Episode Transcript
[00:00:00] Good morning, everyone. Today I'm going to talk about what's going on in the market for AI agents, because there's a whole bunch of stuff coming next week. You're going to see two big announcements from us relative to Galileo. We're going to the Unleash conference. We're going to have about 250 people join us in workshops, teaching them how to use Galileo, showing them what it is. And then there's the huge ServiceNow user conference in Vegas, which we're speaking at, where they're going to be introducing a whole bunch of very incredible things. So let's talk about AI agents. So what is an agent? If you go back in time, when OpenAI unleashed ChatGPT, we had this technology called a large language model, now upgraded to a large reasoning model, that is essentially the intelligence engine behind a chatbot. And the way most of us used it initially was as a way to answer, ask questions, get answers, compose documents, create content, build build pictures, and generate information or analyze information from a large corpus of data. So in the core of these AI systems is this intelligence engine that can analyze and analyze information from various sources and make decisions around it. Well, what happened since then is now we're building systems that not only do that, but they can take action, so they can invoke transactions or make decisions and do things, which means they are not only intelligent bots, but they are intelligent systems of action. Now, that's a massive change to the market because an agent, as I just defined it, is essentially a workflow. And all of our systems and business applications all over our companies are filled with workflows. If this, then this. If the customer says this, do this. If this happens, do this. If this number gets too high, do that. If we run out of stuff here, order this. If this supplier underperforms, you know, use this supplier. If this candidate can't fill out this form, reject them and hire this candidate. If this person finishes this on time, give them a raise. I mean, there's millions of these things and they're very, very complex. And so what's happening within companies and then within vendors is we're starting to build dozens and dozens and dozens of agents as standalone applications or mini applications within bigger systems. And, you know, we've had this agentic type of system for a long time. There have been workflow systems since I worked at IBM. CICS in the mainframe was a workflow system. I mean, that's the way all of our corporate applications work. But we never had anything that could analyze information in Such an intelligent way and then make much more complex decisions on our behalf. So broadly speaking, that's what agents are. Now where are they? They are manifesting themselves as end to end applications and applications within other systems. So in the end to end application market in hr, the most mature agents are in recruiting. If you look at Paradox or Maki People, Maki People, if you haven't heard of them, is one of the most sophisticated recruiting and event talent agents that I've seen. These systems can collect information to create a job description. Create a job description, analyze the job description against the market source candidates, look for skills as we can do in Galileo. Go into Lightcast to identify skills in a particular role. Go into the labor market and find the right city to hire somebody. Understand what the competitive wages are. Open up a job rack and the right candidate sourcing system or advertising channel. Assess candidates against the description of the job. Review the candidates based on various criteria. Send the candidates information, ask answer questions from the candidates as they look for a job and they want to know wages and location and hours. Decide based on the candidate's application which ones seem to be the best fit. Review these internally. Schedule interviews. Create onboarding, create an offer, validate the offer against the market in that city or for that person of that tenure or that experience.
[00:03:58] Issue the offer, conduct and negotiate a start date. Create a custom onboarding program. And then by the way, once it's done all that, it could keep track of your performance, compare your performance against other people, put together a development plan, on and on and on. That's all agentic stuff. And you know, this is all going to get built out by vendors and your IT department because all of the agent vendors, Microsoft ServiceNow, Google, the other enterprise providers, are giving us agent development tools or agent studios so that we can not only turn them on and use their systems, but customize and program their systems. And you don't have to be a software engineer to create these workflows. They're more or less drag and drop visual user interfaces. So there's that whole market that's just beginning to evolve. And in some sense, you know, agents, this agentic idea is the most threatening to the current job market because it's going to change the jobs and the roles that we're in. This is the level three in our four level AI adoption model that you can read about in the super worker research or in a lot of the webinars we're doing. But there's more. The agents also exist within applications, so within SAP, within workday, within Whatever your hr, within ukg, there are agents that are going to make decisions and help you as an administrator or an HR person or an employee do things. If you're using UKG and you're an hourly worker in a retail store and you run out of, I don't know, dresses that people are buying and you want to tell the factory or the supplier that we need more of a certain item, the agent could do that for you. Or you could tell the agent, order more of the small ones but not any more of the big ones because nobody's buying the big ones. Likewise, you might ask, ask the agent, can I change my shift next week because I have to stay home with my kids? Or you might ask the agent, please give me a review of my 401k and if it's a little bit too high, reduce the amount of money I'm putting into the 1k. Now, as an hourly worker, there's many, many things that are much more complex to that. Can I swap my shift with so and so can I come in late today but stay late based on my hourly rate, is there a way I can book more overtime? Is there overtime available for me on and am I certified for this new role? So you can see these agent systems are very, very powerful and there's thousands and thousands of applications of them in our companies. And so within these large HR and ERP systems, the vendors are starting to build internal embedded agents that look like applications. So if you go to workday, for example, they have a bunch of, they have new agent that helps you with job architecture that will help you design a new position so that it a an offshoot or an oddball job based on other jobs in the company. To prevent all this proliferation of job architecture, which I've talked about a lot, agents for finance professionals, to make sure that the accounts receivable isn't too high and that they're going after the right vendors or the right suppliers. I mean, you could just go on and on and on. So what's going to happen very quickly, it's already happening, is your internal organization, your company, your. IT is going to have a lot of these agents. And that leads to two or three issues. Number one, who's going to maintain them, program them, monitor them, train them, which is basically going to be a whole career path for many, many of us in our companies. We're going to be, you know, sort of managing these agents, if you understand that idea. But the second thing, we're going to have to manage and govern them. Now, the word Govern is a big word. And it, you know, it tends to mean organizing and structuring and controlling and, you know, keeping track of things. And so given the fact that there might be, you know, 50 to 75 to 100 agents running in a company in their systems, we're going to need a layer of software to control and manage and monitor these agents. And so there's another market that's been created for the agent management layer. You're going to see a whole bunch of stuff of this come out from ServiceNow. Next week, Google announced an agent management technology called Agent Space. The workday agent system of record is their attempts to do this. There will be others, I'm sure there will be infrastructure technology infrastructure vendors that are, that are really good at this, as usually happens, and they will probably become more standard in it. And then these other guys will have to plug in. So we have these agent programming environments, we have these agent management environments, and then we have this issue of how agents talk to each other. Now, on that last point, there really are no standards yet. There are a series of standards. There's a standard that was created by Anthropic, there's a Google standard called A to A, and there's others being developed. And so the bigger vendors are working with these standards bodies to try to come up with some standard way for agents to communicate. I have a feeling that that's not going to happen anytime soon. Because so many of these agent tools have unique features. It would be very hard to standardize the APIs between them while the agents themselves are getting smarter and smarter. Because what you really want is one agent to say to the other one, hey, I need to make a decision about X. You're the expert on X. Give me the answer to this question. And then based on your answer, I will make the decisions. Because the agents are, you know, a little bit like employees in the sense that I'm not the expert on this. Let me call the expert. And then based on what the expert says, I will make the decision to do what I know how to do in my job. Agent Interact Interoperability is going to be an interesting domain to watch. Agent governance will be an interesting opportunity to watch as well as agent development tools. Now, next week I'm going to give you. I don't want to, you know, kind of jump the gun on all these announcements that are coming, but we're going to announce our agent integration strategy for Galileo. You're going to see some pretty cool announcements from ServiceNow. And then, of course, workday previewed for us, some of which is announced, some of it's not announced. A bunch of the agents that are going on inside the workday system. SAP has formally announced the agents that they've developed, including Joule. So you're going to see a lot of this, including many exciting vendors like Maki People and Paradox and others coming from the dedicated HR tech vendors. By the way, the learning and development market, which I'm going to talk a lot about over the next couple of weeks, is going to have agents too. And there's a lot of agents out there already. We're going to have one couple to Galileo. There's a who I've talked about and just did a really fun video interview with them. So, you know, stay tuned for more on this. Now, what we, what we don't want to do, what we're trying to prevent, is a rat's nest of agents doing things and us not being able to keep track of which agent made what decision and how did this transaction or activity happen so we can audit the transaction. In fact, auditing is a big issue. When I was back when I was at IBM, I keep going back to my IBM days because I learned so much there. Back in the mainframe days when a lot of these ideas were basically done inside of a mainframe, we had really important systems. The one that I learned a lot about was called CICs that did transaction monitoring and auditing. And that was because in a bank and a lot of these early IT systems were financial systems. If you withdrew money from an account but the network went down right in the middle of the transaction, you didn't want the database to be left open. Not knowing how much money this person had. The database had to know that based on the, you know, 5 milliseconds or 10 seconds it took for this transaction to take place, what were the things that were happening at a micro level so that if it interrupted, it could undo or roll back those transactions. Now imagine that across billions of transactions a week, you got to audit that stuff. So there's going to have to be auditing tools for these agents. We're going to have to keep track of what data sources agent has access to and what corpus it's using, how the corpus is being updated and managed, how it's making decisions. So I can't even imagine how complex these agent management tools are going to have to be over time. Right now I think the ones I've seen are pretty primitive from these vendors and I, I don't really believe that the ERPs like Workday are going to probably be successful in this market over time because it's so specialized and there's going to be so many use cases beyond what they do with their systems. Okay, second topic I want to talk about is I want to talk about ServiceNow in particular. ServiceNow is one of the hottest IT tech stocks ever. It's not quite as hot as Nvidia, but it's pretty doggone close. ServiceNow was founded in 2003. It was founded as a IT service management technology to manage essentially cases and call centers. Were you calling the IT department because I can't get my computer to work or I can't reset my password. And they would take a case and then they would route it to an agent and then the agent would log into some database and say, ah, here's the answer. Let me send you the answer. And they were smart enough to realize that that call center automation type of case management technology was actually a much broader workflow system. And they extended the idea of IT to be service delivery of all sorts of services, including HR. And so the ServiceNow Workflo management system, which this is all pre AI, was better and better and better at managing many, many kinds of workflows. And what happened was that as the ERP vendors tried to build more and more applications, we realized that we needed some workflow layer to build workflows across applications because everybody has multiple systems, even small companies. So ServiceNow staked out the space for larger or complex companies to build complex workflows. That includes onboarding, supply chain, legal issues, workspace. What office should I go to today? How do I schedule a meeting in an office and a desk and a slide projector, you know, all that stuff. I mean, just tons and tons of things. And it turns out it was a really good business decision because they are now three and a half times the market cap of workday. They're a big, big gigantic company. Expensive software purchased primarily by it, although they do position it to CEOs and CFOs and CHROs. Also because it is a, you know, a big architectural buy and a very, very sophisticated software company under the covers. I've spent a lot of time talking to ServiceNow, meeting people at ServiceNow as an analyst that I'm really, really impressed with the way they think, the way they execute, the way products. The only complaints I get from customers is that it's very expensive and that it takes a while to implement, which of course, you know, it's a, it's an application development environment as much as it is a service delivery environment. So you do have to implement it. And of course, just like every other major software company, after a while it just gets more complicated and takes a while to, you know, figure out what you're going to do with it. But it is, in a sense the Erector set or the Swiss army knife of enterprise applications because they've built very deep integrations to all of the major business applications that we use. Oracle, Workday, SAP, et cetera, et cetera, et cetera. And because they're such shrewd business people, their middleware or workflow layer is turning into applications itself. So what you're going to see next week is a pretty big announcement about a new market they're going to get into that competes with a lot of these core application vendors. I will not talk about it yet, but you'll hear about it next week. And they've pivoted the entire company to AI. Not only have they built AI layers and they partner with Nvidia for the delivery of AI and they have, you know, custom LLMs and they use off the shelf LLMs, but they've built an AI development environment and AI management system. And it is all under the name now Assist. And we are partnering with them with Galileo. You'll see what we did next week. Because large Companies don't want 50 agents for their employees. I mean, it's okay if all these agents are running behind the scenes inside the company, but if I'm an employee, I don't want to log into this agent for this and this agent for that and try to figure out. Because every single application vendor is building agents on top of their apps. So they're going to be the agent of agents. Frankly, I think that's where they're going to end up. And they'll probably use that kind of marketing. And I think they deserve the credit for, for getting there because this has been a very strategic and important set of investments they've made to get there. The other thing that ServiceNow does that's unique to the application vendors like Workday and the LMS vendors and the ATS vendors and all the ones that we use in HR is they understand the functional capabilities of workflows themselves. And so they realize that ultimately AI for a business, not necessarily for an individual, is a tool set for redesigning, reimagining, re engineering workflows. And that is really what's happening in AI world. In hr, we have in HR hundreds of workflows. They tend to be patterned after the employee experience or the talent management journey that we created or the the pre hire to retire tal talent management journey. There have been hundreds of companies that have built these models and you've seen them all over. If you go to a trade show, you'll see them of what are the transactions or workflows that happen for employees. And sometimes they're different by industry and they're different by role and by level and by geography. And there's legal ramifications of all of them. Well, ServiceNow, because they have this broader context. Instead of building a system that does everything for you, they're building a tool that allows you to build a system. They understand the nature of those workflows. And so I think a very large percentage of the AI automation technology market is going to be IT departments building things, connecting together things to build applications in this agent layer. And this is the big sort of industry change that's taking place. And I think all the application vendors are a little nervous about it. And that is that if you're a big IT department and I run into these all the time and you don't really trust vendor A or vendor B to have all the applications you need, you might just build it yourself. And I know that sounds a little bit strange to many of you, but that's the way it always used to be. When I got into IT stuff in the 80s and 90s, every single system was built by the IT department. IT departments are quite capable of building applications just like software vendors now. They don't want to build stuff that's, you know, highly off the shelf repeatable stuff. But I was in a meeting last week in Europe and with a bunch of heads of talent acquisition and about one out of five or one out of six of them had their own custom agents built by it on OpenAI or another tech stack. So we're going to end up with some companies taking the off the shelf tool from Workday, Oracle, SAP or whatever other vendor and saying okay, that's pretty good, but you know, I think we can build something better on top of it and let's not use this module because it's not quite what we need. We're going to build it ourselves and they're going to have ServiceNow and other vendors all ready to help them. And if you look at the growth of ServiceNow's business relative to the packaged application vendors, it's almost twice the rate of growth. Because this custom building, assembling, Lego block approach is very appealing in a very young immature market when the full end to end systems are not built yet. Now ideally, of course, I'd like this agent to come out of the box and I'd like the agent to be pre made for my company, for my application, for my industry. But they're not there yet. I mean we have, for example, we use HubSpot and I love HubSpot, it's a fantastic system. I would, I'm not a fan of Salesforce, I've never liked it. I got tired of a long time ago. And HubSpot introduced an agent called Breeze. And we're pretty sophisticated user. You know, we looked at Breeze, it's got some good stuff in it, but it's not there yet, so we're not going to use it. So we're stuck using HubSpot as is and doing things with the current system. And that's the nature of this early market is all the vendors are trying to build great stuff, but the IT departments are pretty savvy, they know what their companies need. And I think a bigger, a lot of the bigger companies are going to be building a lot of their own stuff and they're going to really look for these governance platforms to manage all the agents that are being built or sold to them or implemented in all these systems. Okay, now I, this was kind of a, an interesting discussion about what's going on in the market. Let me now shift over to some, you know, kind of application areas that are going to come out next week. So the next couple of weeks are interesting. Next week is both the ServiceNow conference. There'll be some, you know, big announcements there. There's unleash, I'll be speaking on Thursday and then we're going to have our workshops. You're going to see two announcements from us. The following week is another series of vendor activities. Eightfold is going to be announcing some things and other vendors. And then the following week is Irresistible. And let me just put in a plug for irresistible. Irresistible is a 450 person extravaganza of fun and education and engagement and really support at USC campus. We have about, I don't know, 60 or 70 CHROs coming, maybe more. We have some very special guests from all over the world. We're going to have some big awards. We're going to introduce our new learning system which you guys are going to get pretty jazzed about, I think. And we're going to have a couple of people from the federal government there if you want to hear from them. And a lot of discussions about AI, some very strategic vendors, not a lot of vendors. And so I encourage you to come Bring your team. We got a couple more weeks for you to sign up. It's almost full, but it will fill up. And what we're going to do at that conference is not only talk about some new things from us, but I'm going to give you my assessment of the Super Worker organization. Now, last fall and early this year when I, when we launched the Super Worker research, what we tried to do, and it was a lot of forward thinking, was show you where the world is going and where AI is moving to in this kind of complex world of, you know, business transformation. We now know over the last six months what's happening and sure enough, a lot of jobs are being restructured and companies are becoming simpler slowly, but it is happening. And I think there's this new model of a company which I call the Super Worker organization, which is a smaller company, more productive company, maybe not smaller in terms of number of people, but smaller in terms of employees per dollar of revenue. More tight and integrated support functions like HR and others, better use of data, faster time to market, and highly competitive against its peers. And every single one of you are working for a company that wants to do this now. You're not going to do it in every part of the company. At the same time. You're going to have to decide where you apply the super Worker organization effect first, because different parts of the company are going to need at different rates. And then you will implement it through a whole variety of practices that I'm going to talk about at the conference. So stay tuned for that. I also, as I wrote about in the blog this week or last week, still up there, this is going to affect HR a lot. HR L&D, TA, talent management, business partners, call centers, analysts, et cetera, are going to change significantly. AI is going to automate systemic hr. Systemic HR is the operating model, but the number of people in HR is going to change. The rough ratio of HR people per Employee is about 1 to 100, roughly. It varies. Some companies it's way, way higher. Some companies it's a little bit lower. I think that number is going to go down. In other words, there won't be as many people needed, which means that we're going to have this new series of roles which I'll talk about at the conference for HR people that we're doing more routine administrative work. So all of this is coming in the next month or two. And of course, many things I probably don't even know about yet from other vendors. I would also encourage you to read the article I wrote about the skills agent from Microsoft. They kind of slipped it into a whole bunch of other announcements they were doing on the Copilot. But I think it's very, very significant piece of technology. It's going to upset a lot of the skills vendors and gives you some new options on how to manage the skills infrastructure in your company. And come see us at Unleash, come to Irresistible or just come give us a call and we'll come talk to you about what's going on in your company. Have a good weekend and next week I'll go through all these announcements with you. Bye for now.
