Episode Transcript
[00:00:00] Hey everybody. I'm just back from Europe and I want to give you a recap of what I went through and visited and learned about and spoke about at the Unleashed conference in Paris this week. I probably had 30 to 40 client discussions and meetings, gave a keynote, met with dozens and dozens of vendors. We launched our new book, Super Powered. A couple hundred people waited in line and got a signed version. We also did the same thing at the HR Tech conference in Vegas. And what I talked about in the keynote is sort of the, the whole story together as one big issue of economics, politics, technology use cases, implementation maturity and the future of hr.
[00:00:40] So I won't do the whole thing.
[00:00:42] I don't want to give you a two hour podcast, I'm a little too tired. But let me give you the high level since it's fresh in my mind. So taking a step back for a minute, we're in the sixth or seventh year after the pandemic, the fourth year since the launch of ChatGPT, and the world is embarking on a $10 trillion build out of AI.
[00:01:07] It's amazing how fast it's happened. It's resulted in a whole stack of technology behind these tools, including the chipsets, the memory, the interconnect, the compute, the data center, the cooling, the electricity, the power, the real estate, the LLMs and those related software tools, the management tools that sit on top of them, the harnesses that sit on top of that, and then the applications that you buy from other vendors and you don't have to use the Frontier LLMs as you guys know and at every state stage of that stack, there are different competitive issues in different vendors.
[00:01:48] The vendors that build the whole data centers are cloud computing and data infrastructure vendors. They're called Neo Clouds and they have different names. They are taking out massive amounts of debt to build in a sense, the railroad tracks of the future.
[00:02:04] And I'm reading a book, I talked about this at the keynote called 1873 that I really recommend you read. And it's about the build out of the railroads and the role of the Rothschilds and other investors and how the world got so excited about the railroad industry that there was quite a bit of graft and corruption and debt was taken out and some countries went very deeply into debt to build out the railroads because the railroads were expected to be the future of all commerce in the world and about 50% of the railroads that were built went nowhere. But there was a huge recession, huge because, because the bond market was so over subscribed for this debt and it turned out that a lot of these railroad companies weren't even real. There's this, by the way, this is a complete parallel to where we are. And there was a war, the Ottoman Empire collapsed. A lot of really bad stuff happened. And it's a fascinating book because it's about the investors, but it's really about the whole situation. And believe it or not, I really think we're going through this exact scenario right now, and we're in the frothy, exciting stage of it. We just haven't hit the second downside yet, the correction, the bubble, whatever you want to call it. So what it means for us in corporations is that the cost of AI today is extraordinarily high, much higher than it will be in the future. We haven't really optimized it. The engineering of the chipsets relatively new. There's lots of room for optimization in the software.
[00:03:38] And one company, Nvidia, is taking almost 50 to 60% of the dollars that are spent now. It's interesting, Jensen Huang is such a pleasant, positive person, the CEO of Nvidia, that he's in some ways become the godfather of AI. And he's talking about it and he's showing it off and giving us all these grandiose stories as to how it's going to be, the new industrial revolution and all that. I have no question in my mind that it will be a huge market. And in the IT market alone, just corporate, which is about a $7 trillion market, Gartner believes it'll be over a billion within the next year or two. And it's growing at one of the fastest rates it's ever been. Will it take over and transform the whole economy? No, that's just a little bit absurd. But it's going to affect all of us in some way.
[00:04:29] And right now he has captured 60 to 70% of the cost in the compute part of it, which then, you know, sort of stacks up. So that number of what we pay per token is going to go down over time. Now, of course, while that's going on, we have this weird inflation all over the world caused by the post pandemic supply chain issues, but primarily now caused by wars and the energy crisis caused by the war in Iran. So it's a strange combination of occurrences that the United States has decided to go after Iran and create an energy crisis at the same time. We want to build out data centers. I'm not going to make any comments about the federal government at the moment, but they're not thinking very clearly. Obviously and those high costs and prices are resulting in a lot of unhappiness and uncertainty by the middle income and lower workers, including the riots that took place in France when I was there, which is to some degree students, but also just young people and people that can't afford to standard of living. Diesel fuel in Paris was 12 bucks a gallon, so it's about $8 here in California. So it's way worse there than it is here.
[00:05:42] So there's that, which we're all sort of assuming that's going to fix itself. I don't know, maybe it is. And then these same people that are worried about their standards of living are hearing from the technologists that their jobs are going to be eliminate. So isn't that nice?
[00:05:58] Let's create a situation where nobody can afford their food and healthcare very much or their energy in the winter and tell them that we think their jobs might go away because AI is so much more productive than they are. So we have the lowest level of employee engagement that we've had, almost as low as the pandemic. Now, you know that's not true of every employee and every company and every location, but the data is essentially showing that. And then add one more factor that because the stock market has run up so high certainly since the pandemic and we've actually not had a real true stock market correction since 2008. So that's an 18 year run up in the stock market with the exception of the pandemic, it's never gone up that long in my lifetime that the wealthier population of the working class has done great. The wage earners have not. Now in fact, Paul Krugman did some analysis and this has been done by others, that the percentage of us, at least I think it was us, I don't think it was global GDP that comes from labor has gone down and down and down and it's down to about 30%, 35. It's the lowest it's ever been. In other words, this capital investment in AI is squeezing out investment in labor, which means that the salaries that people pay are going down relative to the costs that we spend on AI. And I have a whole bunch of stuff to talk about on that topic on another podcast which I want to get into, because we're going to talk about wages. Because my thesis in Superpowered is that wages are going to go up because we're going to have a talent shortage and basically a labor shortage. We already have a labor shortage now. In fact, it's come out in some of the data that just came out today, and it doesn't matter how much AI you have, the people left in the company running the AI and managing all the agents and doing all the work are going to be leveraging more output per person, so they're going to be paid more. So there'll be some factors to consider here that have to do with the economy. So there's also all of this stuff going on in the meantime. Most HR people I talk to are still pretty confused about what to do with AI, where to apply it, how to use it for RE engineering versus productivity, and what is the right process for moving to HR 2030. So I went through that for at least 20 minutes during the speech. And what's basically going on is most companies are pretty well stocked in terms of general agents, but they're not very far along on the RE engineering and the implementation of solutions. And the vendors are coming along pretty fast. Workday now has 17 what I would call super agents. SAP has more like 60 or 70 smaller agents and a new platform called Joule that sits on top of it. Joule work. Oracle has 1,000 agents in their entire platform and is now verticalizing their HCM solutions hibob as a complete agent that handles seven or eight of the critical HR functions and has now embedded Galileo in Galileo is now embedded in work and will be embedded in ServiceNow and some other tools. ADP's done this. UKG has agents to do scheduling and resource management and optimization.
[00:09:09] Two new companies Gloat just announced an AI native HCM platform which is very cool. It also has Galileo embedded into it.
[00:09:17] Darwin Box has a new HCM that's an AI native HCM and I could keep on going. I have a whole bunch of slides on this. So you're now sitting here as a buyer or a consultant or an HR leader or an executive or an IT executive.
[00:09:32] Seeing this enormous investment pressure from Wall Street, Wall street seems to be telling companies implement AI or else, and then it's telling vendors generate AI revenue or else.
[00:09:46] It's a lot of new stuff, a lot of dynamics going on here at the same time. And so we're very sympathetic to this and a lot of what we're trying to do on our side is educate you so you can make the right decisions. I would say there's a number of things that I'll mention at the moment and I had a lot of conversations about all these things. First of all, you're not buying AI as a tool to cut costs. Just don't think that way. First of all, it probably won't cut costs because you do have to pay for it. And if you're not going out of business, cutting costs is not the way to you grow your company. It's an important thing to do at all times. But reducing the number of people you have in the company may or may not even be a good idea. Because if you use AI the way we expect you to use it, and the way we show you how to use it, it enables growth. And so if your company grows faster and has better customers experience and more intimacy with customers and better employee experience and better scheduling optimization and better resource management and better marketing and all that, you're now talking about a business where the percentage of improvement in growth that comes from AI might be 20 or 30%, but the rest of it's going to come from people. Listen to this data. In the US over the next decade, the size of the workforce is only going to grow by about 3 to 3.5%. Yet the GDP is expected to grow by 25 to 30% somewhere in that range. Where's that delta going to come from? You're going to have to generate more productivity just to stay even with GDP growth. And I know none of you want to grow at the rate of the gdp. You all want to grow at two or three times the rate of the GDP because you want to tell your shareholders that you're a good stock because they can buy bonds if they want to get GDP level growth. So we're going to have to generate productivity out of these AI tools. That's 30, 40% or more if we want to leverage them for growth. Now that is absolutely doable. I mean, we know that's doable from the projects we've worked on with companies. But we have to stop thinking about this as a cost cutting. Let's just make everybody a little more productive type of technology. The second thing that comes up a lot is there's AI fluency, there's spheres of AI, there's capabilities in using AI, and there's the architecture of the AI systems. We're not buying tools for individuals anymore. We're buying corporate enterprise systems. We're buying systems of AI or networks of AI. And please, please look at our HR 2030 materials and the research we're doing on this to get familiar with how these tools fit together. If you go out and you buy Paradox from Workday or a new learning system from Docebo, or one of these UKG tools that optimizes workforce management, or you pick what you like it has to interoperate with all the other AIs that you're going to implement in your company. And you want these AIs to talk to each other just like people. And you're going to have to manage these agents. They don't manage themselves. You're going to have to train them, onboard them, essentially give them performance management, monitor them, make sure they don't get out of control and cost you too much money. And by the way, sometimes humans are cheaper than AI. I had this actually great discussion at one of the dinners that, you know, if you have a deterministic task like computing two plus two, AI is going to be a very ineffective way, an inefficient way of computing that because it can be done by a regular computer in a microsecond and, and the AI might have to compute and sit there and think about it for a while and burn up tokens. So we have a lot of interesting management activities to do in hr, not in IT to build and implement these things. So you have to think about this family of AIs that you're building and buying as an architecture. And by the way, I do think, and I talked about this, you're going to be building a lot of these tools yourself. You might look at a vendor product for coaching, for example, and say it's great, and then realize that, you know, we have a lot of the IP to build this coaching internally. Maybe that's a product we want to build ourselves. And I'm not going to go through all the possibilities there, but that's another topic that came up as well. So we're now not only dealing with data management and security issues and the various GDPR issues and European regulations of safety and bias in AI, but really, do we have an architectural viewpoint and a set of architectural tools to scale this up instead of buying a bunch of agents from a bunch of different vendors that are likely to go out of business business or get acquired. Now, on that topic, while I'm riffing through all of the things that happened over five days, I think the big vendors are in a great position right now. I know we had the SaaS Apocalypse business and all the stocks got sort of hammered, but I think if you look at workday, SAP, hibob, adp, ukg, et cetera, the bigger companies that have larger numbers of customers, cornerstone, they're sitting on the data you need to make your AI work.
[00:14:38] So there is no reason to start over from scratch and try to build this in a new tool just because the vendor's cool when all of your data is already in one place. Now you can complement them with Claude and OpenAI and lots of tools on top of them, including the Microsoft Copilot, of course. But I think the bigger vendors are going to be the home of your AI platform and then you're going to decide strategically where to build and buy other tools as your HR 2030 implementation rolls out and matures. And each time you do an AI project, you should think about it not as an AI project, think about it as a business project. What business problem are we trying to solve? Are we trying to hire faster or are we trying to build higher quality of hire? Are we trying to improve skills? Are we trying to improve go to market? Are we dealing with sales? Are we dealing with marketing? What does our stakeholder want?
[00:15:31] The more clear you are on the problem you're trying to solve, then the project which uses AI, not the AI project, but the project or the agents can be implemented in a way that is appropriate for that problem. So what I want to try to avoid is a bunch of AI infrastructure being purchased. And then you decide, well, it wasn't quite exactly what we needed because the problem was a little different than we thought. So we're going to go back to something we talked about many times, which is falling in love with a problem, not the solution. AI is a solution that comes when applied to a problem. And I think we're at that point of maturity now where most employees roughly know what's going on here and what the stuff is and they know how to use it at their desktop, but they don't know how to do bigger transformational projects. So all of that came up in lots and lots of interesting meetings with lots of companies and there are lots of case studies, but I think we're at the very bottom of the J curve. Tell you guys more about the J curve, maybe if you haven't heard of it and maybe write an article about it. And starting to come up the other side, things are becoming more mature and the products are becoming more stable. I don't think the vendor market is stable at all. I think the vendor market is gonna go through a lot of shakeout and a lot of the tools that seem like up and comers may play out to be huge companies and some may disappear and get acquired. You've seen Workday and SAP have each spent a couple billion each on acquisitions of just HR stuff in the last 12 to 15 months. And there and then of course, Dayforce went private. UKG has been going private For a. There's lots of investment going into this and some of the vendors that were hot a few years ago and that might be hot today are likely to get acquired too. So we have a lot of resources to help you out of the HR 2030 program, our new GHRE program, which is a 12 week certificate program. We'll teach you all about this stuff and you've got hands on experience with a lot of the AI tools yourself. You'll learn how to build things and then we're just more than happy to talk to you as advisors on any of these topics. And then in Europe, of course, a lot of people wanted to talk about the riots and they weren't nearly as bad as they look like they were on tv. I mean they weren't small, but they weren't, it wasn't like the whole country's burning down. It looked like on some of the TV channels like the whole country of France was about to blow up. But their, their reaction to all of the things I just mentioned and I, I don't think you all need to be political scientists. But, but I think understanding the context of this is important because this flurry of noise that's coming out of these big frontier vendors is noise. I'm not pay attention to it. I think they're very, very shrewdly jockeying to make as much money as they possibly can when they go public in an environment where the cost of going to market is very, very high. You know, the stock market will speak on the value that these companies generate when they do go public. But it's a very uncertain economic time for this technology. Like I said, it's just like the railroads. Sometime in the next, you know, some number of years this is going to shake out and we get a much lower cost and more rational pricing model for some of these tools. You know, the, the token cost that you pay to the bigger vendors that you're doing business with are probably going to come down significantly. And they, and they know that I talked to this, the execs at workday about it and they believe we've barely even started the engineering optimization of the LLMs.
[00:18:57] So you know your, your hurdle rate or your cost justification of a project is pretty high at the moment. So you're going to have to look for some big projects. The four big, biggest return on investment projects in HR for AI if you really want to do an AI project for its own sake. And they're all high ROI projects worth doing. Number one is talent acquisition, speeding up and improving the candidate Experience the selection and screening, experience the interview experience and getting much better data and streamlining this huge mess of human processes it takes to do hiring. This is the business of companies like Paradox and Mackie People and Smart recruiters and lots of others. And they've done. They're the most mature in their AI solutions and we have lots of examples and case studies of how people have revolutionized their recruiting. The second is in the employee experience.
[00:19:47] I think all of you, I'm sure, are going to want to give your employees an agent instead of a complicated portal that takes $100 million to maintain and the agent will be able to answer their questions and get them access to the latest policy or documentation, coaching or support. That is something you're going to have to build. The tools will help you. New vendors like Emma or Sana embedded into workday ServiceNow has a whole new manager interface to this as well as an employee interface as well as an IT interface are really spectacular. And whichever platform you decide to use, that is a very high ROI project because not only does it improve the operating efficiency of all the employees and workers in your company, but it allows you to get better information from them on what their needs are and where they're having problems getting their work done and what skills they need. Which leads me to the area of learning and that is the third huge ROI area. The L&D function is a $400 billion market. It's very big. It's probably comparable inside to recruiting in terms of all the spend. And you can move to a model of dynamic enablement using dynamically generated content in AI. And we've written the book on that, that, that was a huge project we did over the last two years. And you can actually experience it in Galileo. If you buy Galileo, you get access to Galileo Learn, which is an AI generated super tutor learning system with about 700 courses in it. So if you're not even familiar with this stuff, for 79 bucks a month you can basically touch the best technology in the world and see it. The fourth area that I think is an important area to consider is redesigning the role of the business partners. Because as more and more agents automate back office special specialist functions in HR and we can show you the whole map of 160 different opportunities for this in HR 2030. You want to spend more time with the business. You want to be. You want to do workforce planning and resource management and performance optimization and talent management and succession management and all those things for the workers in the business in a more intimate way. And that fourth area you might call HR business partner optimization or HR Consulting Optimization means giving the consultants or the business partners the tools and information and the agents they need to do their job. And they won't have to do as much administrative work because the employees will use the self service. So those are the four big areas to get started. But there's lots of other ones. Global onboarding is a huge area for banks for example, which you can save a fortune. It's a big project but there's really good tools in the market from ServiceNow and Workday now for that as well as I know SAP is working on one. And then there's lots of niche ones. There's going to be AI agents for compensation and pay equity AI agents for payroll optimization which everybody wants to stop all that three weeks of work every month trying to get the payroll to reconcile and deal with all the anomalies there. So it's a really exciting time to be in HR and lots and lots of different areas to focus. But I think, you know, taking some time and thinking about this as an architecture and a strategic set of solutions as opposed to let's give everybody a copy of the Copilot and cross our fingers that they're going to come up with some cool ideas and then we'll just tell everybody what our ROI is. That was a good way to get started, but that's the top part of the J curve. We're now at the bottom part of the J curve and so it's time to get much more serious. And we also demonstrated Galileo running in hibob, Galileo running in servicenow, Galileo running in workday, Galileo running in Emma and in the Microsoft Copilot. So Galileo is a tool that is not only an AI agent but also a tool for employees and managers to do some incredible things for their teams, for their careers, for their development and leverage all of these other data sources and technologies and policies that you have in your company. Everybody I talked to over there was excited to just get some clarity about these topics and some clarity about where this is going. I'm almost at the point where I think you should stop reading what Dario Amadai and Sam Altman and Elon Musk have to say. I don't think it's useful and I think it's some form of marketing of some kind that they're doing to better their own well being and Fortune. And I don't have any. I don't mean to be negative at all but, but there's a bunch of monopolistic behavior going on here. Nobody wants to get let out of this, left out of this. I think that the more mature vendors, the SAP workday, Apple, Microsoft, they're not really trying to inflammatize this whole space at all. They're being very rational and they're building things that add value that you will want to pay for. And that's pretty much what this is about. This isn't about scaring everybody into thinking the world is going to come to a crushing end or we're going to all be living next to a power plant. And you know, right now, apparently there's 330 data centers under construction or considered under construction in the United States. I have a feeling I wouldn't invest in that industry at all. I think that stuff is going to get overbuilt, just like the railroads. And the cost that you pay for a data center today might be 10% in five to 10 years, 90% lower. I haven't done the math on that, but I'm just saying that the people that built the railroad tracks early may have sold their bonds or their railroad rights early, but if they didn't, they lost their shirts when the railroad tracks went down in price. That's kind of where I think we are. And I know I'm kind of pushing my credibility with you guys because you don't consider to me to be an economist, but I've learned enough about this that, you know, seeing the supply and demand of real applications, that's kind of where we are. But it's actually great for you, great for us, great for employees, because we are building super worker organizations. Our new book, Superpowered, goes through this. Wages will go up. There will be lots of new jobs created. There will be lots of economic growth for the businesses that implement AI. Well, we'll improve customer experiences. Our companies will be more optimized for customers and less optimized around our back office. Uh, we'll talk about that later. That's another narrative. And all of you in HR are going to be able to add more value and move to the model of HR 2030, which we call dynamic enablement for growth. In Europe, things are a little bit less frothy than they are here, but the expectations are just as high. And maybe there's a little more EU GDPR like regulation, and they're. And they would like to use technology from Europe. So a lot of you that do business in Europe are going to probably want to use Mistral or one of the vendors that is located over there and not pay all your money to Elon or somebody else that's somewhat country specific and company specific. But you've got SAP over there obviously as a great company and others a lot of really cool HR vendors coming out of Europe, including Sana who was acquired by Workday. So I think the European side of this is a little bit more pragmatic than the frenzy going on out here where I live. But it's all good. So anyway, we're going to stay on top of this. We've got a lot of architectural papers coming out and HR 2030 new implementations of Galileo for you you guys to see. And then I'll be happy to do my keynote for any of our corporate clients. So if any of you are corporate clients or you'd like to be corporate clients, we'll take you through the whole story in great detail. That's my quick review for today. I'm going to probably go to bed early tonight and more to come on wages in the next couple of days. Bye for now.
