PwC’s Dallas Dolen: AI Hardware’s Advantage, TokenMaxxing, Automation’s Impact
Channel: Alex Kantrowitz
Published at: 2026-05-15
YouTube video id: 85Et5V2kIUc
Source: https://www.youtube.com/watch?v=85Et5V2kIUc
What does the AI buildout look like on the ground? Let's talk about it with PWC TMT leader Dallas Dolan from the floor of Google Cloud Next in a conversation brought to you by PWC. Dallas, [music] great to see you. Welcome. >> Thank you. It's great to be here. >> So, we are just off the floor at this conference, the Google Cloud Next conference, and our conversation was all about the trillion dollar buildout of AI factories or in layman's terms, a lot of data centers. Yep. and we have this massive investment. That trillion dollars might even just be this year. Um, question for you, is that investment wise or is someone going to lose their shirt on that investment? >> It's like anything. I I think you got to look for at history a little bit and say, "Hey, what's happened historically with respect to other, you know, infrastructure um buildouts. You can look at fiber. You can look at, you know, some of the original internet and even the original telco and what they're having to do too because telco is a major part of the buildout as well, right? I think the forecasts, just to put the numbers in in broader perspective, over a 10-year period of time, something around $7 trillion being spent on data centers. That's one of the forecasts I've seen, there's actually still almost a trillion dollars has to be spent on telco. Someone's absolutely going to lose out from an investor point of view because not every single one of those bets is going to hit. But the other part of it, and I'm sure we'll get into it, is, you know, how many of these projects are really going to go forward. So you have commitments to spend money but the actual ability to benefit from the spend of money and the even the ability to deploy it those are severely limited by actual supplies of chips and workers to build the the the buildings themselves and of course even the te telecom you know uh infrastructure components like copper um so you have all these different like you know minor components so to say or major components they're going to be limited absolutely invest at your own risk I think would be what you'd say in any equity environment and certainly this one when you're talking about numbers like at not everything is going to hit even though we're in Vegas. >> All right. Yeah. And so that's obviously where this Google Cloud Next conference is taking place. Um and one of the cool things about speaking with someone in your position is you're both buying the technology from a company like Google and you're working to implement it uh within you know all the customers that you work with. So you have this like real on the ground view of what's happening with the technology. And of course that involves pricing. And let's just take this discussion a step further. You know, one of the questions that you brought up while we were here and I think is worth asking to you is does the AI business have a different look in terms of margins? Software business obviously looks at 70 80% margins but if you have a trillion dollar buildout what do the margins actually look like for every part of the AI stack whether that might be you whether it could be the data center and the compute providers and the model builders. So why don't you just walk us through what the economics of this business? >> So I love the question because about a year ago I started saying and I think a lot of other you know analysts and people in the industry you know saw this right you had a significant shift in the value proposition. Where was the value coming from? It went from software back to hardware. Remember you know talk 20 years ago maybe even more recently than that a lot of the margin for all the players was actually in hardware. There's a couple companies, the Apples of the world who still have a ton of margin, you know, in their hardware business, but they also do a lot in software, too. Um, so that shift from SAS to hardware that we're seeing in particular in the chip space is really remarkable. It doesn't make those businesses anything less than cyclical like they always have been, but they are on the high side of a cycle, right? We've you've seen this before. So, we're on the high side of the cycle and we have a shift in profitability back to the value proposition which is there because AI has disrupted the value proposition associated with SAS. One thing that I said a couple weeks ago um it you know front facing to the market was you know AI is kind of the new SAS or said differently like SAS is dead long live SAS it's just a different type of SAS and the way that it's happening now is on the services side where was the value the value was in what could engineers create what sort of custom software that you could propagate an enterprise level that you could then sell seat license in particular although it's shifted over time into consumption but what's happened you had a lot of margin expansion because there was a lot of value in what those engineers were doing. You had a few companies who owned a couple subsets of the market. They had moes. Well, guess what? AI has shrunk those moes and so you can now do CRM on your own. You can now do some components of ERP on your own. You can now do customer attraction on your own and marketing and that type of functionality. Well, that's a compressing feature, right? So in addition to the pure rotation of the value proposition from software to hardware, you also have compression that's happening on the software side naturally because AI's come in. You don't need as many people to write the code. You can build things maybe not in a day. I think there's a lot a lot of hyperbolic stories like that, but you can build things really quickly and even the large enterprise software providers are building things really quickly. Their engineers are changing how they do it. And so I I think it's it's a remarkable shift that's both, you know, moved money from one pocket to another pocket, but it's also compressed the ability, you know, from the I'll say the from pocket, so to say, to make money in the SAS place. Um, now the the first part of your question was around AI. How much money will the AI companies make? Right. Yeah, >> that's a really interesting one because I think something you said earlier today really like hit a chord with a lot of people which was, hey, how many folks are willing to spend more money for what they're getting? And I'd say roughly 30 40% of the room said they'd pay up to five times as much what they're paying today for for the services that they're getting. That was not for advanced services. That was for the services they're getting today. They'd pay five times as much. So if it's $19.99 a month, they're willing to pay a hundred bucks a month. That's great news for the Frontier model builders, right? the open ais, the anthropics and the Googles of the world and of course everybody else, you know, the Chinese firms too. That's really great news. How long that stays a reality is going to turn into a question of return on investment. And so if it's an individual consumer making that decision, maybe that person is a great user of AI. If you have someone on the on the, you know, enterprise side, they're going to start asking the question of where's my return on investment. >> Yeah. I I just want to pick up on something you said. It's really kind of fascinating that we have gone from this moment of um the hardware was looked at as commodity and software differentiated and it's kind of flipped. It's >> flipped. >> Yeah. Totally flipped. >> It's wild. So, uh, when you think about the decisions that you make and that your customers make, does this bottleneck of compute, so to speak, come into play, or are you able to find enough good enough, cheap enough models to make what you're trying to do work? >> Yeah. So, there's a couple bottlenecks. I mean, I think you have a trade bottleneck that's coming into play for all the hardware producers, which is really interesting and probably gets talked about like when there's tariff news, but like not talked about as much as you might think. Um, >> wait, what? A trade bottleneck? >> Just a pure trade bottleneck as in like the inability to get products into the country because it costs more or because you're paying more of a tax there's more uncertainty about the supply chain itself. So just that alone has actually disrupted some of the hardware companies who by the way could as you said and as I said could make more money because hardware is the you know the new software whatever you want to say about it right hardware is the new it it new it thing and so by virtue of the same they would love to import more because more people want Mac minis to run their claw bots or whatever it might be the supply isn't there in a lot of cases for most of these companies to expand revenue which in turn would expand you know margin so they're having margin compression with higher cost of trade in addition to just like not being able to meet the moment which is kind of painful for some of the hardware folks. So that's that's one part of it. And then the other bit I think that you know goes goes to your question as well is you know what are we seeing from a like a decision-m process like right now I'm not aware of in an enterprise environment where they're not able to run the things that they've built. Okay. So they haven't run into that you know we'll call it the the the token limitation or the compute limitation. They've run into that we've spent too much money. Yep. Well, we we've had that. We've had to like turn teams off who've spent too much money on tokens. You know, maybe it wasn't for all the right things. We're like watching what people are doing, wait a second, I don't think we want you to use AI for that. You know, go back to doing it the way it was before. That costs more. And so, we're making these really interesting decisions on ROI and on like what the right, you know, use cases and flows are for the >> not pure token maxers. >> There is no token maxing going on. There is some token maxing going on and there are some competitive leaderboards within the organization too. I see that. Absolutely. I mean you it is fun like so we have about 300,000 people globally right so when we want to gamify something like that we find that that is actually a really great way to get people into doing something new gamification we definitely have the construct of gamification going on as it relates to AI so both training and usage so there is a little bit of the you know the token maximization happening but it's done within our own pretenses and we have some you know some guard rails around that the most important of which is security right just making sure they're not going outside the the confines of the environments that we've set up. And so we've been very very specific about that. Um but you know, back to the question on on the enterprise bit, I don't think I've seen a lot of companies saying they can't get done what they need to do because they're hitting some sort of wall. Unless of course they're 100% AI native, in which case I have heard conversations about where they simply need more and they're very low on the depth chart, so to say, as far as like what they're getting from the provider. That's come up. We've seen a couple news stories about that one where they're going in and talking with the CEOs of these companies saying, "Hey, I just simply need more. I need your help." Um, you're also in an interesting position because you basically have to decide or PWC decides who do I work with. Um and very interesting moment uh that I think uh sort of captures where we are on the Dwarish podcast with Jensen from Nvidia where um you know Dwarvesh was talking to him about uh the fact that if you look at the worldleading models two of the three uh were trained um outside of the Nvidia stack. >> Uh this [clears throat] of course all built up to then you get to the export control question. Jensen gets frustrated enough, he goes, "I didn't wake up a loser." And that's sort of the headline. Uh, you're not talking to someone who woke up a loser. And that's the headline of the conversation. But, um, that point really stuck with me, which really sort of comes down to this question, which is there's two competing ecosystem, the Nvidia ecosystem and the accelerator ecosystem. Um, with a couple of cloud providers, Google being one of them. Yep. Do you have to work with everyone? Do you have to make a bet or one on one or the other? And then is there is there going to come a point where you see one as winning and just say all right in there? >> Yeah, that's a great question. I mean, I think the the way we look at it because we have a a very large, you know, strategy system integrator and an AI business in addition to the things that we do just traditionally as a firm, right? We have audit and tax and sort of our traditional, you know, business advisory services including cyber is we see a place currently for all of those firms to work with us. So we have use cases for all the hyperscalers. We have use cases for all the providers of Frontier models and we do that for a couple reasons. One, some things are just better than others given a certain environment. So when we think about security, you know, all the things that we're doing with Google, we have a big bet on that, right? We do have a very large GTM team um and and doing a lot with, you know, some new leadership that we have there. Morgan Damsky being one um here with PWC who just recently joined us. So that's one good example of it. And then we also have situations where we look at our engineers. We say, "Okay, you know, our engineers, they're going to pick the highest and best for them. So if they want to use cloud code, they're definitely going to go do that." And that's okay, too. Um, so that's one bit. And then the other thing that we're very responsive to is we look at the market and say, "Hey, what's the market demanding from us, right? >> We have demand that's really coming in." And it's, you know, it's funny. You see those usage charts and they're really, "Oh, but look at this model uses, look at that model." the the the the use cases and the demand is coming in in about third third as it relates to those right and I'm thinking and and speaking a bit from a um you know I say North American perspective here right and what I see with my client base within tech and telco but I also see it across industries being about that there are certain you know companies are playing a little bit stronger in a given industry you know I think cloud's had its moment with fs recently and you see a lot of FS companies who have thousands of engineers yeah sorry financial who have thousands of engineers on the uh you know on their payroll saying hey we want this and so they're having their moment there you at the same time right like if you look at the engineering capacity and capabilities you know of cursor and of Gemini like it's right there too so these things are all kind of neck and neck and their ability to do deals is is really functionally a part of you know maybe what's hot at the moment but also what's their long-term bet I could see a long-term bet decision-m process edging in favor of the models who have something underpinning them either financially or structurally, right? The actual [clears throat] infra itself. So where are the data centers or the AI factory? Who owns the pipes? You know, who can make sure it gets to them, you know, in in the most effective way because eventually going back a little bit on our combo, hey, we are going to run into structural limitations, right? We don't have the nuclear power. As it so turns out, we don't have the number of data centers. So all of these things are going to be limits. The question is, how long does it take for us to reach them? Is it 27? Is it 30? I'm not quite sure, but we're already going to start to see some of the supply chain limits stopping the production of some of the data centers at which point, you know, the acceleration of that limitation probably comes into play much faster than anybody assumed. So, you're sparking a question in me. I was in a uh a briefing with large AI lab today and one of the reporters asked um the executives there when you automate all your AI researchers what are they going to do like when they use the technology to automate their own work what will they do and you know I think the headline around a lot of this has been well AI will um you know replace software but I also wonder from like your standpoint think about systems integration for instance We've seen so many instances where these AI models that are getting really good at code. If you say refactor my code in another language, it can oneshot it in some in some uh instances. So what does that mean for businesses like systems integration? If you could for for instance say integrate these systems and maybe there's a world where claude or codeex or Gemini can handle that. >> Yeah. Um, look, I I think the reality is uh we have a completely different operating model as it relates to humans and technology. >> I don't know that anybody has the exact, you know, perfect point of view on what that is. If if structurally, you know, the organization is a pyramid, >> there's a shrinking of base. The question is, does it look like a spindle, which is a good term I heard last week. I like that one. Does it look like a monolith? I know there's a couple people written about that. Is it just a narrower pier? Or is it like the sail of a of a boat, >> right? Meaning that the top is the is the the narrow part and then as you >> Yeah. The question is how how wide is the base going to be on your organization, right? Engineering, sales, front office, back office, right? All those things like it's just going to be different. Um you have to keep in mind though, we still have an I'll call it a you know, I'm going say like a a need for a consumer and a need and actually I had a funny conversation this morning um with with one of our consumer markets guys. Uh, we went for a run. He actually tricked me into doing a five mile run and doing an interview on the five mile run. That was a whole thing. And then we did the last bit of the interview on like mile number five, like five five and a half miles into it. I'm like, I'm out of breath, man. Like, I'm going to need an IV after this and maybe laying on a stretcher. But nonetheless, he we were talking about how far is Agenta Commerce going to take things. If you look at Agent Commerce, it's the it's going to do some of the shopping. I don't think anyone's going to say, I want it to do all of my shopping for me. like there's still something enjoyable about that. Take the concept of enjoyment and supplement that with the concept of quality and the concept of human in loop. I think the same thing is going to happen within enterprise regardless of the size of the business and also regardless of the function itself. And so yes, we're going to go from pyramid to more like you know more like the the spindle or the the super narrow you know sale itself. But that's going to be the transition that happens probably in the next two to five years. it will be influenced by the amount of compute because if you're taking all these bits and pieces out, right? The cost of compute today, which I think you and I both agree is really, really low, the number probably goes up if demand's going up, right? Just natural basic macroeconomics. So, if that happens, we're going to find ourselves in a position where you might be saying, "Oh, I could replace that person with some level of agentics, but you know, it's actually cheaper to have the person because if this thing runs at $12,000 a month, I might be better off paying somebody to do it." Yeah, I think the some of the latest models the prices after going down for a while are starting to go up. >> Absolutely. Y um let's talk a little bit about Agentic technology, Agentic AI. Um I'd love to hear your perspective just on um first of all define what an agent is if you could and then I'm curious to hear if these things actually exist. How far are your clients in sort of putting them into action or are we still just at this like time where it's just talked about but not a real thing? >> Yeah. So, um, as you mentioned, I I do tech media and telco. I'm going to use media as my example for what an agent is. In the movie business, actors have agents, right? And what do the agents do for them? The agents job is to make their actor employable, right? They're going to go find scripts for them. They're going to go get them employed. They're going to go find other deals for them to do. it's advertisements for cologne or cars or whatever it might be, right? Um, you've given them the authority to act on your behalf in a specific area, right? Some level of agency, you know, for you. Take that concept of agency, apply it to technology. You're telling a piece of technology, in this case, almost like, you know, SAS in a box, like this little mini this little mini person who's sitting there. You're giving it authority to go do certain things for you. In this case, it's going to operate within a, you know, relatively closed environment. Perhaps it's on your desktop or perhaps it's within your, you know, your platform, application, ecosystem, something like that. But you're giving it the authority to operate in there. And you're hopefully giving it rules to operate under and you're also giving it skills. Skills that might tell it like where to get information or how to do certain things. And then of course, you're building security around it, too. I love all the stories about agentics, whether it's the claw bots or just the pure agentics that people are building on these open models that are funny stories, right? like I told my agent to read my emails for me and then when it got done reading them it deleted them because it was done reading them. Um and of course there's the you know the the mult story of starting the religion and everything else. But the basic premise is you know I have something that can do a lot of tasks that otherwise I could do and I don't really want to. Um and I think that last part is really an interesting one because it's how far do you take the concept of it's a task I could do but I don't really want to. You start applying that to enterprise. what's happening today and I have I have I think eight agents that I run on a regular basis like on a daily basis that do I'll call them random but important tasks for me. So every morning I have one that runs a headline review and I have it look at things around regulatory tax the markets and broad economy globally and it runs a list for me including citations of everything that's going on and it fires into my inbox at 512 a.m. >> Where does it live? I mean where is what technology are you using to do that? Okay, so that one actually lives on Grock. >> Okay. >> And the reason it does because I have it scouring X to go find that information for me because all the news providers post everything they're doing to X, interestingly enough. So it's a single source that they can a single pane that they can go to, right? And I have limitations on it like don't pick up anybody's posts because some posts are not news, their opinion, right? So I don't want opinion. I just want to know what's happening, right? And it's not reading my feed. It's reading all feeds. It's looking for everything that's out there. Um and then I actually take it. So here's the fun part of what I do with it because I then curate that and I send it to about a thousand partners within our firm. I take it and I run it through Gemini and I say hey is there anything missing in here that you see in the headlines course because Google is a great accumulator of news as well functionally. It also I think it more slightly more cleanly gives some of the links and some of the things that it did. So I say tell me am I missing anything? And then I ask it to like clean up links to do a few other things to format it for me. And then I take that and I throw it into funny enough I use a third technology. I throw it in Teams and and you know fire it out over the over the universe to all my partners. Is that sorry where do you set up that workflow? Uh it's all set up on my iPad. >> But like are you you're not manually taking the out are you manually taking the output from the email and then dropping it into Gemini? >> The only thing I have to manually do is I have to take it from Gemini and put it into Teams because from a security feature I don't want my Agentic to have the ability to just write Teams messages anywhere. >> Okay, got it. But you are able to orchestrate the multi- aent >> correct >> workflow with >> a technology. That's right. What technology do you use to do that? >> It's all I have it all sitting uh within within the the you know actually so it's the third one. I have it all sitting within the Microsoft environment. So it has that there too. Yeah. >> Yeah. It sort of goes to like I mean just briefly like the we were talking before about who's going to win. >> Yeah. >> There's pieces that each one can do. >> Um all right. So then when it comes to to clients, right? >> Yeah. Um, I'd love to hear just your perspective on like is there an ideal use case in enterprise because we hear um we hear anthropic and we hear open AI talking about how they want to start building for enterprise and you're at the center of that. Is there an ideal use case in enterprise that a company could come in and say we want to go full agentic and they can and it can actually change the way that they operate or do business. Does that exist now? And and what could it look like? Well, what how has it worked most successfully for you? I think the three the the top three areas I'm seeing right now that the back office specifically around finance okay and a lot of the source to pay procure to payroll those types of functionalities um that you can right now build aentics that they're interoperating with your ERP and interoperating with some of your customuilt tools to do a lot of the things that historically you had people doing right you're you're checking the you know the actuals to acrruels you're providing updates you're building dashboards all those types of things agents simply running 90 to 100% of that right doing that that functionality So that's one. Um the second area I see it is in the front office. So specifically around marketing and sales campaigns, the actual building of customized campaigns for customers. Instead of having you know 30 50 60 person teams that are doing this in your B2B space in particular and supporting your sales teams, you actually have identification of it. So it's I'm going in I'm talking to automotive company. We want to sell them on you know our you know latest and greatest software. Help me write this pitch. And so you know you have basically identification that's going in it's looking at all the product functionality it knows everything about the company knows everything about the pitches you've historically done it's updating that it can even do pricing right so the ability to do some of the pricing as well that's coming through so those that that's a second one and then the third one I think is a little bit well it's in every company right but it's also its own industry and that's in the legal space so I think you see a lot of things happening around legal um where the ability to help you know do research and draft and maybe even sort of do some intakes on information that's coming through and summarizing that that's more it might be more basic in some ways because we've seen a lot of tool sets that do that do research but I think that space in particular highly disrupted reviewing contracts and maybe almost a procurement functionality as well like find the discrepancies in these contracts I mean think about how much time people spend redlining things you know whether you're using docs or word or whatever you don't have to do that like you can absolutely have an agent do that and then you have it summarize it for you and then you're not spending time scrolling through and reading everything right, scroll on the iPad, whatever it might be, it's right there for you. Here's the top, you know, the top five things. And because it's natural language, tends to be pretty darn accurate. Those are actually some of the best use cases I've seen those those three areas from enterprise in particular. >> So, I I heard someone uh today refer to one of the latest models and when they were unable to access it um or this is sort of I heard this as a from secondhand when they were unable to access it, they said they they complained and said part of their brain uh it feels like part of their brain is missing. Um, so then what does the nature I mean not to get too deep into this, but I'd love to hear your perspective on what the nature of work uh looks like in in this case. >> No, I think it's a great question. I think there there's absolutely a philosophical dynamic here. Um, and I I contend with that. I have I have three, you know, call youngish kids or a couple teenagers and and one one almost there. And I think every day about what are they going to do, right? So, it's very personal in some sense of like what kind of jobs will my children do when they get older? um you know this whole concept of the the work that we've done you know really in the last you know 20 years or so um it's all net new right so and if you think back if you go in I made a historical comment earlier on is rel to technology like you could talk about like humanity going back in time like what was work and what was satisfying work um you know the reality is the space that we operate in today around technology did not exist right just simply did not exist 40 years ago and then you go through this huge spike in human capital being spent on a given thing and then it goes away. We've seen this cycle, right? We've seen this in agriculture. We've seen this in industrialization and manufacturing. We've seen this in other places. Um the the odds that this goes and follows a similar trend is very high, right? I don't think as we live through it, it will feel like this. I think it'll feel like a much more natural curve. It's not going to be some like massive moment of oh my gosh, there it happened and all the jobs went away. I think actually in fact you know recording this on a Thursday today the payroll data and unemployment data came out it was in it was in trend like with what they expected as far as job losses for the month. I look at that and say hey that's a sign that we're not going anywhere like super terrible as it relates or draconian as it relates to the economy as a whole which as a separate matter confuses the heck out of the the macro folks and and uh you know perhaps you know what we do around monetary policy but maybe that's a separate issue. Um, and I think this this natural cycle though, the ethereal concept of like how much are are going to be, you know, how much these tasks are going to be automated, I think we just have to take this as it goes and just note that for everything that is, you know, likely quote destroyed by disruption, there's a natural retraining or gravitation to a new new like a new job. And I I saw a report yesterday, you know, I think it was uh um a couple of construction companies were saying that they are roughly 500,000 people short of workers to build the requisite data centers and some of the infrastructure is necessary. You have 500,000, right? That's a lot of people who are simply not in the job market at all right now. Um and the reality is too, you have a lot of people in those jobs who are going to make a lot more money than if you were working in tech or in tech sales or, you know, some of these other things. So I I I don't I I don't say this in a like, oh my gosh, we have to embrace everything that happens around technology. I am equally, you know, nervous or skeptical, so to say, but I also look at it and say what we're experiencing as far as disruption is totally okay. It's happened before. The good news is we're highly adaptable creatures. Like we will figure this out. I heard I think it was Sam Alman say that like humans have people have started to talk to him about this. We have a right to adversity. I I don't know how I feel about that. Like would it be better to live in a world without any adversity? And is that something we want to suffer through? I guess you can get growth through adversity, but that's definitely one that I've been thinking about and I'm just like I I don't know. I mean, we we're clearly not close to a world without adversity. Yeah. >> But I don't I don't fully understand the people who are like, "Please give me some of it." >> No, I think I I'll take a different different view on that. Um and that is to say like and I love um I love uh some of the quotes from like Jaco Willink, right? Um it's like something went bad, good, right? It's an opportunity to learn. It's an opportunity to do better next time. Um I take that and apply it to you know to to the world we're facing here like you know the old quote may you live in interesting times right like I think we are living in probably the most interesting times >> no doubt >> which is so fascinating right because you could have said that you know in the 1960s and '7s some of the things that were going on society you could have said that certainly in the 1940s or World War II I mean these these are you substantially more dramatic societal disruptions that were occurring there than what's happening here we live in the microcosm where it's like having the most impact I think and so it feels the most dramatic in a way, but like in some ways I think I I I won't I won't say I subscribe, you know, perfectly to to you know, Sam's commentary, but boy, isn't it fun to be challenged? So, it's fun to be challenged to then meet that challenge and, you know, exceed maybe what your expectations are. And that's not because I'm a particularly optimistic person, right? Like deep down inside I'm probably a pessimist, but at the same time, like I love fighting past that. It's almost like a fun thing that you have to like get through. And that's part of I think that's part of like the challenge of being a human in a way. Um, again, not to get super philosophical, but like this is fun because it's hard. I mean, even going back to the conversation we were having earlier on how do you make a decision and what you should do? Like these are really hard decisions. Like we have so many decision makers even within my own organization and then we're meeting with, you know, so many different enterprises who are have not, you know, entirely differentiated pitches around what we're going to do and it's hard to pick winners and losers. And so, you know, you do as best a job you can and you try to influence it in the best way you can. You walk away from it and say, "Hey, did I do the best job that I could have in that?" I think so. And then you move on and you do the best thing. So, I say good except challenge. Um, I don't want it to be terrible. Like, nobody wants terrible, but I don't view this as terrible. I view this as like just, you know, sort of an evolution in a way. >> Yeah. And it is interesting because I think people hear about the capabilities and they're just like straight shot to AGI and we won't have work in two years. Um and on one hand like the capabilities are advancing to the point where AGI is something that doesn't sound crazy anymore, >> right? >> Um but on the other hand, and this is something that I think you see, change management is really tough. >> Really tough. It's it's it's actually the hardest part, right? Like changing on on on the behaviors front. Um it's interesting to see the people who are most receptive to change. I I um I told somebody the story last night, but I'll just retell. It's quick quick story. My dad calls me the other day. I said, 'Hey, how's it going? He's he's uh shockingly turning 80 years old. He's in great shape. He's like fixing up a house. He's flipping it just for fun because it's like what he's doing. He's been retired for a while. Um and he's fixing the downstairs bathroom and he's done all the work himself. He's done all the counters and the tile and he's the last thing is to set the toilet. Okay. And he apparently has a 1% toilet which has the wrong offset for every other toilet that's ever been built, right? as far as the way that the pipe the pipes were run. And so he said how frustrating it was. He finally had to call a plumber and the plumber came in and they fixed, you know, made a workaround and blah blah the whole way through. I said, "Well, dad, I mean, you know, not I wasn't trying to talk down to him, but I said, "Didn't you like check with like some technology like Google it perhaps or look at some YouTube?" And he's like, "Hey, son, like who do you think you're talking to?" I asked Enterprise GPT what I should be doing and I asked him for examples and I went to YouTube and I watched the videos. He's like I couldn't find the answer in any of those places. That was a place where only the plumber who I, you know, could call and talk to knew exactly what to do cuz they were a specialist in that area. Which maybe gives you hope on a couple of fronts as it relates to like what jobs will exist in the future. That's one part. And then the other part of it is it goes to show that people are highly adaptive. This is a guy who was born in 1946 saying, "Hey, the first thing I did was I went to AI and I asked AI like,"How do I fix this problem?" Unbelievable. So, if you think about like our transition, right? Like our transition should actually be a whole lot easier than that. It's funny to just maybe have more personalities and dynamics involved. Maybe that's the bit that uh that is the hangup. But I like I like that story because it's one of those funny things here. I think I'm a smart guy and I'm like telling him what to do and he's like I'm so way far ahead of you, it's like not even funny. >> Three technologies. >> Yeah. So, you know, the the grasshopper still has not snatched the pebble from the hand in that case. >> That is pretty cool. That's such a cool story. And you're right, like once you start using the AI, then all of a sudden it it can become it's not perfect, but it can become less scary and quite useful. >> Y >> All right, let's end with this. uh there is something uh called shadow AI which is um when you know there's going to be prescribed uses from a company uh of artificial intelligence and then I think stvers uh who see the potential they don't like the boundaries that they have uh and then they go do things on their own maybe on their own cloud subscription for instance >> uh as long as they're as long as they're like not including customer data >> uh in their own instances. I sort of I not even sort of I celebrate these people. Yes. Uh I think that they can be incredibly impactful in an organization. Um and we might even see a divergence where you're going to have the shadow AI folks rise to the top of companies where everyone who's following the rules so to speak trails behind. So just talk a little bit about that dynamic because you know you're again working within PWC and with so many companies that have to be careful about their AI implementation. So I imagine that you have like a real perspective on where this goes. >> Yeah. Um, you know, it's funny like it it just just the the tea up there kind of reminded me of of a a very old quote um which is if you know the way broadly, you see it in all things, right? Um and it's Japanese philosophy, samurai philosophy. Yeah. Um, and what's interesting about that is I I look at these people to your exact point. The ones who are using this stuff to do really cool things, they are the ones who see the way. The rest of us who aren't doing that kind of holding them back. Now, the reason that they are held back is a function of things that people think are more important, right? security cost, token cost, um, you know, regulatory dynamics, um, you know, certainly not being disruptive, right? Like, oh, like that person over there, they're doing their own thing. Like, why are they doing their own thing? But I had someone explain to me on Monday. It was actually really funny. Um, one of our partners, Khalil, he was telling about something he built in Claude. Um, you know, some some functionality built he trained himself on it. um he was writing all the all the all the in input Python code himself and then he was taking bringing into Claude and he was saying I want to automate the process that I'm currently you know doing with my team um and I want to take a bunch of these you know skill sets that individuals on my team have and I want to identify them as much as possible and then I want to create like an environment where I can basically load tasks that generally speaking my team has to do and the output will be frankly the thing that I get from them to review afterwards. So that was his goal. He said it was five weeks of nights and weekends, right? So like I was joking earlier, not building a company overnight. You're probably not building a company overnight. So five weeks >> just a few weekends. >> That's right. But he's a smart guy and you know he was doing it all himself which is pretty amazing in addition to that like having a full-time job and a wife and kids. Um and he said in the end he said the output product is about 99% right. He's like I would still review it and I'd be happy to review it. really really good. That's the type of personality that's going to change the world. It's going to change the way professional services operate. It's going to change the way software companies operate. It's going to change the way other industries do things. Right? We see this a little bit in health. We talked about that earlier today. Whether it's health research or the things are happening financial services and the customer experience journey. Um certainly you know in the case of industrial and you know how managing the manufacturing process and supply chain and the like you can build agentics that do these things. It's not easy. You need one or two people. You need the explorer. You need that person, the entrepreneur, the solo, you know, the the solo uh uh venturer, if you will, who's going to go there and build the thing that shows everybody that there is a way through in an industry that's old. Our company is 175 years old, right? And here's somebody who's doing something completely different than it's ever been done before. That's incredible. That's a person who sees the way. That's a person that we will follow. And so is that creating some level of risk or shadow you know shadow IT AI you know AIdriven shadow IT? Yeah a little bit. Um but you know they're all doing it within the confines of of the rules of the organization. I think it's possible to do that too. We just have to get more comfortable with some of the nomenclature being used and the how people are doing it and make sure we're encouraging it in the right way. Not everything's going to be gamification, right? Some things are just going to be um well these one guy I can't believe we've gone so far into philosophy. You've you've like you've inspired me Alex. I appreciate that. you know um Confucious water finds its natural level. The way we're going to find the people who are the most successful is they're the ones who are going to lean into this right they will find that natural water level and every challenge right is met by people who are willing you know to accept it and to overperform against the same and that's I think what we're going to find in this new era of of AI and so you know that I encourage that within the rules so to say I encourage people to push the limits right because that's actually where progress is made >> great note to leave it on Dallas thanks so much for coming on the show >> thank you Alex everybody Everybody, thank you so much for watching and we'll be back here soon.