Episode Transcript
[00:00:00] Good morning, everyone. I'm speaking to you from New York City this week where we have a company meeting.
[00:00:06] And it's a beautiful, cold, clear day.
[00:00:10] And I want to give you some clarity about something that I learned and have been discovering around the world this last few months, and that is confusion. We had a big workshop yesterday with a bunch of companies on the AI blueprint we're working on and met a lot of chros. And what I explained to them is three things that I want to run through this morning for you, and that is everyone is confused about AI, and there's good reasons for this.
[00:00:41] So I think this will hopefully make you feel a little bit better. The first is, as corporate HR leaders, we are in some sense expected to suddenly adapt and adopt this new technology to reduce costs, to improve efficiencies, to get with the program, to be contemporary, to be cool, to adapt to what the company's trying to do. Fine, that's okay. But I want to make a point. We're not adopting AI for the sake of adopting AI. And this is a big point I was making to the folks I was with in Singapore. This is not an AI transformation. This is a process of learning to use AI to solve problems. And there are many, many, many problems we can go after. We can reduce routine work and simplify jobs if that's worth the time. We can improve hiring and skills, we can streamline onboarding, we can improve sales effectiveness, we can build better leaders, et cetera. I was sitting next to one of the chros of a large financial services company yesterday, and she told me that one of their initiatives they've and struggling with is building a better leadership platform for leaders. And we talked about Galileo for managers and how they could build dynamic learning experiences and support programs integrated with performance and all sorts of other management tasks in an agent. And she just jumped up and said, wow, that's what we want to do. So it wasn't about the AI, it was about the problem. And so rather than thinking of AI as a project to implement AI, which is the old way of thinking about hcm, by the way, don't think about it that way. This is a technology that's capable of doing many, many things. So your role here or job is to understand what it can do and start to figure out where to prioritize these applications, now you can do that from the top down or the bottoms up. I think you'll find doing it from the bottoms up is much more effective because there are so many applications for this. And I keep going Back to this idea of the Microsoft spreadsheet. When the PC was launched in the 1980s, we didn't have systems we could type data into. We didn't have systems that we could program ourselves. We didn't have systems that were interactive before the PC. So when we got our hands on the PC, we suddenly discovered all sorts of things we could do with it, but we had to know how to use it. And I laughed about this multiple times with many of you. Many of the executives during the early 80s, when the PC came out that we met with, didn't want to use the PC. They wanted their secretaries to use it. Well, this is the same situation. You use it, you get people to use it and you discuss what are the best projects to work on. And that's where the agent, super agent AI blueprint comes in. Because we can show you exactly what these applications look like and how to turn them into level one, level two, level three, level four level solutions. And the vendors are getting there, but the vendors are not there yet. So this isn't a matter of off the shelf stuff. This is a matter of problem solving, prioritization, and exploring the technology.
[00:03:52] Okay? So that, I hope, will fix or help with that problem. Second problem, the vendor market.
[00:04:00] Every vendor, including us, is learning about this technology and what IT can do and the impact on its business. The startups, of course, are starting from scratch. So they're going to come to you with agents and chatbots and tools and coaching tools and development tools and all sorts of things that look great. And if they're small companies, they probably have prototypes that are working immediately and you're going to be enamored by them. They're very fascinating tools, but you have to think about the architecture. Do you want to have 20 vendors implementing 20 different tools or a hundred different tools in your architecture? Will your IT department tolerate that and will IT work and will you be able to coordinate the data? Probably not for a long time. And if you remember, this is what happened in talent management when we had a bunch of performance management vendors and a bunch of onboarding vendors and a bunch of ATS vendors and a bunch of other things. So take a look at them, experiment with them, learn about them, but remember that those are early stage point solutions. Many of those companies will either go out of business or be acquired. And you need some sort of an architecture. We have a good sense of how to do that. Your IT department's going to want to be involved, so that's sort of them. Second category of vendors, the big ones Workday, SAP, Oracle, Ceridian, now Dayforce, adp, whoever you may have as your payroll provider. They're all building agents. The first agent they're mostly likely to build is a self service agent. But remember that self service, which is a big, big, big use case, is not about hcm. It's much bigger than that because the self service for HR has to merge with the self service for everything else. We're working on a project to integrate Galileo with a company called Lena AI that has integrations to all of the major HCM systems and other systems that we use in our companies. And you'll see what this is like. And I think they're kind of a groundbreaking company. They're about 10 years old, but they've been doing a long time. And you'll see this when you come to some of these conferences when we introduce it. But that's what you have to envision, the self service for everything, not just hr. I'm not saying that the self service tools and the ATM vendors won't be good, but they won't get you everything you need unless you think about them architecturally. Second thing they're going to show you is all sorts of embedded agents in their HCM applications, which are great, but they're mostly type one and tape type two apps. They don't transform your workflows because they work within the workflows you already have. And you know, I'm pretty convinced that the real benefit of AI is not doing tasks faster or eliminating tasks, but re engineering how we do things as a whole. Task automation is not turning out to be a high ROI in if you don't reorganize jobs and reorganize work, and if you insert task automation tools to your ATM system, you might inadvertently miss the opportunity to rethink how you do things and eliminate steps. Anyway, they each have their own strategies. Lots of the stuff they're working on is great, but it's coming along. And remember that they are under a lot of pressure. A lot of pressure. You saw the stocks of Workday Salesforce, all of the major enterprise software companies plummeted in the last two weeks. The market has pretty much spoken that a new generation of these kinds of systems is needed. And some of these guys might make it there and some may not. Incidentally, I'm sure a lot of you know this, but Anil Bashuri, founder of Workday came back. He was sort of stepped away, turned over the CEO role to someone else and recently, this week, just came back. He clearly wants to be involved in the next chapter of that company. And that's what's going on in all of these bigger software companies. So the confusion for you as a buyer is no less than the confusion for them as a vendor. So it's okay. This is just where we are. But it's your chance to get to know them and what they're working on and share your direction with their direction as much as you can. Number three, the process of changing jobs, changing roles in reorganizing work. Now, we've done a lot of work on this last year. We spent almost nine months talking to about 50 companies about it. And in the early stages of this, what we thought was going to happen is you were going to sit down with your job architecture and you are going to buy a tool and analyze all the jobs and tasks that people are doing, and then you are going to push a button and implement the Microsoft Copilot or some other tool and instantly reduce all that work. Well, I mean, that sounds great, but it's really much more complicated. You've really got to sit down and workshop this out to figure out what to remove.
[00:08:57] Two examples. I mean, Microsoft are doing a lot of work with Microsoft. You're going to have Galileo on the copilot pretty soon here. They have been doing this for a while. We're working with them on a bit of it, and they're using Galileo for a lot of it.
[00:09:11] Every time they try to automate a process, they find out there's more than 70 versions of the process because the company is so big. And if you work for a big company, that's not unusual. You have locations, geographies, business units, use cases that just aren't clear from the org chart of which the knowledge is embedded into the people in your HR department that are doing them. So better to work from the ground up than to sit from the top down. And assume that by simplifying your job architecture, you're going to simplify the work. The job architecture will follow. I think the way this is going to work is your large, complex, highly frustrating job architecture will have to adapt to the new flow of work. It's work architecture, not job architecture. If you fix the work architecture, the job architecture will follow.
[00:10:03] So it's not as maybe simple as you thought, but in some sense it's easier than you thought because you have the people in the company that know how things get done. And we use the philosophy of falling in love with the problem. Every workflow or every process or every org you have has goals and Think about those goals from a customer standpoint, whoever the customer may be, and be clear on them. And then go back and look at how you're doing it today and question why you're doing the things you're doing now when many of them could be automated by AI. To give you an example, I've talked about this before, but I'll give it to you briefly. In our case, for 25 years we've been doing research in a fairly traditional way which involved hypotheses, data scans, surveys, interviews, et cetera, et cetera, et cetera. We still do that, but we're going to do a lot less of it because we can do research much faster now. Much faster. We can interview dozens of you over the phone or over Zoom and we can take those interviews and we can throw them into Galileo and we can identify trends very fast. I mean, we don't have to wait for a survey. Not saying we won't do any surveys, we'll do some, but we're not going to do as many. And pretty soon we're going to be doing surveys verbally with the AI, which we're experimenting with now. All of your HR practices are filled with things like this, where you can really rethink how you do things. Okay, so. So there's a lot of confusion about that, but it's actually clearer than you think. And if you read the work we did on dynamic work design, we have a whole research report on it, you'll understand it better. And in fact, Galileo can show you how to do this because it's trained on the process. Number four, change. Now, one of the companies was with us this week was IBM. We also had actually quite a few large companies and the IBM folks made a really interesting comment.
[00:11:56] I really love IBM, by the way, and we'll have to be doing more with them. But one of the things they were talking about was that at IBM in the past, if you did a project and it wasn't perfect, it's a career limiting experience. When I worked at IBM a long time ago, it was like this too. Such a big company and it has such a legacy and it has such a reputation that the internal projects really needed to be perfect. And I remember when I worked there long ago, things were rolled out in an extraordinarily high quality way. The problem with that, of course, is it takes a long, long time to make things perfect. And they never really are perfect, even though they look like they are. But it slows things down a lot. And last year or so they replaced Workday with SAP, with SuccessFactors for a bunch of reasons. And the mandate from, and that could be, you know, like a five year project easily.
[00:12:50] And they basically said, we're gonna do it in 18 months and we're gonna turn it on even though everything won't be ready. And Arvind, the new CEO, has pushed a new culture in the company that going forward, speed is more important than perfection. And the way they described it is there's three parts to their process. Eliminate, simplify, automate, eliminate as much as you can, simplify the process that's left and automate next. And I think that's a really good philosophy. And if you do that, it's not going to be perfect day one. But after all, you got to remember that AI technology is not done yet. It's early, early days. So whatever you do, relative AI, the tools you're using are changing under your feet. So as I said in the luncheon we had, I think you're going to end up with solutions that are 80% working, 90%, but you know, 10% of the time they're not going to work. And you're going to have to learn with the providers or on your own how to make them 92%, 93%, 94% correct. Because if you try to get all that done before you roll it out, you won't even know if it's ready. So that's the next confusion is don't wait for perfection.
[00:14:00] Probably impossible to find it. We're just not at that stage. This is a different type of technology. It isn't button down and 100% deterministic. In fact, it's not deterministic because the technology's behavior is dependent on the data that you give it. So regardless of how well organized your data may be, you don't really know how it's going to behave until you use it. And we can tell you all sorts of stories about our situation with this too, but you will learn how to use it fairly quickly. I know that because we know how Galileo behaves. We've just used it a lot. So we know where the rough edges are and we know what to do to keep it working correctly. Okay, there's that. Now there's the issue of culture of the employees.
[00:14:48] And we talked about this a lot and I know all of you are aware of it.
[00:14:52] We're in a situation here, this particular period of time where the economy is pretty good, not great, it's definitely slowing, but there's no recession. Hiring has slowed significantly for a Lot of budgetary reasons and uncertainties and tariffs and other reasons and people are really nervous about their jobs. You see it in the data, you see it in the surveys, you see it in the consumer sentiment. The standard of living of most people is flat to declining because of inflation.
[00:15:22] And all of this bombastic talk by AI techies about jobs going away is scaring people.
[00:15:32] I have no evidence of any kind that we're going to run out of jobs. There's going to be so many new jobs from this AI stuff that we're going to not have enough people. But things are going to be more productive. We're going to grow our companies faster, we're going to have more revenue per employee, et cetera. There's lots of good things that will come out of this and we're not going to run out of jobs. But we had an employee leave our company the other day who was a graphic designer because he didn't want to learn how to use AI. And there will be people, instructional designers for example, or people doing clerical work who find out that the thing they were doing or the job title they had did go away and lo and behold, they got to do something new. Is that so terrible? No, it's great. It's a good opportunity. So if you frame it that way and you explain it and communicate it that way and stop talking about layoffs as much as you can, you're going to find a significant number of people in your company are ready to go, they're ready to be retrained, they're ready to try something new, they're ready to learn a new role or a new job. And you're going to have lots of interesting experience moving forward.
[00:16:38] Of course there will be some layoffs. And I think one of the reasons a lot of companies have been laying off goes back to this issue of what I call talent density. We basically ran our companies as talent supply chains instead of skills based organizations. And so we tended to hire people based on linear scale. So we overhired, we said we need twice as much revenue, so let's hire twice as many salespeople, we need twice as many products, so let's hire twice as many engineers. And that's just not the way to grow a company anymore. You have to grow exponentially through technology, through automation, through productivity improvement, through skills. You may not need as many people, you may not need as many headcount. This idea that your salary or your role or your position is based is valued on by the number of people that report to you is going to have to sort of disappear. We're going to have super worker people that are doing the work of five people. In fact, one of the most interesting things I think we're going to run into here is wide disparities in performance per person. Gonna write more about this. I have a kind of a big piece I'm working on this. But you're gonna have some people that are extremely productive with AI and you're gonna have to pay them more, which is fine. You have to sort of break some of the bonds, you have to salary bans and performance reviews and stuff. But that'll be good for a lot of people. And that whole dialogue is important to get people on board with the changes that are happening. You know when you read about Amazon laying off 20,000 people and everybody's sort of in a panic about it. Amazon has a million employees.
[00:18:15] 20,000 people is not that many. And who knows why they laid them off. I don't know, maybe a distribution center moved from one city to another or they got a new kind of a truck that could handle more capacity. There's all sorts of reasons for that. And then there was a big freak out last week that the new version of CLAUDE was going to eliminate all the work done by financial analysts. We've seen this before. Financial analysts will now use the new version of Claude and they'll do more financial analysis and more sophisticated things for you.
[00:18:48] I have no worries about my job going away.
[00:18:51] I just think I'm going to be able to do more and better and more interesting things at a higher speed with these tools. And I think most of you probably feel the same way.
[00:19:00] But you have to communicate this and you have to make it clear from the CEO down that we're not implementing AI to reduce headcount. That is keeps coming up. But I don't think that's really where this is going. That's a very temporary phenomenon. And yes, we want companies to be more efficient, but ideally you want people to re engineer their work around these new tools. And the people that don't want to transform can leave. And there will be some, there will be some people that just don't want to change because they may have to change roles or location or managers and for some reason they just might not see it in their best interest. And I think some of that will be voluntary. So that's that topic. I guess the final thing I would talk about that came up in these meetings, there's many, many other things is trust. I do get in a lot of these meetings Questions from people about the future of the world and can we trust these things and how do we make them compliant and how do we train them, et cetera. Well, you know, as an engineering type person, I think you just have to be aware. These are algorithms. They're not human, they're not alive. They don't have genes and cells. They don't learn like we do. They learn through data. You train them, quote, unquote, by giving them data and labeling the data by humans. A lot of the training of these systems comes from humans who have told them and labeled them about what some of this data means. So if you take the time to govern the data well and make sure you know who's responsible for what, and you keep it up to date and you deal with issues like data quality, your AI systems are going to work and you are going to be able to trust them. Now, I'm not sure about what to say about the consumer world because we have these wacky things coming out of OpenAI about personalities and porn and strange things like sora, et cetera. Those are different business models. But inside of a company, you have a lot of control over the quality or trust of these systems, much more than you think. They're not going to jump out of the computer and bite you, but you're going to have to learn how to do this. I know what we've been through. It hasn't been as hard as we thought, but there were some glitches along the way. And you'll find tools coming out of the market to do this. There's lots of tools being introduced to evaluate the trustworthiness of an, of an AI. I don't know how good they are, but I'm going to talk to some of the vendors. A bunch of them are calling me, and if any of you have information on that, please reach out to us.
[00:21:32] So trust is going to be a manageable problem. It's not out of your control.
[00:21:38] The vendors who make the LLMs are not trying to make things difficult for us. They're just building something that can adapt quickly to the data that we give it. So the answer to all this is try to clear the dust of confusion. Now, I think maybe the most tricky part of this is your incumbent technologies. Every company I talk to has a hundred different tools already, and every one of them was developed far before, long before ChatGPT came to market. So the next couple of years, there's going to be a lot of debate, discussion, vendor analysis, ma disruption in the market.
[00:22:20] So in the middle of these confusing issues. Stay focused on your problems. What are the business imperatives of your company? What are the areas of performance improvement, profit growth, cost savings, productivity where you see big opportunities? Get your teams together and prioritize on those things because they're all complex, none of them are simple.
[00:22:43] And we have a tool set now that's more flexible. And by the way, the only thing I left out of this podcast was your ability to build things. These are tools you can build with. You don't have to wait for vendors to build all these applications. Find some of the people in your company that like to do hacking. They don't have to be software engineers that like to hack things together and you'll be amazed what people can build. It's a new world and we will do everything in our power to give you clarity here. Okay, that's it for now. Talk to you guys later.
