. We made some changes to even our partner incentives. Our partners were out at Ignite a couple weeks ago as well and we conveyed where we see them growing more in those middle market segments.

And then finally, 5 is a little bit like the real estate rule. Instead of location, location, location, for us it's migration, migration, migration. Because there's still an awful lot of long tail cloud growth out there and we're still really only in the middle innings of cloud. You know, everybody, sort of the hype cyclists switched to AI, but let's not forget that there is an awful lot of money yet to be had in migrating SQL Server, Windows Server, VMware.

Karl Keirstead: Any early signs of that picking up a little bit, Judson?

Judson Althoff: Picking up a bit. I think VMware did the world a favor, Broadcom with some of the punitive licensing strategies they've had, and it's got a lot of customers wanting to lift those environments forward.

Karl Keirstead: Now also the Oracle on Azure, a lot of SAP customers moving over.

Judson Althoff: Yeah, we still see mainframe migration opportunities as well. There's a lot of headroom on the migration front. So those are our 5 priorities. I expect it to stay that way, frankly, for the remainder of this fiscal year and into the next as well. We're pretty excited about the opportunity.

Karl Keirstead: Well, Judson, that's all the time we have. Congrats on an amazing last year for Microsoft. Thanks for being such a great tech partner for UBS.

Judson Althoff: Hey, thank you for the partnership. I appreciate it. Thanks, Karl.

Karl Keirstead: Okay, my pleasure.