Episode Transcript
[00:00:00] Hey everybody.
[00:00:01] Today I want to talk about Claude code and you as a citizen developer. There's a sweeping trend going on in the AI world driven largely by Claude code, but by other tools to enable you as an HR or a business person to build things using AI. And the theory behind this and the technology behind this is that the thing that AI is the best at doing is writing code.
[00:00:33] It's pretty good at answering questions, it's pretty good at chatting, it's pretty good at teaching you stuff, but it makes mistakes.
[00:00:40] But it's actually very, very, very good at writing code. And so what Anthropic has done, and the other vendors are going to figure this out, is they've said code generation is our key to doing everything else.
[00:00:58] Rather than train our models to do every possible scenario of every video, audio or other workflow that we decide we want to use AI for, we're going to let it write code that does that. Now, this has been going on for a couple of years.
[00:01:15] There's a two to two and a half billion dollar industry already of co generation, what are called vibe coding apps. But those are really kind of used by software developers, not really individuals and business people. But it's turning into a business tool and there's many implications of this for HR technology, HR management and leadership.
[00:01:39] First of all, it really does work. If you open up Anthropic or you use Galileo and you use Claude, it works best in Claude and you tell it to write an application to do something, it will do that for you and it will display it. I am going to be demonstrating next week on the webinar I'm doing on Wednesday how I did this for a performance management application and it took me literally 10 minutes to get it to work and it worked right out of the box. It had one glitch.
[00:02:13] I took a screenshot of the glitch, sent it back to Claude and said, why is this not working? And it fixed it so I didn't even have to debug it myself. Now, there are many ways to do this and the software industry is all over this.
[00:02:28] Microsoft now generates more revenue from the GitHub copilot than the business copilot, according to the research I've seen, and you've seen more and more vendors, including Workday, get into the Vibe coding business.
[00:02:42] But I think the bigger implication is for you, the way you think about AI and corporate applications in the traditional enterprise AI market, going back 30 years into applications, back in the mainframe days, we built them in cobol. So we had large teams of database developers and COBOL developers, not really user interface developers, because they were mostly green screens building apps.
[00:03:11] And those apps run our banks and our insurance companies. And they're still there. The mainframes are still there too, and we've layered on more and more stuff on top of them.
[00:03:21] And then in the 1980s and on into the 90s and 2000s, packaged application software took off.
[00:03:29] The early vendors are all gone, but there was Computer Associates and many others that built corporate applications, eventually leading to Oracle, SAP, Workday, ukg, Dayforce, ADP and hundreds of others. And very, very few companies developed their own payroll systems, their own time and attendance systems, their own scheduling systems, or even their own performance management systems. And of course, what happened was there were a flurry of HR apps in the talent management space in the early 2000s, hundreds and hundred of them, including learning management, applicant tracking, various forms of talent management.
[00:04:09] And the same thing happened in sales. And this is when salesforce.com took off. And the big ones got big and the small ones got acquired. And eventually the HCM vendors figured this out and did it themselves.
[00:04:22] So most of you are going down the path of buying human capital software for your HR function and you're relying on medium to large vendors to do what you want.
[00:04:33] But as you know, it's never quite what you need, it's never quite finished, it's never quite customizable to the way you want it. Customizable. It's hard to learn, it's hard to use, it's hard to integrate, and you're dependent on this vendor's roadmap. And they have many competing priorities. And so you have to sort of debate with them what you need versus what another customer needs, or go to their user conference and meet the CEO, et cetera. And of course they have product managers, each of whom are learning on the job, and their product managers are coming and going and sometimes they don't know what to build or they have ideas and they don't tend to be the right ideas. And so the whole idea of packaged software is sort of problematic. I mean, I'm not saying it doesn't work because these are pretty big companies now, but it hasn't been easy for them and it hasn't been easy for us.
[00:05:24] Along comes this new technology and the whole focus of how we do this is about to change.
[00:05:32] There's going to be a massive magnetic force towards you building your own solutions.
[00:05:40] We're going to demonstrate some of this at Irresistible in June. We are, we have a large Very large company coming that's going to show you all the stuff that they have built to run their HR department without the use of vendor software.
[00:05:56] Now, that happens to be a company that has a lot of engineers, but most of you don't.
[00:06:01] Well, if you look at what these AI coding tools can do, you can build your own apps, but they're not called apps, they're called agents.
[00:06:11] So in the old world, we had a monolithic system that had multiple functional areas. It had a user interface, it had a bunch of modules and a bunch of product and engineering teams building those modules and connecting them together. And what would happen was most of the interconnectivity worked, some of it didn't work. Some of the modules were built out in great detail, some of them were not. And they were all interlocked. So if you wanted a feature added in one of them, you usually had to wait period of time for a bunch of features in the other ones to be built because they were really linked together behind the scenes with different workflows. And I'm talking about all these vendors, they all work this way. That's just the way the software works. In the new world of agents and super agents, which you're going to hear about next week, this is completely different.
[00:07:02] Rather than thinking of a business workflow like onboarding or recruiting or career development or promotion or location mobility or job change or salary change or whatever it may be as an integrated process, it's going to be a series of independent agents, each of which is dedicated to doing the thing that it knows how to do with the data that it has access to, connected together with what we call super agents. And instead of having one interlocked application where you can't really fix it easily without dealing with all of the product issues of all the connectivity to other functionality in it, each agent will have its own developer or owner or roadmap, and the agent will be available to talk to the super agent, which we call it the coordinating agent, or the user interface agent, to produce a better and better outcome. You know, a perfect example of this is onboarding. If you look at onboarding, and we've been talking to a client about this this week, it includes getting somebody a desk, getting somebody a security card, getting them a computer, getting them a login id, getting them into the payroll system, getting their benefits filled out, getting their various HR forms filed, telling their manager when they're coming, giving them information on how to get to work and what time to get to work, and what are the policies of getting to work, giving them some education on their job, getting them started in their job, teaching them what their first week and first day will be, connecting them to other people more, and on and on and on. Every company's got their own version of this.
[00:08:44] Well, all of those steps are located in other systems. There's a system for facilities, there's a system for PCs, there's a system for training, there's a system for filing forms, there's a, there's a system for payroll, there's a system for onboarding training.
[00:09:02] And you know, maybe there's a dozen different systems. So if you were to build an onboarding app in the old fashioned traditional approach, you would be building essentially an enterprise app. And that's why there aren't very many of them, because the onboarding process is a very custom process to each company.
[00:09:20] And if you even had such an app from a vendor, you would be spending most of your time connecting it to other apps and making sure that the interfaces between the two work. And by the way, every company has this problem. And if you're a global company, multiply that by another, maybe 30 or 40 by the version in France, the version in England, the version in this city, the version in that city, and the version in this country. Well, in the new world of agents, you will have an agent responsible for all of these little processes. And this is what the blueprint's going to show you. Those of you that are clients are going to be able to get our blueprint, and the blueprint will show you a template for all of the agents that are likely to be needed. And some of these agents you could buy, but a lot of them you could build using coding like what I mentioned earlier. So this new world is going to be much, much more controllable and manageable by you.
[00:10:15] Now, we're not going to go back to the world of IT departments building big monolithic applications and having hundreds and hundreds of COBOL programmers. But we're going to go back into some direction like that because I know the cost and expense and complexity of buying vendor software is high. Now, some vendors are going to adapt, of course, and probably newer ones will give you much more modular applications for doing this. But if the incumbent systems you have don't move in this direction, there's no question that many of you are going to start doing this on your own and you're going to get many, many tools from the AI platform providers and you're going to be in the control center of building much more of your HR automation and Human capital systems and solutions yourself. Now Galileo, the new version of Galileo just coming out in March, which is called Mars, we went to Venus and now we're going to Mars. We're going one planet at a time through the solar system here is all about building things. And you're going to see a whole bunch of features in the new version of Galileo for building courses, for building content, for building applications, for building workflows, for building solutions, for building agents. Because we of course want to help you do this. Because if you use Galileo, it's already trained and knows everything it needs to know about hr. It's not context empty, it's context aware. Now, from the standpoint of the software industry, this is even more disruptive. And I'm going to be writing some articles about this, but I don't quite have enough information yet. If you look at the revenue rate, revenue growth rate of the big HCM vendors, and this is Oracle, SAP, Workday, Dayforce, ukg, Lattice, all, all of the other software vendors, some of them are growing at pretty good rates, but the big ones have slowed down a lot and their stocks have all declined. The reason for that is the external market, the analysts, the stock market, the folks that are out there looking at these industries realize now that the Claude code AI generation market is going to eat into the revenues and the growth rates of these big HR and other and CRM and other software companies.
[00:12:36] Adobe stock is down, Salesforce's stock is down. And if you look at the package software industry, it's really flattened out, not going away. We're going to need this stuff for a long, long time, but this growth is happening in this new era. You could argue that the business that the big software vendors are going to have to get into is the business of giving you code building tools of your own. Which means that not only are they going to have to build and deliver LLM type of solutions of their own, because you can get these from the third party LLMs already, but they're also going to have to invest in infrastructure. Interestingly enough, Oracle has now spent almost $400 billion on AI infrastructure to build a massive AI data center. And I think the reason they're doing that, and they're way, way, way down the path here compared to everybody else, is I'm sure Larry Ellison is thinking to himself, I gotta be in the token generation business, not the software business, because the token generation business is going to replace the software business. That's my guess. I've never talked to him about it and he's never said that publicly, but why else would Oracle turn itself into an AI infrastructure company?
[00:13:54] The bottom line for you as an HR person, as an IT person, as a business professional is you have to get your hands dirty on this new technology, play with IT, and start talking to other people about how they build these agents or us. This is going to be the path through the enterprise AI journey in 2026 and beyond. Given the money, the investment, the intelligent people and the strategies behind the scenes that are going into this, I think this is going to be a massively interesting and exciting opportunity ahead. Stay tuned for my big webinar on Wednesday where I'm going to talk about all of this stuff in the context of the 2026 imperatives, and I will be presenting this information at the trade shows around the world when I do my various speeches. Call us if you'd like. Help get your hands on Galileo. Everything you need to know is there. Galileo will teach you about this stuff and have a great week. Bye for now.
