رونوشت کامل Moonshot first, then tools: The way to AI value

Platform, that I don't think people fully understand the breadth and depth anymore. 33 00:01:45,038 --> 00:01:46,438 And it's moving fast. 34 00:01:46,498 --> 00:01:52,198 I mean, tens of millions of people use just a Power App in the last 28 days inside their company, 35 00:01:52,618 --> 00:01:55,598 and a lot of those were built by non-traditional technologists. 36 00:01:55,598 --> 00:01:58,218 A lot of those were built by technology teams, right, 37 00:01:59,098 --> 00:02:06,658 who just don't want to have to maintain a full stack of microservices and would much rather put the velocity into repeatability 38 00:02:06,889 --> 00:02:09,516 and, you know, much more high-value solutions. 39 00:02:09,516 --> 00:02:14,378 The key is unlocking the relationship between the two. You know, a lot of what 40 00:02:14,378 --> 00:02:18,089 makes software expensive and slow is not the software part. 41 00:02:18,089 --> 00:02:22,791 It's the, "what the heck are we building in the first place" part. And is it good or not? 42 00:02:22,891 --> 00:02:34,176 And the more you can get practitioners and people with process expertise on the same toolkit, around the same table, on the same page as technologists, the faster you go. 43 00:02:34,176 --> 00:02:38,038 That is all the more true in a world where agents are writing a lot of code and doing a lot of the processes. 44 00:02:38,402 --> 00:02:39,123 Yeah, that's great. 45 00:02:39,123 --> 00:02:46,609 You mentioned this earlier and I thought it was great because I talk about something similar. When you talk about what enterprise applications are and they're essentially 46 00:02:46,609 --> 00:02:48,991 data delivery environments, right? 47 00:02:48,991 --> 00:02:51,433 I mean, really it's process on top of data models. 48 00:02:51,433 --> 00:03:00,742 And the idea of unlocking that data across and having, being able to orchestrate an application that is unique is much faster to value, 49 00:03:00,742 --> 00:03:01,592 I think, is really important. 50 00:03:01,592 --> 00:03:06,062 Can you talk a little bit about the conversations you're having with clients around what they can unlock now with their data? 51 00:03:06,062 --> 00:03:07,932 Yeah. You know, I think the 52 00:03:08,262 --> 00:03:14,085 challenge has been we've built up all of these systems as architectural silos unto themselves. 53 00:03:14,085 --> 00:03:23,561 You know, your HR system is often very separate from your ITSM system, very separate from your CRM system, very separate from your ERP system. 54 00:03:23,561 --> 00:03:24,226 And that's natural. 55 00:03:24,226 --> 00:03:26,052 I mean, this is a very fragmented market. 56 00:03:26,052 --> 00:03:30,795 A lot of the vendors that are strong in one of those categories are not strong in the others. 57 00:03:31,115 --> 00:03:35,397 The problem is, real work doesn't care about your categories, your architectural silos. 58 00:03:35,478 --> 00:03:36,118 All the real stuff 59 00:03:36,118 --> 00:03:41,190 people do inside of an organization requires a lot of hops. 60 00:03:41,190 --> 00:03:43,601 You know, let's onboard a new supplier. 61 00:03:43,601 --> 00:03:50,304 Yes, technically that'll be a data entry into a table in an ERP system somewhere, but that doesn't come out of thin air, right? 62 00:03:50,304 --> 00:03:52,085 It's, what are all the meetings we're going to? 63 00:03:52,085 --> 00:03:57,587 The decks we're reviewing. The spreadsheets we're comparing. The sourcing systems we're consulting. 64 00:03:58,558 --> 00:04:03,820 You know, an end-to-end process requires moving across a lot of those different systems. 65 00:04:04,094 --> 00:04:08,408 And so, when it comes to unlocking, part of that is an insight question. 66 00:04:08,408 --> 00:04:12,514 How do you see across those boundaries and understand trends? 67 00:04:12,514 --> 00:04:14,043 But part of that is also operational. 68 00:04:14,043 --> 00:04:23,141 How do you automate work across those hops? Take humans out of the loop. When humans are in the loop, make much more cohesive and better experiences for them. 69 00:04:23,141 --> 00:04:24,102 And that's really how people are 70 00:04:24,102 --> 00:04:26,004 applying Power Platform. 71 00:04:26,004 --> 00:04:30,808 That's how they've been applying it for the last decade. But it's certainly how we're seeing a lot of acceleration in 72 00:04:30,808 --> 00:04:32,016 an era of 73 00:04:32,142 --> 00:04:37,346 agents that do that manual work, and also agents that write a lot of new code for better experiences. 74 00:04:37,346 --> 00:04:37,732 Yes, so 75 00:04:37,732 --> 00:04:38,364 the agents 76 00:04:38,364 --> 00:04:47,137 are interesting because I think it is a natural evolution from the Power Platform. But the concept of an agent can be difficult for a normal user to sort of understand, because you 77 00:04:47,137 --> 00:04:49,448 hear people talk about, oh, it's a digital human. 78 00:04:49,448 --> 00:04:50,406 It does a digital task. 79 00:04:50,406 --> 00:04:51,358 Can you talk, a couple of things: 80 00:04:51,358 --> 00:04:57,340 One, how would you define an agent in a normal mode? And then also, scale of an agent? 81 00:04:57,340 --> 00:05:00,601 I think sometimes people start to think of agents as being broader. 82 00:05:00,601 --> 00:05:03,916 You should think about the task orientation and then think about orchestrating those agents, right? 83 00:05:03,916 --> 00:05:05,172 Yeah. 100%. 84 00:05:05,172 --> 00:05:06,934 And so yeah, we got a lot of weird 85 00:05:06,934 --> 00:05:09,058 nouns in this space. 86 00:05:09,461 --> 00:05:10,213 Agentic. 87 00:05:10,213 --> 00:05:10,965 Yeah. 88 00:05:10,965 --> 00:05:12,327 Adjectives, too. 89 00:05:12,349 --> 00:05:13,140 I think, 90 00:05:13,479 --> 00:05:14,542 look, 91 00:05:14,934 --> 00:05:16,620 in very plainspoken terms, right, 92 00:05:16,620 --> 00:05:22,126 the difference between an agent and a chatbot is

the agent doesn't have to wait for somebody to type a message to it, and it can do more than type 93 00:05:22,126 --> 00:05:22,987 a message back. 94 00:05:22,987 --> 00:05:31,716 It's like delegating work to a member of your team, right? And so if you think about that, 95 00:05:31,716 --> 00:05:33,208 you know, that's a really powerful concept. 96 00:05:33,208 --> 00:05:40,604 How can I delegate work to a wider set of assets? And what does it mean for them to be good at those tasks? 97 00:05:40,610 --> 00:05:45,629 You know, Researcher is an agent built by Microsoft for a broad horizontal set of research work. 98 00:05:45,629 --> 00:05:52,338 There are also agents that need to be created for every single organization that are much more specific to their tasks. 99 00:05:52,338 --> 00:05:54,360 How do I go investigate a customer case? 100 00:05:54,360 --> 00:05:57,492 How do I go process a grant application? 101 00:05:57,492 --> 00:05:59,964 How do I go onboard an employee? 102 00:05:59,964 --> 00:06:05,718 These are all skills that right now are distributed among a lot of software systems, and lot of know-how. 103 00:06:05,718 --> 00:06:07,751 If we can encode them into agents, 104 00:06:07,751 --> 00:06:09,036 then we can move just a lot faster. 105 00:06:09,036 --> 00:06:14,566 Yeah, the interesting point there is in the past, that would have required developers around the table. 106 00:06:14,566 --> 00:06:19,441 We would have been talking about, even in the older days before cloud, we would have been talking about OS and DB and where we were on kernels, 107 00:06:19,441 --> 00:06:22,293 and now we're talking about, but still, not everybody understands the context. 108 00:06:22,293 --> 00:06:30,917 I think the idea of, and you talked about this earlier, getting that development closer to the practitioner is very important. 109 00:06:31,057 --> 00:06:33,474 So maybe speak a little bit about that. But also 110 00:06:33,474 --> 00:06:36,536 the change needed to help get people across that threshold 111 00:06:36,536 --> 00:06:39,938 because it is not, and in the federal government it's maybe a wider threshold. 112 00:06:39,938 --> 00:06:40,458 Sure. 113 00:06:40,458 --> 00:06:43,220 Just because of the nature of some of the applications they have. 114 00:06:43,220 --> 00:06:45,301 So how have you been talking to clients? 115 00:06:45,301 --> 00:06:47,822 So, those that are on the leading edge, how are they doing it? 116 00:06:48,142 --> 00:06:51,731 So, I would say everybody thinks they're behind right now. 117 00:06:52,656 --> 00:06:56,259 And yes, there are some organizations that are meaningfully ahead. 118 00:06:56,259 --> 00:06:58,050 And that gap, I fear, is actually 119 00:06:58,050 --> 00:07:02,172 widening between the organizations that are really moving in a modern way and those that are not. 120 00:07:03,273 --> 00:07:07,215 But nobody's actually figured this out yet. An everybody's on a journey. 121 00:07:07,215 --> 00:07:09,293 So its, you know, it's never too late to start, right? 122 00:07:09,293 --> 00:07:09,796 Is what I'm saying. 123 00:07:09,796 --> 00:07:10,026 Good. 124 00:07:10,026 --> 00:07:11,547 Best time to plant a tree is 40 years ago. 125 00:07:11,547 --> 00:07:13,118 Second best time is today. 126 00:07:13,318 --> 00:07:14,698 So, get started. 127 00:07:14,779 --> 00:07:18,441 You know, I will say, you know, some of this is technology constraints, right? 128 00:07:18,441 --> 00:07:21,062 What does it mean to get to a place where 129 00:07:21,240 --> 00:07:26,533 there are MCP servers in the organization so that as I build an agent, it can actually work with those systems? 130 00:07:26,533 --> 00:07:29,345 What does it mean to have authentication figured out? 131 00:07:29,345 --> 00:07:31,696 What does it mean to have organizational knowledge in a place? 132 00:07:31,937 --> 00:07:34,958 Like, there are some real tech things to go tackle. 133 00:07:35,399 --> 00:07:40,762 I would say the much bigger blocker for a lot of people, though, is cultural and organizational 134 00:07:41,803 --> 00:07:45,164 and knowing, what are we trying to achieve in the first place, 135 00:07:45,345 --> 00:07:47,512 because that actually focuses a lot of the 136 00:07:47,512 --> 00:07:48,728 technology questions. 137 00:07:48,728 --> 00:07:51,030 And I think sometimes 138 00:07:51,030 --> 00:07:51,750 we get confused. 139 00:07:51,750 --> 00:07:55,130 We think the AI use case is, you know, I need to go ship a chatbot. 140 00:07:55,230 --> 00:07:57,590 And that's the use case and let's go run that through. 141 00:07:57,590 --> 00:08:00,370 It'll have a start date and an end date and a technology provider. 142 00:08:00,890 --> 00:08:05,330 All the interesting AI use cases and organizations I'm meeting don't start at technology at all. 143 00:08:05,510 --> 00:08:09,030 I met a customer in Australia last month who said, here's my AI use case. 144 00:08:09,030 --> 00:08:11,150 Today it takes me 36 hours to generate a quote. 145 00:08:11,330 --> 00:08:13,650 I want it to be two hours by June. 146 00:08:14,150 --> 00:08:16,130 That is an AI use case. 147 00:08:16,550 --> 00:08:19,430 And that ambition of 36 hours to two hours, 148 00:08:19,564 --> 00:08:23,011 you don't get that just by, you know, working people a little harder. 149 00:08:23,011 --> 00:08:23,605 You have 150 00:08:23,605 --> 00:08:27,896 to fundamentally rethink, how does a quote get generated and slash 151 00:08:27,896 --> 00:08:29,256 a whole bunch of stuff out of the process. 152 00:08:30,237 --> 00:08:35,038 Doing it by June means I don't have the luxury to overthink this thing. 153 00:08:35,038 --> 00:08:41,540 You know, I don't have the time to go make an 18-month Gantt chart and hire a bunch of process experts to go study. 154 00:08:41,540 --> 00:08:49,838 Like, I got to start tomorrow, which means I have to get capability and expertise to the

place where the people understand the process the best 155 00:08:49,838 --> 00:08:51,598 today, right? 156 00:08:52,638 --> 00:08:55,438 And that is as much cultural as it is technology-based, right? 157 00:08:55,438 --> 00:08:56,506 How do I use 158 00:08:56,506 --> 00:09:01,406 that idea? 36 months to two or 36 hours to two hours by June. 159 00:09:01,406 --> 00:09:07,406 Use it as a rallying cry to motivate people, and then a focusing effort to go say, what needs to be true? 160 00:09:07,426 --> 00:09:13,226 That also means I've front-loaded the ROI so I know what I'm willing to spend or risk to make that outcome happen. 161 00:09:14,266 --> 00:09:19,466 I don't have to wait for a brainstorm and then some post-rationalization of AI value. 162 00:09:19,466 --> 00:09:21,226 I know the value up front. 163 00:09:21,266 --> 00:09:22,872 Now I need to get really curious about how to 164 00:09:22,872 --> 00:09:24,119 make it happen, and I think 165 00:09:24,280 --> 00:09:32,581 the organizations that are embracing that mindset, then they find the tools. They find the experts to help them. They find the people inside the organization to equip, 166 00:09:32,581 --> 00:09:33,533 and they make magic happen. 167 00:09:33,533 --> 00:09:39,394 They may or may not hit the goal as they originally intended, but being on the trajectory towards it is the valuable part. 168 00:09:39,394 --> 00:09:46,017 Well, I think number one is the cultural change of thinking that way unlocks value across the board, regardless of whether you hit the original goal or not. 169 00:09:46,017 --> 00:09:51,319 So, your goals may be, timing may not be working, but where you're trying to get to makes sense. 170 00:09:51,319 --> 00:09:53,900 I think the cultural piece is really interesting. 171 00:09:53,900 --> 00:09:58,962 The next piece of that really gets to, once you've unlocked that, you unlock a ton within the organization. 172 00:09:59,643 --> 00:10:06,872 Now there's one thing about POCs and building something out yourself, which we've seen be successful and not successful because of the way people went at it. 173 00:10:06,872 --> 00:10:08,113 But POCs to enterprise, 174 00:10:08,113 --> 00:10:09,803 and this is where I Microsoft differentiates. 175 00:10:09,803 --> 00:10:13,408 I'd love to get your thoughts on this, because I think this is where the entire stack comes in. 176 00:10:13,469 --> 00:10:19,154 Enterprise-deployed applications being driven by citizen developers is the goal. 177 00:10:19,374 --> 00:10:21,005 Yeah. Well, I mean, I think it is 178 00:10:21,005 --> 00:10:22,150 one of 179 00:10:22,228 --> 00:10:27,072 many tactics towards the goal. Like, making enterprise work more efficiently is 180 00:10:27,072 --> 00:10:28,169 the goal. 181 00:10:29,568 --> 00:10:32,240 You know, I think there's a lot of myths around citizen developers 182 00:10:32,561 --> 00:10:36,635 and frankly, a lot of people in IT roles have been terrified of this idea for a long time. 183 00:10:36,635 --> 00:10:38,306 However, here's the reality. 184 00:10:38,426 --> 00:10:43,250 Very few regular people woke up this morning and said, I want to build an agentic business application. 185 00:10:43,531 --> 00:10:46,133 Like, that is just not a thing that happens to real people. 186 00:10:46,394 --> 00:10:50,257 A lot of people did wake up this morning and say, man, this part of my job sucks. 187 00:10:50,758 --> 00:10:52,720 And I really wish it was better. 188 00:10:52,720 --> 00:10:55,732 And they know exactly what better looks like, 189 00:10:55,808 --> 00:10:57,449 and they're going to do something about it. 190 00:10:57,449 --> 00:11:04,465 Some portion of them will do something about it. Whether you have approved the technology or not, whether you like it or not, whether you manage it or not, 191 00:11:04,465 --> 00:11:11,551 everybody who's got a deployment of 1000 Power Apps and 100 AI agents, none of that stuff came out of thin air. 192 00:11:11,551 --> 00:11:14,083 All of it was preexisting in the organization. 193 00:11:14,083 --> 00:11:15,678 You just couldn't see it before. 194 00:11:15,894 --> 00:11:23,560 You know, and so for a lot of organizations, it's how do I make the right thing, the easy thing, so that people don't do something even riskier? 195 00:11:23,781 --> 00:11:25,812 And how do I go create a path 196 00:11:25,848 --> 00:11:27,921 where I can start to contain risk? 197 00:11:27,921 --> 00:11:32,346 Why are major global financial services institutions all in on Power Platform? 198 00:11:32,346 --> 00:11:36,130 Why are major government institutions pushing this at huge scale? 199 00:11:36,130 --> 00:11:43,040 It's a risk reduction tool. Not a risk acceleration tool, if it's used the right way. 200 00:11:43,040 --> 00:11:49,596 Yeah, and when I think about the application, one thing I've talked to boards and others and C-suites about is this idea, well two things: 201 00:11:49,596 --> 00:11:53,439 One, sometimes they'll say, well, I'm told I can't start because my data is not good. 202 00:11:53,439 --> 00:11:56,442 And I'm like, great. So you're making decisions with bad data today and you know it. 203 00:11:56,442 --> 00:11:56,702 Yeah. 204 00:11:56,702 --> 00:11:58,203 Not being elevated to the right place because 205 00:11:58,203 --> 00:12:00,185 it's sitting in those silos we talked about. 206 00:12:00,185 --> 00:12:10,958 I think the idea of, when we look across the environment, I want to make sure that we're applying the intelligence of the organization at every level, whether it's an agent, 207 00:12:10,958 --> 00:12:20,958 my most senior or my most junior employee, customer care agent, you know, whatever, that we are applying consistency across the organization. 208 00:12:20,958 --> 00:12:22,578 And so, I think it's going to be really interesting. 209 00:12:22,578 --> 00:12:24,358 I'd love to get your thoughts on this. 210 00:12:24,398 -->

00:12:30,498 Things like, you know, customer care or, you know, procure-to-pay or all those kinds of things. The process is the same. 211 00:12:30,498 --> 00:12:32,058 Procure-to-pay is procure-to-pay. 212 00:12:32,198 --> 00:12:34,218 But every organization will tell you their piece is unique. 213 00:12:34,218 --> 00:12:36,218 They do have a unique way they do it. 214 00:12:36,218 --> 00:12:36,674 And so 215 00:12:36,674 --> 00:12:46,146 how do you see the bifurcation? And it may be more than a bifurcation, of the core process, which is procure-to-pay or whatever, supply chain, to the "how we do it" layer? 216 00:12:46,146 --> 00:12:47,943 Yeah, I think that's a huge part of it. 217 00:12:47,943 --> 00:12:49,139 Look, this is why, you know, 218 00:12:50,658 --> 00:12:55,242 there's never really been a one-size-fits-all solution in business software. 219 00:12:55,242 --> 00:12:55,512 Right. 220 00:12:55,512 --> 00:12:59,505 And we've spent a decade or more trying to get to that holy grail. 221 00:13:00,186 --> 00:13:01,637 We customized the heck out of them. 222 00:13:01,637 --> 00:13:02,097 Yeah. 223 00:13:02,097 --> 00:13:10,104 And maybe some of that is frivolous and whatever, but a lot of that is, you know, what it takes to sell running shoes is very different from what it takes to sell wind turbines. 224 00:13:10,104 --> 00:13:13,736 And what it means to do that at one company is very different 225 00:13:13,797 --> 00:13:14,257 than another 226 00:13:14,257 --> 00:13:18,701 Like, this is why there's competitive advantage in a capital market. 227 00:13:18,701 --> 00:13:19,531 And that's okay. 228 00:13:19,531 --> 00:13:20,846 That's a good thing. 229 00:13:20,846 --> 00:13:26,426 The thing is, you got to be conscious about what is the real IP layer. 230 00:13:27,386 --> 00:13:28,786 You know, where is it repeatable? 231 00:13:28,786 --> 00:13:32,986 Where should I push something to a platform because it's not actually differentiated for me 232 00:13:32,986 --> 00:13:38,946 versus where is it worth it for me to take on the technical debt of having something custom and unique? 233 00:13:38,946 --> 00:13:49,726 What is it about this one particular procure-to-pay process that is going to make it ten times better than what you had before, and ten times better than any competitor's, you know, 234 00:13:49,726 --> 00:13:50,230 version of that, 235 00:13:50,230 --> 00:13:54,105 and what is just repetitive, and you want to be 236 00:13:54,105 --> 00:13:57,820 able to compress as much as possible, because when you compress as much as possible to an out-of-the-box product or a platform, 237 00:13:58,602 --> 00:14:04,946 you're buying yourself the bandwidth to be able to iterate and have cycles on the things that are more differentiated. 238 00:14:04,946 --> 00:14:05,886 I love that. 239 00:14:05,886 --> 00:14:06,826 That's brilliant. 240 00:14:06,826 --> 00:14:11,046 I think, if I'm sitting in front of a CEO and others, like, what is core to your business? 241 00:14:11,046 --> 00:14:14,286 I mean, today you don't know what's core IP and you're spending a lot of money on stuff that isn't. 242 00:14:14,286 --> 00:14:14,546 Right. 243 00:14:14,546 --> 00:14:14,806 Right. 244 00:14:14,806 --> 00:14:25,246 And I think that, I think getting to that concentrated layer of what is core to what we need to maintain, what, you know, how many times have you been, you know, been in 245 00:14:25,246 --> 00:14:27,186 organizations, like, IT is key to you organization. 246 00:14:27,266 --> 00:14:29,606 Is ITSM key? 247 00:14:29,606 --> 00:14:30,986 Is that strategically differentiated? 248 00:14:30,986 --> 00:14:31,366 Right. 249 00:14:31,366 --> 00:14:31,786 No. 250 00:14:31,786 --> 00:14:31,986 Right. 251 00:14:31,986 --> 00:14:33,206 I mean, it's not that you don't need it. 252 00:14:33,206 --> 00:14:33,998 You absolutely need it. 253 00:14:33,998 --> 00:14:34,520 Absolutely. 254 00:14:34,520 --> 00:14:37,213 But it's, what is the piece of how you do it? 255 00:14:37,213 --> 00:14:44,866 And I think if companies haven't started to articulate the strategy in that way, they're going to have to do it going forward, especially as the world starts to move pretty fast. 256 00:14:44,866 --> 00:14:45,844 Right, and I think that 257 00:14:45,844 --> 00:14:49,076 is what leads to these concepts of ambition and constraint, right? 258 00:14:49,076 --> 00:14:57,542 If a leader has a good idea of what is core and what is not, then I need to start compressing what I'm spending on the things that are not. 259 00:14:57,723 --> 00:15:03,447 And that's not just software dollars. It's also time and attention and opportunity cost. 260 00:15:03,447 --> 00:15:11,442 But if I can go start creating those constraints, then I can start creating bandwidth to be able to do things in other places. 261 00:15:11,442 --> 00:15:13,908 And that's, this is a very different mindset than how 262 00:15:13,908 --> 00:15:18,665 a lot of people are approaching AI right now, which is, let me wait for all my data to be perfectly organized. 263 00:15:18,665 --> 00:15:19,236 Then guess what? 264 00:15:19,236 --> 00:15:20,108 It's not going to happen. 265 00:15:20,108 --> 00:15:24,371 And let me wait for the perfect vendor solution to come along 266 00:15:24,371 --> 00:15:26,096 and have all this figured out. 267 00:15:26,346 --> 00:15:28,467 Usually you're waiting way too long at that point. 268 00:15:28,683 --> 00:15:29,713 You'll miss it. 269 00:15:30,082 --> 00:15:34,756 Let me go in and say, how do I compress my ITSM by 50% but make people happier? 270 00:15:34,756 --> 00:15:37,769 Yeah. 271 00:15:37,829 --> 00:15:38,630 Exactly. 272 00:15:38,630 --> 00:15:38,860 Right? 273 00:15:38,860 --> 00:15:43,374 Now I have a much more focused version of what data I need to clean up. 274 00:15:43,374 --> 00:15:44,819 What agents do I need to put in. 275 00:15:44,819 --> 00:15:46,076 What

software do I need. 276 00:15:46,076 --> 00:15:49,480 Where should I build differentiated value versus rely on platforms. 277 00:15:49,480 --> 00:15:51,742 I've made the world and the problem a lot smaller, 278 00:15:51,742 --> 00:15:53,283 and so it's actionable now. 279 00:15:53,884 --> 00:15:56,144 And that's the kind of shift that needs to happen 280 00:15:56,144 --> 00:15:57,698 in a lot of organizations. 281 00:15:57,698 --> 00:16:04,338 How do you see organizations changing their structure from an organizational perspective to take advantage of where things are going? 282 00:16:04,338 --> 00:16:05,537 Because it's moving rapidly. 283 00:16:05,537 --> 00:16:06,018 It is moving 284 00:16:06,018 --> 00:16:11,750 really rapidly. And I will share some that that we're doing too, because I think this is, 285 00:16:11,864 --> 00:16:15,884 you know, look, we're all on this journey, and Microsoft is experimenting as well. 286 00:16:16,147 --> 00:16:20,142 I would say, you know, it's maybe ironic, but one of the least productive things you can have 287 00:16:20,142 --> 00:16:21,767 right now is an AI team. 288 00:16:23,884 --> 00:16:25,027 That's a provocative statement. 289 00:16:25,027 --> 00:16:31,094 Well, I know, because, look, when there's an AI team, then everybody else assumes that their job is not 290 00:16:31,288 --> 00:16:31,850 to be the AI team. 291 00:16:31,850 --> 00:16:33,098 I love that. 292 00:16:33,254 --> 00:16:40,777 And then this AI team becomes the bottleneck for everything across the company. And like, we are past that, boy. Every team is an AI team right now 293 00:16:41,077 --> 00:16:43,078 whether they realize it or not. 294 00:16:43,598 --> 00:16:48,761 And like, really practically, you know, we used to concentrate all of our data scientists and applied scientists into one organization. 295 00:16:48,761 --> 00:16:53,042 We have made sure every engineering team has that capability within it right now, and 296 00:16:53,382 --> 00:16:56,544 it has raised the bar for engineers at Microsoft. 297 00:16:56,544 --> 00:17:02,228 It has raised the bar for applied scientists at Microsoft because they're working in a truly applied context, 298 00:17:02,228 --> 00:17:05,523 and everybody is responsible for the outcome. 299 00:17:05,523 --> 00:17:14,005 So, some of this is, you have to really mainstream capabilities that previously were set completely apart or separately if you actually want to go fast 300 00:17:14,005 --> 00:17:15,271 and do it at scale. 301 00:17:15,271 --> 00:17:16,505 Yeah. 302 00:17:16,505 --> 00:17:18,796 This is also why platforms are important, right? 303 00:17:18,796 --> 00:17:24,368 Because you're not going to instantly turn everybody into a PhD AI researcher overnight. 304 00:17:24,368 --> 00:17:24,598 Right. 305 00:17:24,598 --> 00:17:25,618 And you shouldn't have to. 306 00:17:25,618 --> 00:17:26,168 Nor do you want to. 307 00:17:26,168 --> 00:17:27,919 Right. 308 00:17:27,919 --> 00:17:31,960 You want to build on the shoulders of platforms that have already implemented 309 00:17:31,960 --> 00:17:33,094 those pieces. 310 00:17:33,094 --> 00:17:35,856 You know, I would also say along with that, it's like, 311 00:17:35,856 --> 00:17:41,092 how do we change layers of decision-making and, and sort of process cruft? 312 00:17:41,092 --> 00:17:44,456 You know, there's a lot of that in our organizations still. 313 00:17:44,972 --> 00:17:46,152 Yeah, yeah, I agree. 314 00:17:46,152 --> 00:17:48,884 And that's where, I know, some of the narratives have been out there on AI. 315 00:17:48,884 --> 00:17:52,435 Obviously, there's a lot of misinformation and we've seen the percentages of people that think it's positive. 316 00:17:52,435 --> 00:17:56,797 And I think part of that's because of how it gets measured both in the marketplace and then how it's been applied. 317 00:17:56,797 --> 00:17:59,408 I think people are, ahh cost takeout. People takeout. People's jobs. 318 00:17:59,408 --> 00:18:03,179 And the reality is, where you were going is exactly right. 319 00:18:03,179 --> 00:18:07,141 We need to encourage people to make their lives easier by driving, 320 00:18:07,141 --> 00:18:08,672 does that lead to more efficiency? 321 00:18:08,672 --> 00:18:13,674 It does, but hopefully it also leads to more business, which leads to more value, which leads to more people. 322 00:18:13,674 --> 00:18:15,174 Right? 323 00:18:15,214 --> 00:18:19,634 I mean, I think that's incumbent on organization leadership. 324 00:18:19,634 --> 00:18:21,694 I had these conversations many times with executives. 325 00:18:21,694 --> 00:18:23,454 I'm like, yeah, you'll be more efficient. 326 00:18:23,594 --> 00:18:31,334 But hopefully we're doing more of what we do, whether it's, whether you're a CPG company or you're serving Veterans or you're serving taxpayers. 327 00:18:31,334 --> 00:18:37,794 Like, you know, I went to the DMV last week and I'm telling you, in Florida, it was a joyous experience. 328 00:18:37,834 --> 00:18:40,054 I was in and out in 15 minutes to get a license 329 00:18:40,634 --> 00:18:42,094 because the process was well done 330 00:18:42,094 --> 00:18:43,114 and the notification was good. 331 00:18:43,114 --> 00:18:43,970 I'm like, 332 00:18:43,970 --> 00:18:44,421 things can be done. 333 00:18:44,421 --> 00:18:45,262 Things can be better. 334 00:18:45,262 --> 00:18:47,054 This is so important. 335 00:18:47,055 --> 00:18:53,472 Like, this is a really touchy subject. There's very few technology executives that want to go be the doom and gloom 336 00:18:53,472 --> 00:18:55,999 person talking about all that like, 337 00:18:56,110 --> 00:18:57,250 jobs will change. 338 00:18:57,250 --> 00:18:57,770 It will change. 339 00:18:57,770 --> 00:18:58,490 Things will change. 340 00:18:58,490 --> 00:19:07,690 But the least interesting question we can ask of this technology is how do we do everything exactly

like we did it before with just a couple fewer FTEs? 341 00:19:07,690 --> 00:19:11,650 That is just the fundamentally most uninspired question you could possibly ask. 342 00:19:12,090 --> 00:19:14,910 Like, what does it mean to be ten times better at customer service? 343 00:19:14,910 --> 00:19:17,430 What does it mean to have ten times better citizen engagement? 344 00:19:17,430 --> 00:19:20,570 What does it mean to have a ten times better DMV experience? 345 00:19:20,570 --> 00:19:22,990 Those are the interesting questions, right? 346 00:19:22,990 --> 00:19:26,112 And those will create a lot more value 347 00:19:26,112 --> 00:19:29,074 under the curve than any kind of cutting will do. 348 00:19:29,074 --> 00:19:34,228 So yes, it will require change and reskilling and a lot of cultural shifting. 349 00:19:34,808 --> 00:19:42,173 But gosh, I hope, I hope we have something better to show for that on the other side than a one-time balance sheet adjustment. 350 00:19:42,754 --> 00:19:44,505 And I believe that that's possible. 351 00:19:44,505 --> 00:19:48,178 I believe we can build far better, more impactful things. 352 00:19:48,178 --> 00:19:50,559 You got to come at it from a perspective of that 353 00:19:51,080 --> 00:19:53,942 future abundance though, and not just the short-term scarcity. 354 00:19:54,650 --> 00:19:57,192 I always say the generative part of generative AI is the human. 355 00:19:57,192 --> 00:19:57,592 Right. 356 00:19:57,592 --> 00:19:58,022 Right. 357 00:19:58,022 --> 00:19:59,813 I mean, the AI is, we've had AI. 358 00:19:59,813 --> 00:20:00,474 Right. 359 00:20:00,474 --> 00:20:02,795 It's obviously more accessible now. 360 00:20:02,795 --> 00:20:04,616 And again, we can drive it. 361 00:20:04,616 --> 00:20:09,009 But the idea of building applications, and this is where Power Platform, to me, unlocks so much value. 362 00:20:09,009 --> 00:20:14,562 I mean, what percentage of companies do you think are actually taking advantage of it in a way that you would say is leading? 363 00:20:14,562 --> 00:20:16,417 Yeah, I mean, I think, 364 00:20:17,070 --> 00:20:22,190 look, there's something like 97% of the Fortune 500 are actively using Power Platform at big scale. 365 00:20:22,190 --> 00:20:26,110 So a lot of percentage of companies are, hundreds of thousands across the world. 366 00:20:26,590 --> 00:20:33,530 But in terms of a "leading," you know, who's doing more than checking the box and saying they did a thing or perpetuating the last generation. 367 00:20:33,890 --> 00:20:35,150 I would say that's still very small. 368 00:20:35,150 --> 00:20:36,292 And that's where, 369 00:20:36,292 --> 00:20:37,633 you know, 370 00:20:38,168 --> 00:20:40,386 the ink is not dry on this chapter of history. 371 00:20:40,386 --> 00:20:44,678 There's still a lot of room for impact and a lot of room for evolution. 372 00:20:46,146 --> 00:20:49,474 You know, a lot of people feel like, man, we've been talking about this AI thing for years now, right? 373 00:20:49,474 --> 00:20:51,698 Gosh, it's still early in the shift. 374 00:20:51,960 --> 00:20:53,132 It's very early. 375 00:20:53,132 --> 00:20:53,682 Yeah. 376 00:20:53,682 --> 00:20:58,676 So maybe talk a little bit about, you know, the private sector moved faster. 377 00:20:58,676 --> 00:21:00,568 They do move faster in general. 378 00:21:00,568 --> 00:21:08,664 I think there's been obviously a shift, especially in the U.S. federal business, but also state and local and others, of driving, you know, really trying to get more commercialized 379 00:21:08,664 --> 00:21:09,415 and drive faster. 380 00:21:09,415 --> 00:21:15,470 What lessons can the agencies learn, you know, in both modernizing their processes and driving this kind of cultural change? 381 00:21:15,470 --> 00:21:16,151 What have you seen? 382 00:21:16,151 --> 00:21:18,574 Because you've been part of this environment for a while. 383 00:21:18,574 --> 00:21:20,270 Yeah, I mean, look, one of the 384 00:21:20,270 --> 00:21:28,496 other organizational challenges that I think we see a lot in both the public sector and the private sector is, today, you have this very organizational separation in a lot of places 385 00:21:28,496 --> 00:21:35,982 where you have one team that owns all the benefit of going faster, and another team that owns all the risk of ending up in the news for the wrong reasons. 386 00:21:36,422 --> 00:21:38,314 And that creates loggerheads, right? 387 00:21:38,314 --> 00:21:46,764 Because if somebody who owns all the upside doesn't own any of the downside, then they have no incentive to be safe and go slow. 388 00:21:46,764 --> 00:21:50,516 If somebody owns all the downside and none of the upside, then they have all the incentive to put up blockers 389 00:21:50,516 --> 00:21:52,821 and make it hard to do anything. 390 00:21:52,821 --> 00:21:57,821 And so, you get these logjams, you know, where it gets really tough to make progress. 391 00:21:58,092 --> 00:22:04,965 And so I think one of the things to really look at is, first you got to understand, like the world is moving really fast. 392 00:22:04,965 --> 00:22:06,715 It is risky to go too fast. 393 00:22:06,715 --> 00:22:09,556 It is also very risky to go too slow. 394 00:22:09,777 --> 00:22:14,709 You know, from a pure infosec perspective, threat actors are not sitting on their heels here. 395 00:22:14,709 --> 00:22:20,731 They are not throwing up the internal AI review board that's going to sit on something for six months. 396 00:22:20,731 --> 00:22:22,776 Like, they are moving super fast. 397 00:22:22,776 --> 00:22:33,104 But also just from a pure operational perspective, expectations are accelerating in the industry, in the market, from citizen perspective, much faster than a lot of organizations 398 00:22:33,104 --> 00:22:34,146 are moving. 399 00:22:34,146 --> 00:22:36,558 And so there's real operational risk to not 400 00:22:36,558 -->

00:22:37,653 moving fast. 401 00:22:37,718 --> 00:22:40,109 And I would say speed of being uncomfortable, but not risky. 402 00:22:40,109 --> 00:22:46,661 You don't want to push it past the idea of what the company can absorb, but you want to push past the comfort zone. 403 00:22:46,661 --> 00:22:58,384 Because if you're not, your competitors are in the commercial space. And, you know, and actually even in the regulated space, there's still scrutiny. 404 00:22:58,384 --> 00:23:04,066 The scrutiny comes from either regulators or, whatever, your utility commissions. 405 00:23:05,058 --> 00:23:11,466 And when you get asked the question of, what are you doing to serve the customer better or the taxpayer better, is do you have an answer? 406 00:23:11,662 --> 00:23:20,482 Well, and this is where, in particular, in highly regulated spaces, there's huge opportunity for massive de-risking to happen with AI. 407 00:23:20,482 --> 00:23:23,435 Oh yeah, even more than deregulated spaces. 408 00:23:23,435 --> 00:23:23,904 Absolutely. 409 00:23:23,904 --> 00:23:30,809 An agent, unlike a rotating cast of humans, will document what it does every single time. 410 00:23:30,809 --> 00:23:31,740 Right. 411 00:23:31,740 --> 00:23:34,812 It will take instructions and process them every single time. 412 00:23:34,812 --> 00:23:46,641 It will use a system in the exact same way every single time. Like, there are parts of these things that we can make far more deterministically automated, because we've been able to 413 00:23:46,641 --> 00:23:51,478 add agent capabilities and tools around traditional automation tools 414 00:23:51,478 --> 00:23:59,925 in a way that actually increases consistency, increases auditability, increases throughput, and actually lowers our risk quotient. 415 00:23:59,925 --> 00:24:07,572 Similarly, there's a lot of things humans are doing today under their desks. And it's not only inefficient, it's very distributed. 416 00:24:07,572 --> 00:24:13,237 And this is why a lot of financial services institutions and other regulated industries are moving onto Power Platform. 417 00:24:13,237 --> 00:24:16,740 It's all those apps and tables. Again, they didn't come out of thin air. 418 00:24:16,740 --> 00:24:18,621 They already existed in the organization. 419 00:24:18,621 --> 00:24:19,778 They just weren't 420 00:24:19,778 --> 00:24:23,230 very easy to track. And they certainly weren't easy to be consistent. 421 00:24:23,230 --> 00:24:34,686 So how do I move that end user compute and that actual shadow IT onto a platform that is visible and is governable, and then use agents and AI to make it far more repeatable, 422 00:24:35,447 --> 00:24:36,347 and, you know, auditable. 423 00:24:36,347 --> 00:24:41,450 There's a massive amount of risk benefit that we can gain if we do this right. 424 00:24:41,790 --> 00:24:48,914 If we get stopped at the sort of AI moral panic headline fear, then we actually miss a massive opportunity 425 00:24:49,026 --> 00:24:49,802 to go improve 426 00:24:49,802 --> 00:24:52,418 our regulatory stance in a lot of these places. 427 00:24:52,418 --> 00:24:53,439 I think that's a great comment. 428 00:24:53,439 --> 00:25:01,256 I mean, I remember the early days of cloud, the last ones to go were the most regulated, whether it was utilities or financial services or healthcare, but they were the first ones 429 00:25:01,256 --> 00:25:07,012 to drive, I think, and adopt AI because the promise, because of the regulated, and the process orientation in the data. 430 00:25:07,012 --> 00:25:10,245 It actually allowed them to move much faster, which I think was great. 431 00:25:10,245 --> 00:25:12,076 So I think that's a key lesson. 432 00:25:12,277 --> 00:25:13,882 So, Ryan, we call this GuideLine. 433 00:25:13,882 --> 00:25:20,189 The whole idea is really thinking about giving some insight and information to our clients, our customers, and to the business. 434 00:25:20,189 --> 00:25:25,075 So if there's one guideline we think about, sort of, your mission and the statement of what Microsoft's doing, 435 00:25:25,075 --> 00:25:31,441 what would you, what would be the guideline you give to executives or others out there who are now going on this journey? What would you give them? 436 00:25:31,630 --> 00:25:36,330 Yeah, I mean, we've been talking about it this whole session, but I think to sum it up in a nutshell, it's know your moonshot, 437 00:25:36,570 --> 00:25:45,150 you know, front-load and ambition and constraint. Then give the right tools to the right people and then create that culture of iterating to climb the hill. 438 00:25:46,310 --> 00:25:47,470 That is the recipe. 439 00:25:47,470 --> 00:25:55,370 That is the combination that's providing a lot of value. And that will guide you to far better technology decisions and partnerships and product implementations. 440 00:25:56,550 --> 00:25:58,895 Without that, it's all just shiny objects 441 00:25:58,895 --> 00:26:00,396 and science projects. 442 00:26:00,396 --> 00:26:00,907 Great. 443 00:26:00,907 --> 00:26:01,327 Thank you. 444 00:26:01,327 --> 00:26:06,526 Thanks again for coming and thanks for spending time with us. And I look forward to playing with some of the tools. 445 00:26:06,894 --> 00:26:08,774 Likewise. Thank you very much. Cheers. 446 00:26:14,286 --> 00:26:18,712 The GuideLine is a podcast brought to you by Guidehouse. Subscribe wherever you listen to podcasts so you never miss an episode. 447 00:26:18,712 --> 00:26:22,737 To see more from Guidehouse, head to our website, guidehouse.com.