Episode Transcript
[00:00:00] Okay, today I want to give you guys a little bit of education on the word semantic layer and the AI battle brewing for the context layer of business rules that all of our new AI agents use.
[00:00:18] Now, the word business rule is kind of a weird phrase because there's literally thousands of things that fall into the category of business rules. For example, at Deloitte, because it's a partner centric business, you're not supposed to call clients without asking the partner for permission. I don't know if that's written down, I don't know if that's in the HCM system. But it's a rule, believe me. And if you violate it, you'll know. And so you take that idea and you proliferate it throughout HR and sales and marketing and everything else that companies do. There's thousands, if not tens of thousands of these privileges, approvals, checkpoints, workflows and security rules that companies have. And they're embedded in your legacy systems. Let's call all the cloud systems legacy for the moment. Even though I know that's a little bit of a negative term, but really it's not so. Because what's really happened for the last 20 years, since the Internet sort of started, is we built our company business processes into these big platforms. The sales organization runs on Salesforce or HubSpot. The marketing does on its own tools. All of the HR stuff runs on Workday, on SuccessFactors, on Oracle, on UKG, on Dayforce, on ADP, including all the payroll rules, the tax rules, the accounting rules, what we decide to give people as benefits, and the scheduling rules that might be in another system. The learning approvals and the learning charges, they're in the learning system. The way candidates are treated, the applicant tracking system. I mean, this stuff's everywhere. And it's very glib to say we're going to redo it all in an agent system, but you can't really do that, especially if everybody in the company is building their own agents without some layer or application interface to all of these rules. Which means all of the big vendors want to be your, your context layer. They call it a context layer or semantic layer. And it's not as simple as pulling the business logic out of these systems. If you look at Workday or SAP or any of these tools, there's a lot of logic in there. Some of it's implicit in the application, some of that stuff that you've added over many, many years. But then there's higher level constructs like the career paths or the skills or the Definitions of our five levels of performance or our leadership models, I don't even know. I mean, there's hundreds of things that companies build to operationalize their strategy that sit around in different places that are used by everybody. And I can't possibly imagine what they would be in your company. But I'm sure you've got many of your own and some of them are written down and some of them aren't. Then you add to that the third party regulatory or legal policies that are also enforced, like if you're in banking or insurance or oil and gas or other places, then you've got the safety policies, then you've got the leadership principles that change every three to five years. There's a lot of business rules, quote unquote, that make up your company and no two companies have the same rules. You take two companies in the same business selling the same products and they're very different. And that's what's the human nature of business is the way we decide to organize ourselves and, and how much freedom we give individuals versus the system. And you know, there's some companies that are highly distributed and highly decentralized and highly empowered and there's others that are the complete opposite where everything has to be controlled, especially if you're in banking or a highly regulated, you know, sector. So, so anyway, so you start building agents and ServiceNow this week announced an announcement called the Context Engine, which is a beautiful demo of a fairly simple example of how you could build those rules into their tool.
[00:04:15] And I don't know if they built this or bought it. It's a little bit unclear where it came from. And they use all the connectors they have to other systems. Gloat announced something very similar, but much more sophisticated for HR last week. And then there are other companies out there that do this for testing. Like if you hire a firm to test the implementation of Workday or Oracle, SAP or some other system, they're going to crawl through the system with their agents and they're going to see what the rules are and they're going to try to test them to see if they're implemented correctly. And I do believe that this, this, this sort of intermediate semantic layer space is going to attract a lot of players. You have companies like Okta that do security and you know, and then you have the agent force Sana Joule offerings from the big erps that are trying to do this. It's very hard for any of these software companies to thrive if they don't own the rules that are put into now, ServiceNow is a big company and they have enormous power in a sense, over larger IT departments because they've done such a good job of providing case management, IT service management and other functional management tools back in the automation era. Now in the AI era, they're trying to do it again in hr.
[00:05:31] They're very successful, not as big as they'd like, but they're growing a lot in IT service delivery, IT service management.
[00:05:38] And so they're claiming on the announcement that's coming out today that the Context engine is the place to put all this stuff. And I wrote an article about this and it's a pretty big move by them and it makes perfect sense. And I don't know where it's going to go, but I would say that there's issues here. The first is, do you want to hard code or soft code everything into a new vendor layer, which you now have to pay for? Because ServiceNow also announced a pricing model that sort of shocked me out of my chair. What they're essentially announcing, I believe, and I could be wrong, but let me just tell you, what they told us was that you're going to pay for service now by token usage. So you're going to be metered, and there are going to be three types of tokens, depending on the types of agents you build. The information agents will have a sort of a regular price token, the autonomous, or rather the automation agents will have a higher price token, and then the autonomous agents will have an even higher price token. How they enforce this is beyond me, but what they're kind of saying is we are going to capture value from the amount of value you capture from your agent, which is almost like they're trying to share your cost savings with their revenue. Now, I don't know, you know, maybe that'll work, but it's a little bit odd.
[00:06:58] You know, it's hard to price software to value because every company delivers value in different ways and it's proprietary to that company. And I doubt they're going to go in and inspect what you're doing with your agents, but somehow they're planning on doing this. And maybe Wall Street's going to like it, maybe the customers will like it. I don't think I would like it. And that's another part of their strategy. Now, you can't argue with ServiceNow's success. They have a lot of customers, a very smart company, great engineers, they have a lot on their plate at the moment. I mean, this is a lot of engineering going on there, but you've got to hand it to them. To boldly move from cloud workflow to AI is a big move and they've done a pretty, pretty spectacular job of putting it all together. So, anyway, that's big. Gloat is another one to look at. A big, big fan of a company we've been a fan of for a long, long time. We've worked with Gloat since they were founded. I know the management team very well. I think they're going to have a really interesting time here. We're going to be doing some integration work with Galileo there, where we are already integrated with ServiceNow now at the assist level. So Galileo, Galileo, by the way, just let me add that comment. If you think about that semantic layer of business rules, in some ways Galileo is the intelligence layer on top of the semantic layer that you can talk to. So the example that ServiceNow gave yesterday was an employee trying to book a business class airfare to go to a conference, and the business rules lighting up and saying, you know, you're not entitled to do this, you're not the right level, or whatever. That's a very simple, almost pedestrian business rule that's not very hard to understand or figure out. But what would normally happen before that would happen is the employee would go to their manager or the employee would go to Galileo and say, I'm kind of stagnant in my job, here's what I do. And maybe Galileo already knows what you do. What do you suggest I do as a development plan for the next six months to get promoted or to get a raise? And Galileo would look at your skills and would look at the skills in the market and would look at the jobs in your company. And it would say, well, here's a couple of really good opportunities for you. If you'd like to go this direction, I can build you a development plan. If you'd like to go this direction, I'll build you a development plan. And then you pick one and it would build a development plan. And we are doing this right now for clients who have taken our capability assessment. For example, we've had a 94 practice HR capability assessment now for, I don't know, seven or eight years. And a lot of companies have gone through it with their HR teams. Now we can take that capability assessment output and we can generate personalized development plans for every HR person in your company from Galileo. And then they can use Galileo to learn because it's got the training in it. So. And then in the middle of that, you might decide to book a Flight and then you get the business role. So you can see where I'm going is that this semantic layer or context layer, it's got a lot of names, is just one layer of basically your business written down on paper, what you do. And you know, my experience in my 40 plus years of working is that these things are constantly under change and discussion. You know, this policy, we don't want to use it anymore because in this case we want to override it. And why do we have it this way when we could have it that way? And this thing is getting in the way and people don't like it. So it's not a simple decision to stick this stuff in one place. I think the way business rules tend to work is there's a home for every business rule. And that home isn't just a system of record, it's a system of business.
[00:10:38] And you implement it there because that's where it belongs. And then the context layer might ingest it in to be aware of it, but you don't want to write it in a centralized system and then deploy it into an operational system later because it may not work right in the operational system. You know, Salesforce is very unique and problematic and HubSpot is different and you know, different case management tools are different. So. So you're not going to write that rule at an abstract layer and then try to figure out how to implement it in each of these systems. I think that's probably backwards the way most companies would do it. It all reminds me of my younger years at IBM in the mainframe days when all of this was coded in COBOL in mainframes. You know, mainframe systems were extremely well organized systems. The COBOL structuring programming language was exactly designed for this.
[00:11:30] And there were systems like CICS on top of it that embedded these rules into transactional systems. So you could write a rule and it would be enforced by the IBM infrastructure. You know, now that that's all been blown up and distributed into hundreds of open systems, we got to figure out how to do it in a new world. And I think ServiceNow is going to make a huge amount of money at this and they'll probably be very successful with it. But it's such an important problem that I don't think they're going to be the only ones. And they're a little bit early to this. And there are very arcane manifestations of it in different areas of HR that I think you guys are going to need to be aware of. So stay tuned. We're going to keep up on this. And this is part of the HR 2030 story. You know, we got, if you, if you like my 2030 narrative, we got three and a half or four years to think this through.
[00:12:21] Probably a lot more than that, of course, as the world evolves to agentic AI and agentic hr. Okay. Call us if you have any questions. We can walk you through this in the blueprint work that we do. Going to have more blueprint sessions around the country and around the world in the next couple of months. And if you really want to get into this face to face, come to our conference in June at usc and you'll have a great time, I promise. Thanks a lot.
