Discover a world where technology is the key to a better future. Where progress brings enlightenment and unlocks new opportunities just like letterpress once did. Where human ingenuity and technological advancement helps us to reveal the workings of the body, the mind and the universe, empowering humans to heal and economies to flourish. Let us not just do what is possible but what is right. By using resources to create and to shape a planet in balance in a world in which everything is possible. We make consciousness a given. Empowering growth and creating experiences that unite is what drives us. Welcome to the Microsoft AI Tour 2026 in Helsinki. The future is in our hands.
Please welcome country general manager Microsoft Teimu Vidgrren.
So, just a second. Just a second. Let's switch to English. So next up, it's my massive pleasure and honor to welcome to Helsinki and Microsoft AI Tour stage the CEO of Microsoft Global Business, Judson Althoff.
Thank you my friend. Thank you.
Well, thank you. Thank you.
Well, good morning. Good morning and welcome to the Microsoft AI Tour here in Helsinki. I'm thrilled to be here with each and every one of you today because it's my first time in Helsinki and there's not too many places in the world where I get to speak to our customers and partners where I can say it's a first for me. So, I'm very happy to be here. It was awesome this morning taking a brisk jog out by the waterfront and seeing the glory of the city and the sun rising so early which is nice compared to what we experience even in Seattle. So fantastic to be here with each and every one of you. I have a job to do here in front of you. I want to do three things for you this morning. One, I want to talk to you about a topic I call frontier transformation, which is all about demanding more from AI, making sure AI is contributing positively to society and unlocking growth. The second is we're going to show you all the latest capabilities that we have built into the Microsoft portfolio to help enable your success on that journey of becoming Frontier. And then third, enable you to hear from some of your peers and how they're navigating all of the transformation opportunities today, which is my favorite part of the show. So, but before I begin, one of the things I do before I go to each city, and we're doing 40 cities this year, will touch over 100,000 customers in venues large and small. One of the things I do before I come to each city is I use Copilot to run a bit of a prompt to help myself get indoctrinated into how things are going here locally. And so my prompt is hey I'm going to give a presentation tomorrow in Helsinki. Tell me how people in Finland are thinking about AI. You know what do they see in terms of opportunity? What are the concerns? And you can see the prompt here and some of the responses. And here in Helsinki, it's fascinating to see that, you know, while the majority of people are very excited about the opportunity around AI, there is a mix. There's a balance. 60% say, hey, we're very excited. We see opportunity. 40% cautiously optimistic, more concerned about how AI is going to impact society and making sure that the appropriate regulatory measures are in place to get the most out of AI whilst at the same time doing so in a trusted way and delivering on the positive promises. So if you look at the potential for impact, there's no question. It's not a question of if, it's a question of how. Making sure that you have the appropriate guardrails in place to get the best out of your AI capabilities whilst at the same time adhering to governance, manageability, cyber security, and frankly the ROI of AI for your business across all industries. So at Microsoft, we take this quite seriously. In fact, if you think about the first wave of AI solutions that have come out over the last 3 years, much of it was around productivity. In fact, I don't think our industry has done too much of the world a great favor by focusing on productivity and efficiency and leading to some of the concerns about the impact on jobs and career obsolescence when in fact if you actually push AI to do more for humanity you can unlock growth, creativity, democratize intelligence. You think about the gaps in society that have existed throughout time. Those who have had access to knowledge, the opportunity to learn, harness their potential, have outpaced those that have not had that access. So, putting AI to work for the world's best educator, the cure for the incurable, unlocking creativity and growth, the world's best film creator, the world's best musician is out there. And that potential can be unlocked through AI. So we have focused a lot of our attention over this last year on building a platform to enable this potential so that we can demand more from AI as a society. But like all businesses, you need a framework, a framework for success. And we've put forth this four-pillar framework at Microsoft. In fact we use it internally. I'm the proud owner of it inside of Microsoft so I use this framework for success to measure and think about the outcomes we can drive using AI. And again, it's been fascinating. The pace of adoption has been unlike any other wave of technology we have ever seen. Three years ago, everybody thought, wow, okay, we can kind of see this as being some cute parlor tricks. You know, maybe it can help us write our holiday cards, but I don't think anybody actually took it serious in the context of business. Then as time went on, fear, uncertainty, and doubt set in, and folks would come to Microsoft headquarters saying, hey, Judson, we think the AI robots are coming to kill us all. What can you do to help? And now today, people have a thousand big ideas of all of the things they think they can get done with AI. If they've paid an advisory firm $25 million, they have 4,000 big ideas of all of the things they think they can get done with AI. And about 98% of them aren't really very good ideas because they're grounded first in technology versus business outcomes. And we believe AI transformation, frontier transformation has to be grounded in business outcomes, empowered by technology and in that order. So this four-pillar framework helps us stay grounded. You have to ask yourself, is the AI solution in question going to enrich the experience of our employees? Are we going to be able to attract the best talent to come to our company once they're here? Can we provide them with the best tools to nurture their potential to help them realize their career opportunities and goals more effectively with the AI that we've put in place? We measure this inside of Microsoft in our own AI transformation. I've rolled out Copilot to all of my people. And you might say, well gosh, that's obvious. You make the product, but actually the opportunity cost is quite high because of the cost of the infrastructure and everything we have to do to prioritize these days. Everything I implement internally is just one less thing I can implement externally. So the payoff must be there. So weekly I track the utilization and the enrichment of the employee experience and measure it back. And I can tell you that within our sales force, for example, the folks that are the top users of AI are 10% better at producing pipeline, have 23% faster close rates, and generate 9 million more revenue per head. So the business results and the outcomes and the nurturing of the environment is critical. When it comes to customer engagement, you also have to think about how you're going to personalize the relationship you have with your customers. And Microsoft has tens of thousands of enterprise customers where we have great one-on-one relationships, but we also have millions of SMB customers. I would love it if someone with a Microsoft badge could talk to each and every one of those customers, but it's just simply not practical. We can use AI to drive hyperpersonalization and better engagement, customer self-service. And as a result, our NPS scores have gone up by over 12 points. Our customer self-serve is up 35% year-over-year. And so the realization of growth through better customer engagement powered by AI is there. The third pillar is about reshaping business processes. And we all have legacy processes that need to evolve. Microsoft recently celebrated its 50th anniversary, which is not that old. Many of you work for companies that are older, but in technology, 50 years makes us ancient. So, we have to remember the fact that when Microsoft first started distributing its products, we printed software onto a CD and put it into a box and handed it to distributors and had them send it around the world. If you stare closely at my order management system today, you can see artifacts of that past. Using AI to reshape those business processes so that people can spend less time on the overhead of operations and more time working with our customers, more time innovating new products. We've committed to our board that we'll continue to grow revenue at the same rates for the next three years with the same headcount. So, it's not about reducing our workforce. It's about shifting to the left and having better engagement with our customers and shifting to the right and building better products for our customers. Which brings me to the fourth pillar, bending the curve on innovation. We've all seen what AI can do in terms of writing software. And in fact, at Microsoft, on average, AI writes 35% of the code that we have today. And that spans across our entire portfolio. Some of our newer products like the ones you'll see today, Copilot, Co-work, more than 80% of the codebase for that product was written using AI. So we can see how it helps you get better product, more product to market faster than ever before. But it goes so far beyond coding into basic science and discovery. Our quantum lab teams that are staffed with very intelligent PhD researchers that are the last folks in the world that would admit that AI could help them do something better than they could do it themselves. They will tell you that we're 5 years ahead on our quantum delivery. We're now bringing chipsets to market. We never would have been able to do that without using AI and the discovery process, new states of matter, new materials. And it's not just in technology, it's in other industries as well. We see it in healthcare delivering on the promise of AI drug discovery, shortening the cycles to bring new medications to market, better cures, lower costs, better outcomes for society. So this four-pillar success framework when used properly can help drive frontier transformation, true growth, expecting more from AI. In addition, you need to have the right approach inside your company. These three traits we think represent high potential for the frontier firm. First, you have to put AI in the flow of human ambition. We can't spend our lives cutting and pasting into prompts and getting data back and then figuring out what to do with it. It has to show up in the tools that people use each and every day. And you'll see this throughout the demonstrations that we're going to have for you this morning. The second is you have to unlock ubiquitous innovation. There's a maker in every one of us. And more often than not, the people on the front lines closer to the business are actually in the best position to solve these problems versus having to rely on professional IT or development to build products for their own use. So putting AI development tools into the hands of the citizen developer and enabling that business potential is super critical. But if you do that, you also then have to have this third trait which is observability at every layer of the stack. What is AI doing in your environment? Is it helping you to produce the outcomes that you expect? Is it operating in a well-governed, well-managed and secure capacity? And are you getting the ROI that you demand? So, we're going to show you how you enable these three traits through the power of a platform. So when I travel to all of these locations around the world and have great privilege to talk to so many customers and so many industries, I often ask, what do you think the most important aspects of an AI solution are? And if I were to have asked you to do a little bit of a survey, I'm sure we'd get a lot of different answers. But two that most often come up are, gosh, Judson, isn't it about models? Like every day, every week, we hear about a new model dropping and new capabilities with the model. So models must be one of the most important things in these solutions. Then some of you might say, well, it's silicon, right? You need GPUs to make AI come to life. So it must be models and silicon must be the two most important things in an AI solution. This morning, I want to debate that with you. I think the two most important things in any AI solution are intelligence and trust. You have to have a platform that harnesses your intelligence as an individual, as an organization, as a company operating in an industry. And AI needs to amplify that intelligence, not take it over, amplify it, help you reach your potential while you continue to build out the platform of intelligence that differentiates your business. The second thing is you have to have a trust platform that enables this governance, manageability, security, ROI, the things I've been talking about as an aspect of frontier transformation. We believe in this at Microsoft. So much so that we have pivoted our entire portfolio of capabilities around intelligence and trust because these are the two most important things. I mean, case in point, 6 months ago, all anyone was talking about was the latest GPT-5 models. Then a month later, it was Gemini 3. Then a month after that and even through today it's Anthropic. Next month it'll be something else. So do you want to change out your entire information work platform, your entire development platform, your security and operations platform, your supply chain platform, your finance platform, your system of record every time a new model drops? No, you want to build a platform for intelligence and a platform for trust and leverage the latest and greatest AI to amplify the same. So in November at our Ignite conference, we announced over 70 products, but all really anchored around five key products that enable intelligence and trust. The first three all relate to what we call the Microsoft IQ platform. It's a platform that enables your IQ to remain front and center as the top priority in your AI strategy and your top priority in your frontier transformation strategy. It's far different than simply using models to connect to data. I'll give you some examples. Work IQ is the brain inside of Microsoft 365 Copilot. It knows how you work, with whom you work, the content over which you collaborate, your most important business flows and it knows this definitively not through inferencing or guessing. The second thing is Fabric IQ which provides the same kind of semantic context over your data. Now what we've done here is we've taken the brain out of Power BI which was already a cross-cloud, cross-data service analytics solution. So if you have data in Google BigQuery or Amazon S3 data stores or Azure data services or your own on-premises data environment you can use Fabric IQ to understand all of your data and enable models to reason over it to produce better outcomes. The third is Foundry IQ which is the middle tier. Think of it as the application server layer that allows you to harness your own knowledge bases and nurture your IQ. Again, your differentiation. Put it all together in a scenario like supply chain management. So, if you're going to build an inventory agent, an out-of-stock agent, an availability to promise agent, you're going to need to reason over the system of record. Lots of data sources. Fabric IQ shines there. You're also going to need to then look at external data, data from third-party logistic companies, weather patterns, tariff scenarios, and there you can use Foundry IQ to harness all of your data. And then finally, you're going to want to work across your distributed workforce and their Work IQ can help you understand how to get the right information to the right employee at the right time to take the right action. So the IQ platform enables model diversity, openness and heterogeneity to really bolster your IQ, keeping it front and center in differentiation. The next major announcement we made was around Agent 365. It's an observability platform, a trust platform that enables you to see every agent operating in your environment, whether it was built on the Microsoft platform or anyone else's. An open registry for all of your agents, connections to your security environment, and a mapping visualization that allows you to see how agents are coming to life in your business. When we turned this on inside of Microsoft before we announced in November, the preview, we had over 150,000 agents being used by 80,000 employees on a weekly basis, which is materially greater number than we ever thought we had. And we're a reasonably competent technology company. So this agent sprawl is a very real thing. Getting your arms around it, being able to govern it, that's what's behind Agent 365. And then finally, we announced Agent Factory, which is a way for you to build agents across line of business using every asset that Microsoft has in one single meter so that you can build consumption patterns that map back to the real work being done and pay for value. But we didn't stop there. The innovation continues. In March, we announced wave three of Copilot, bringing agentic capabilities into Copilot that I'll show you in just a moment. We also announced the preview of Co-work, Agent 365 becoming generally available earlier than we had planned because of the adoption. We already have 30 million agents out there worldwide in the agent registry in Agent 365. And then we announced something called Microsoft 365 E7, the frontier suite that has all of these Copilot assets and Agent 365 built into the solution so that you can deliver on the promises of AI transformation. Now, we've brought all of these scenarios to life today for you in a scenario through a company we call Zava. And I want to play a video for you to introduce conceptually everything that we will show to you for the rest of the session this morning. So please roll the video.
Zava hits a new high. Leading innovator Zava. Zava Slide. Startup Zava. Zava. Zava. Zava does it again. What a journey from scrappy startup to global game changers. What if clothing could think? Weaving together technology and comfort. Up and coming startup Zava is launching its first smart textile, Thermore. Designed to turn heat into data and fabric into intelligence. So like, why strap a sensor on your wrist when you can weave it into what you wear every day. Thermicore was just the beginning. Introducing Zavacore. It doesn't just stretch and breathe, it thinks. And we're bringing it to the world's best athletes. Zava goes public today. The intelligent textile market continues to heat up and leading innovator Zava is upping the game. Zava hits a new high. From professional locker rooms to the runway, Zava blurs the line between fitness, fashion, and innovation. You've seen the pros rave about their Zavacore jerseys. Now Zava is bringing it to any athlete, including your local little league team. Their AI-first approach is a huge reason for their meteoric rise. And now we're bringing Zavacore technology to the pitch with a brand new line of intelligent soccer cleats, allowing both players and coaches to track stats and performance direct from the match. So tell us, what's the secret to your success? 18 months ago, we made a conscious decision to really weave AI into the fabric of Zava as a company. All in our pursuit of becoming Frontier.
Okay, so before you all pull out your mobile phones and try to buy stock in Zava, we made it up. It's not real, but we made it up based on the inspiration that we see from so many companies around the world. And many of the scenarios in the Zava story are coming to life today in customers that we work with in so many different industries. And so we're going to unpack how you build this kind of innovation in the spirit of becoming Frontier. And we're going to start in the marketing department and I have Ann Krepky here with me. Ann, take it away.
Thanks Judson. So as the marketing lead for Zavacore, it's my job to help our engineering team get our latest innovations to the market. And we had a big update this morning. We made a breakthrough in our Zavacore technology and we've got to take it and get it ready to go. I'm super excited. I'm going to show you today how we're using a brand new feature called Copilot Co-work to help me navigate this launch process. Right inside the Copilot app, I'll click on Co-work. And you can see we've got a very familiar interface. We have our prompt bar here at the top. Underneath we have some suggested prompts to get me started. And underneath that, we can see all the different tasks that I have running in parallel, which is just one of the amazing things that Co-work allows me to do. So, let's put in our prompt. I want Co-work to do three things for me. I need to schedule a 30-minute meeting with Jason to finalize the value prop for Zavacore v2. I want to pull all the launch details into a Word document and I need to start building a PowerPoint so we can go pitch this launch to our exec team.
Super. So Copilot Co-work basically allows you to multitask using agents working on your behalf.
Exactly. Yes. And full transparency, this did take about 20 minutes live, so we sped it up to show you all the goodness of what Co-work can support with. Let's kick it off. So Co-work gets started working just like you said. It's multitasking on these different pieces I asked it to do. And it's using Work IQ to look at my work context, pull all of the information surrounding this launch and get going for me. What I love in particular about Co-work is if it doesn't know, it checks in with me. You can see it found multiple Jason Gregories at Zava and it did recommend the correct one based on the fact that we've worked on Zavacore together. Great example of Work IQ. I'll confirm that it's got the proper understanding and it'll kick it off. Co-work takes all those different requests I asked and break them into individual tasks which I can monitor progress on in this nice progress tracker. And what that means is it can work on each task in parallel and give it to me as soon as it's done. For example, we see it found 30 minutes with Jason and I can edit the meeting invite right from the prompt bar and when I'm ready I click create and this goes and creates the Outlook invite. Now, that's a great interaction example with Co-work. And it's not just responding to it following up with me, but I can actually also ask Co-work to work on additional things for me. You know, we really need to get finance in the loop to work on the business case. 100%. So, I'm going to ask Co-work to send an email to my finance colleague Nerja and get her up to speed so she can start building the business case. And like you said, we're not stopping Co-work in its tracks. It's taking that additional request and adding it to the list of tasks it's working on in parallel. So now it's going to start writing that email and work on those two files I asked it to build. And again, it gives me each output as soon as it's finished. When those files get completed, they are showing up in this output folder. There's the Word document and it's uploading them right away to my OneDrive. They're cloud-based and protected by Purview right away and there's the PowerPoint here. We see that email for Nerja getting her up to speed. Again, I can edit it right from the prompt bar. And when I'm ready, I click send. Absolutely. Yes. And now that it's completed all four tasks I asked it to do, I get this summary of everything it's done for me. So I can review and make sure I've got a complete set of outputs. And let's take a look at the PowerPoint that it created just to show you an example of the kind of quality that Co-work produces. Here's that PowerPoint. And look, it's this super polished deck. I asked for something ready for execs. And I feel great about how this looks. So I better go jump into that meeting about value prop. And I think Nerja's got to pick it up for our finance business case.
So, you can basically interrupt the flow of work just like you were talking to a human saying, hey, listen. Actually, I need one more thing done. And add it to the flow without disrupting what you've already asked it to do. Awesome. So, it's right in the flow of how you work. You're not popping in and out. The context is there just as if you were, you know, going into Outlook and scheduling the meeting. It's right there in the flow of Co-work. Awesome. Super. Awesome. We'll go check in on the numbers and I'll be back with you. Thank you so much.
Hi Judson. How are you doing today?
I'm doing great Nerja. How are you?
I'm well. Thanks for being here with me.
So we just saw Ann kind of prepare the basic pitch deck and the power and that Word document. But the thing that matters most is can we make money on this solution. So let's talk about the numbers here.
Yeah, exactly. And as the finance lead, I also care about making money. So here we have an Excel sheet of our V1 launch. And so this launch has already happened. So you could see here that this is a general file that we have. And so this launch I'm going to use the assumptions from this launch to pressure test our V2 proposal. So you'll see here that on the right hand side before we get started I have access to model choice which means I can choose the right model for the job whether it be our latest GPT models like GPT 5.5 which we just announced yesterday or our latest Claude models. And so before I start, I'm going to move over here and then start with a quick prompt to show our V2 spend figures and use an opportunity brief to pull in an opportunity brief that I have. So using contextual search, it's going to pull in the most recent file that I have. Again, powered by Work IQ. Work IQ will surface the most relevant responses based on the work that I've been doing. And then I want to highlight in yellow the incremental additions that I have so I can quickly see the changes. So this is our V1 launch. And then in just a few moments I'll be able to see our V2 incremental changes. Now you'll notice that on the top left side that the confidentiality label changed from blue to orange. Now that's not just visual. That means that the security label was applied throughout the new folder because the V2 business opportunity brief that we used that's a confidential file. So that was carried over to this Excel sheet. Now let's take a look. Exactly. Exactly. So all happening in this one workflow and we see here super clear to able to see the highlighted changes that were added here. So now I understand not only the V1 assumptions but what's going to happen with this V2 launch. And if I needed a little bit more analysis on the right side I get a full in-depth analysis about what was added. So I can see here the key changes. So it tells me the 10 new items and then it also tells me the total combination of the spend. So now it's not just important to understand if this is going to be cost effective. I also want to know if it's going to be profitable. So, I'm going to use the context from another revenue projection sheet that I have and then make sure that again using contextual search pulling in the right data to see is this going to be a profitable venture and I also want to make sure I want to understand the best case and worst case models. So, you'll see on the grid as it starts filling in it's going to be live code, filling in live tables, live formulas, live syntax, all in the grid. And then on the right side, I can see its chain of thought reasoning as it works. So it tells me the final verification, making sure the syntax is accurate, and then also showing me the analysis. Now, let's take a look. Now, I'm not lying. Here we go. This is live formulas. So, we all know that things change, assumptions change. So, as they change, I can go in and edit. And again, nothing in the grid will break. And if I scroll down a little bit more, you'll see that these are also live charts as well. So, not static images. All right. Now, let's check out this analysis. Again, every single chart is accompanied with analysis. So, if I scroll down, I can see key revenue projections. I'm also able to see the Q1 investment breakdown as well. So it shows me exactly what everything in this chart reflects. So the content marketing, the product demos, the strategic partnerships, very differentiated allocations across this go to market. Scroll down. I see the full analysis. All right. And so here at Zava, we're trying out new technology. We're still very, very early in our adoption journey. And so here I have the Claude in Excel plugin. And so you could see here it's that exact same prompt that I use. No funny business. So you could see it's using the context from this revenue projection report. I want to see if this is profitable. And I've uploaded the revenue projection report here as well. But the difference is there's no contextual search. So no connection to Work IQ. So I have to go in and download this document. So it's a static document that doesn't reflect the changes if my stakeholders made them. But let's give it a go. So, this takes about 10ish minutes. This is sped up quite a bit. For a reference, the Copilot in Excel demo took in reality about 5-ish minutes. So, about double the time, but you still get that in-depth analysis. If I scroll all the way down, you also are able to see inside the grid summaries and tables, but you just don't get that security. You can see here that that security label was not applied. So you don't get that security right in the flow of work. It's not integrated with Purview. But if I scroll down again all the way down, so you see all these tables, which is great, but they're not accurate. So you see here that this V1 GTM investment allocation that we just saw, you saw that there was many differentiated workflows. We saw demos, we saw strategy, but here you just see one specific allocation, which is just not correct. Exactly. Exactly. Right. Thank you.
Cool. So basically if you think about this you have this V1 doc that you know you use from prior launch that could have been created with a different model or frankly created by hand. You now have V2 context sitting in a different document. You're asking now a new model to go operate against that and take V1 context and update it with V2 thinking. All within the boundaries of a trusted environment. Got it. Mhm. Yeah. So I think you're highlighting something really important here because you know while Anthropic is a great partner to Microsoft and they've built a nice plugin for Excel, the power of Work IQ is really differentiating here because it's faster, more accurate, and more trusted. You're able to produce the results in half the time. They're far more accurate based on what we're seeing here and they're more trusted because it picks up on the security context, the data and information that you have the rights and privileges to access and the things that you don't. Super awesome. Very well done here. Super. Okay, so now that we know that our marketing plan is actually financially viable and we're gonna kind of check back in with you and take these numbers and put them into the GTM context and the documents and PowerPoint presentations we're going to then show to our board and get approval to launch this product.
Let's go for it. Yeah. So Nerja got that finance case pulled together while I was in my meeting finalizing value prop and of course work being as busy as I also got an email while I was in the meeting with all the pricing details proposed for the launch. What we're looking at here is the Word document that Co-work created for us and notice it got a confidential sensitivity label from the start again because Copilot and Purview are natively integrated. Here's that document and we can see Co-work highlighted all the remaining sections. So we can see that value proposition still needs to be filled in. And if I keep going, our pricing strategy needs to be added as well. I'm going to use Copilot in Word to help me do just that. So I will ask Copilot to update the value prop section based on what was discussed in my meeting and to update pricing strategy based on the email I got. What you notice I didn't do was explicitly link the meeting or the email. Copilot has Work IQ so it has access to my entire context and it's able to find that content for me and as we see here edit the document in line and apply it. And we notice first of all that the sensitivity label was updated to highly confidential because the pricing email contained a more confidential sensitivity label than the original document and Co-work Copilot made the edits directly into my document. We can review the value prop it pulled everything we discussed in the meeting. And if we go all the way down, we can see that pricing strategy came through. All the numbers from my colleague. Looks good to me. So, I'll keep it. Now, just like Nerja, I'm also experimenting with different AI tools. So, I thought we would try out this exact same prompt in ChatGPT. So, I uploaded my document with the sections unfilled out and I did the exact same prompt as I did in Word. And I'll click submit. ChatGPT will get to work, take my prompt, apply it, and there we go. There's an output. Let me break down what just happened for you, though. So, while ChatGPT does have an Outlook connector, it was able to find the meeting invite from the meeting I just had, but it wasn't able to access the Teams transcript and recap. So, it wasn't able to pull exactly what was discussed. Similarly, it found that email from my colleague, but because it was highly confidential and encrypted, it wasn't able to access the actual pricing details. It did however create a document for us. So, we can take a look at what that looks like. Now, you'll notice it didn't edit the document that I gave it. It created a brand new document, which it applied a general sensitivity label to. So, it didn't maintain that confidential label from the beginning. And if we look down, it did fill in a value proposition and it did also, if we keep going. There we go. It provided a recommended pricing strategy, but you'll notice this doesn't match the numbers of what was in my email because it wasn't able to access that content from my work environment. Exactly. So, I personally prefer Copilot because it's embedded into my flow of work. It has that Work IQ and it has that built-in security. So, we've got the Word doc in good shape. We've got all those details of the launch together. Now, we just need to wrap up the PowerPoint and we should be good to go.
Really cool. Got it. Right? So, the model's trying here. It's inferencing or guessing over the financial data and putting it in the document. But to the reader, it allows you to make mistakes with greater confidence than ever before because the data is actually not right versus when you use Work IQ, it's the direct access to the information, your intelligence that you've harnessed. Cool. I can see Nerja's already put the business case from finance in a slide, but it's just a little bit basic and this is going in front of our executive team. So, I'm going to use PowerPoint inside Copilot. Copilot inside PowerPoint and ask it to simplify the slide so it's exec ready and make sure those key details popped. So, just like I did in Word, Copilot in PowerPoint is going to edit that slide directly based on what I've asked it to do. That's super cool. So what you've been able to do here is actually use a different model with the same context to update the PowerPoint information here because you started off originally in Copilot Co-work using an Anthropic model. Now you're in PowerPoint using an OpenAI model, but your Work IQ is maintaining that consistency across everything you do. Awesome. Super job, Ann. Thank you so much.
Exactly. Yes. And there's the output. We've got the big numbers popping. It's polished and ready for our executive team. And now we're good to go. And I love how Copilot didn't just help me create those assets, it also helped me enrich them so we could go forward with some really high quality content.
Thanks, Ann. So what Ann has shown you here is that you can create these assets using Work IQ in a model diverse and open and heterogeneous way that your output is faster, it's more accurate, and it's trusted and you can use it across multiple modalities across chatting with the data, across creating new assets or new artifacts as we call them but also augmenting those artifacts versus the models that will just simply create a new one that is outside of the context of your security controls. The next big modality though is actually building agents on top of that data and we're going to check out what's next here with Stephanie.
Yeah. How's it going Judson?
I'm doing great. We just got some fantastic news. I'm an operations manager at Zava and we just received the green light to move toward product launch. Now, part of every successful launch is ensuring that the material that we need is available. And that's my job within the process as a procurement manager. So, let's take a look at how by leveraging agents, I'm able to automate this process. So, here in Outlook, when I receive an email confirming that the launch plan has been approved, we're going to see the outline of everything that the go to market strategy is going to comprehend here. And now, from here, this is going to trigger the inventory management agent. And this agent is going to run an inventory audit. And that audit is going to help me better understand whether or not we have the material that we need to manufacture a new product. So here's the result of that. We can see the inventory management agent has sent me a Teams adaptive card. And if we take a closer look at this, we can see the summary here. Now, this is Work IQ behind the scenes helping the agent by helping it understand the product teams that I'm a part of, the different meetings that we've had, and of course, the deadlines that we're driving towards. And if we take a closer look here, we can see that the agent also identified some supply risks. It looks like we have a sensor module and a battery pod that are both running low on inventory. And so, it's not just going to stop there. It's going to take the next step and it's going to identify recommended suppliers. And then from there, it's going to go even further and it's going to generate an RFQ or a request for quote to request bids from those suppliers. Exactly. Exactly. Now, you might be wondering, how does an operations manager put something like this together? So, enter Copilot Studio. Now here I simply describe my ask and then the agent is going to execute it for me. So the first thing that I'm going to do is I'm going to provide my description and then from here I'm going to select a model. That's really awesome. Now, and beyond this, I also have over 11,000 models that I could choose from in Azure Foundry. So, after this, I need to provide my agent with some instructions. Now, this is just as if I were providing instructions to a co-worker. So, I'm going to tell my agent to look at the launch plan approval, to look at our go to market documents, identify any kind of supply chain risks, generate that RFQ, and send it out to our identified suppliers. Now, just like when humans execute this process, we need to give them some knowledge. And so here, I'm going to provide my agent with all these different data files, resources, and most notably this RFQ template. This is going to allow the agent to adopt the branding and the format that we typically use. And then we also have specific tools that we're connecting the agent to. So I want to call out here Work IQ. This is that intelligence layer that's really going to personalize the agent to me and the organization. It's going to connect the agent to all of our work apps. So, it's going to better understand the business, the job I do, and the teams that I work with. We also have this specific tool here, the Dynamics 365 ERP model context protocol server or MCP server. And this is our system of record. It's going to allow the agent to pull in inventory details. Exactly. And speaking of that, there's a lot of capabilities that I don't know how to do. And so, here we can kind of see that I've been able to leverage an agent that was built by the IT team. And this is going to allow us to create a comprehensive go-to-market strategy document. And it's been built on a custom model and it's been fine-tuned to create that super detailed and professional document. But this agent also operates autonomously. And we can see up here that we've got a few different triggers that are going to help the agent activate. So, let's test it out. All right. So, I'm going to trigger the agent by simulating the inventory review being complete. And then what we're going to see on the left side is the agent is going to reason in real time. All right, it looks like we got a little glitch here. So I'm just going to run to one that was already completed. So if we take a look, we can see that the agent leveraged Work IQ as we talked about. So this is the agent reasoning in real time. It's going to be searching across my documents, my teams, my organization so that it can better understand me and the business that we're running here.
Okay, cool. Great. So, it's a full agentic workflow. Cool. So, you're reasoning over the system of record. You're reasoning over Work IQ. You've added some other files to use as templates. You've built that all here without writing a single line of code. Really cool. Awesome. Super. Great. And then that D365 ERP server comes in or it's going to pull in the specifications so that we have information about the materials. Now if anything looks unusual, I can also look
Here to see the rationale of specific steps that the agent took. This is going to help me better understand why the agent did what it did.
So then on the right side, we can see the results of what the agent put together. And here's our summary. It's very clear. It's concise. I can get a good understanding of what the agent did. And then we can also take a look at the go-to-market document that I put together. So let's click on that. Now this is a living document. So it's going to open right up in Word online. And as I'm looking through this, I can see that it's super comprehensive, very detailed. It's got all of the different sources that Work IQ had pulled in. And this is a document that probably would have taken my team days to put together. And it put it together right here on stage.
That's awesome. So the power of Copilot Studio helped you leverage all kinds of data sources, Work IQ, your system of record, allowed you to use multiple models, frontier ones as well as some fine-tuned open source models to sort of reduce the cost associated with the agentic flow. And then from left to right, you've built something that automates your entire launch process all without writing a line of code.
Awesome job. Thanks, Stephanie. Well done.
Okay, so now we've gone from the early stages of marketing and conceiving of a product, running the numbers from the finance team, getting approval from the board by creating all kinds of rich artifacts that we could present. We've gone through the supply chain analysis and produced the GTM doc, and now we're going to talk to Seth and the professional development department about getting that product built and taking it to market.
That is correct. But before, I want to say something. Moikka kaikille. That's like hello everybody. I know it's the first clapping we've got in the whole presentation.
Hold on. I have another one. Menen.
Oh, I don't know what I said, but it's probably say you're an idiot.
No, no. I was laughing at you saying I'm a professional dev because I got an email on Friday from Ann and she said we need to do the new V2 website. And when Ann sends an email, we all have to just do what she says. And so instead of, because you know how I love reading emails and marketing documents, instead I'm going to use Copilot and Work IQ to make this all happen. What do you think?
Before we get there, I want to show you a couple things. So the first thing is that when you are looking at GitHub Copilot, this is GitHub Copilot and the CLI. You can choose from any number of models, and I'm talking about all of them. Recently GPT-5.5 came out. That's the one I use. It's pretty cool. But the other thing is you can also do a couple of things like you can shift how you work. So, for example, if you don't want to do anything, you just want to plan, you can send it into plan mode. Or sometimes when I have a very important project that I need to make overnight, I basically put it on autopilot and go to sleep and then wake up and see what it's done. And it actually, I have done this. This is a real thing. It's really cool. And then beyond that, for example, you can hook up things like skills, and you're going to see some interesting skills in here that help me bridge the gap between emails, documents, and work and dev. And so, those are all in here. So, I did this last Friday and when you look at it, I'm going to speed it up, but it took all of 10 minutes to do this. So, I'm going to speed it up. Here is me going in and saying, 'Hey, use Work IQ to find the latest email from Ann about the Zava store. Read the linked document. Then do what she asks because that's what we do, you know, at Zava, right? You're trusting.'
Yeah. Yeah. I'm trusting. And notice that it's using Work IQ, but it's using it in a very special way. First, it's saying, 'Let me locate the email.' I found the email. There's a full message because there's a document linked in SharePoint, right? And notice that usually you can't just pull a document from SharePoint because DRM and other issues. And notice that it says it has a linked document rather than a traditional document. I've got a SharePoint link. I'm resolving it in the graph. And notice Work IQ does something special because you can't just look at a document that's protected. It's going to ask Work IQ, 'Hey, can you read this for me and tell me what I can look at?'
Yeah, it's really cool. And so now I'm getting this document from Work IQ. It finds the two things. There's landing page content themes and future of AI and another future AI buying. So two things she wants us to do. So it's going to say, 'Hey, there's the two things that I'm going to make.' It puts them into GitHub as issues. And then I say, 'Hey, can you just take care of issue 23 and do that for me?' And then it goes ahead and delegates that work into the cloud with GitHub Copilot, which is pretty amazing, right? That's awesome. And notice it's doing, I stopped it there but I want to show you the actual issues that it created. Here is the first one: build the product document page. And then the second one we'll look at a little bit later. So as I assigned it, it actually created a PR and you can see in this PR when I click over to it, it actually implements the entire thing and gives me screenshots of what it actually looks like. And notice it has the design theme and everything. And just so you can see, I actually went over and I executed it so you can see the actual website right here, which is pretty cool.
I do have one question for you though. So if GitHub Copilot and Work IQ are doing all of this work, what exactly are you doing?
Yeah, it's an important job as a dev to supervise your agents.
Yeah, that's what I'm just telling everyone, that I've been supervising.
We'll talk about that later.
Okay. So, experience number two though. And this is where I make it. So you are actually doing this is the professional dev because I am going to do an actual AI experience for design. Should we try it out? Yep. All right. So let me go over here and again this is all you're going to see over here on the right hand side all of the debug statements because I want you to see that things are happening. So let's, here we go.
Maybe I should refresh the page.
Hi there. Great to have you here. I can help you with questions about Zava store shirts, analyze or edit your logo designs, and provide info on product policies and ordering. What would you like to explore today?
I want to make a brand new Zava store shirt, but I want a special logo. Can you capture and analyze an image so we can put it on the shirt? Go ahead and do that.
Let's see what you've got. Click the camera to capture your image. Working on analyzing the image. Hang tight.
All right. Here we go. Here we go. So, again, this is all happening live.
Hmm. That image didn't give me much to go on. Could you try capturing a clearer logo or design? Go ahead.
I want you to use it anyways. I want a blue Zava store shirt. And we're soccer fans, so we want the Suomi Kickers as the actual text. Can you make that for us? Well, oh, by the way, and I want just our faces in black and white cartoony line art. Can you make that for us?
Here we go. You still there? Can you do that?
She hung up on me. She looked at our faces and said, 'I do not want to do that.' But you know what? I did this earlier and I want to show you the actual session that we did with the Suomi Kickers. So, let me show you this playback because it is absolutely delicious. So, here we go.
Hi there. Great to connect.
I want to make a brand new Zava store shirt, but I want you to capture an image.
Absolutely, Seth. Let's see what you've got, please. Working on analyzing the image.
Here we go. Oh, here comes.
Whoa, that selfie is beaming with stage conference chic vibes. Let's...
I want to do a blue shirt with the Suomi Kickers as the sort of text underneath. And I want us to use only our faces, but like in a cartoony line art, black and white kind of style. Can you do that for me?
You got it. Blue Zava store shirts, Suomi Kickers text, and a black and white cartoon line art of your faces. What color do you want the text to be?
I'm thinking white stylized text.
Sweet. Last check. Does the cartoon line art style work for you as the logo style?
Yep, that's correct. Let's do it.
