Episode Transcript
[00:00:07] Hey, everybody. Today I want to talk about AI and some really interesting things that have been happening that are very important to understand relative to the architecture of all of our HR technology. So in the last several weeks, the folks at Microsoft have done a lot of introductions and launch of tools for using and integrating the Microsoft copilot into different systems. And at the same time, that's going on. Products like Galileo from us, Joule, J O U L e from SAP, and others are coming that are also chatbot interfaces to HR tech. So how does this all come together?
[00:00:50] Let me start by just talking about architectural issues, and then I want to talk about SAP. I want to talk about what we are doing relative to Microsoft and other systems, and then the implications on this on the market.
[00:01:03] The architectural issues are as follows. The AI platforms being developed that are chat like interfaces, like Chat, GBT, Gemini, Anthropoc, etcetera, are full stack, top to bottom AI platforms. They're built on vector databases. They use large language models of different sizes and shapes, and they interpret language and can access and analyze massive amounts of data, as you know, and I won't go into the details of that. They're very useful for expert information retrieval, like we do in Galileo. They're very good for generating content, in other words, creating documents, creating artifacts, including pictures. They're now becoming good at talking to you, thanks to the Omni technology that's coming out of OpenAI. And they're also good for what are called actions, implementing transactions in other systems. In the case of our platform, Galileo, we have huge amount of corpus of AHR data in there. We have our trusted content partners with benchmarks from visir skills, data from lightcast, leadership models from Heidrick, and global employment practices from oyster. The system will be able to invoke transactions on other applications. We do that work with Sana, because Sana is our platform, and Sana has already done this in Salesforce, and we're starting to talk to workday, and we will also talk to SAP about that. But for the SAP oriented, or workday oriented companies who have transactional systems, they want their chatbots to, to look like assistance, in a sense, to using their application. So what SAP has done is actually quite extensive. They've built an entire AI set of services in the SAP infrastructure, manifesting in a front end called Joule. Joule. Joule looks like a chatbot. And when you use Joule, you can ask it questions. It'll get data out of SAP like, what was my last expense account? Or, you know, how long has so and so worked in the company, or, you know, things like that. You can load your HR documentation and policy data into the system so that it can make that available to employees and the system can do transactions for you. So you can ask Juul to create a new job, open a rack, you know, update somebody's profile, update somebody's salary, etcetera. It's an extremely big undertaking for SAP and really strategic because SAP started early on this. They understand that this new system will replace many of the user interfaces they have. And because it's built on a model that allows you to use multiple LLMs, you're not dependent on one LLM under the covers. Under the covers, they select the LLM. That's appropriate. By the way, we do the same thing in Galileo. Galileo. Even though you don't realize it. The reason it's so good at answering questions is that Sana has multiple LLMs under the COVID And the question and the question path that you send it will go to a different LLM depending on the type of question you ask, as opposed to going out and buying OpenAI, where basically that's what you're getting, is OpenAI. So SAP is well ahead here with hundreds of transactions mapped into Joule, and they are mapping it in the context of sort of a lifecycle of an employee, because if you go through every transaction in SAP, there must be 10,000 transactions. But what they're doing is they're looking at the prehir to retire cycle and finding ways to add AI interfaces to those cycles. So I think for those of you that are SAP customers using SAP, considering SAP, you're going to be astounded at what Joule can do, because not only have they upgraded the user interface itself in a very significant way and added all these new capabilities for talent, intelligence, skills assessment, and other tools in the system, you don't have to find all those buttons to do it. So that leads to the larger architectural issue of what is the relationship between Joule, Galileo, Microsoft, Copilot, and let's suppose a copilot that comes from your ATS paradox, or whatever it may be. All of these chat bots are user interfaces for an employee or a user. Are we going to end up having the same mess we had a few decades ago when we had all these different interfaces of different systems? Well, actually, yes, today it is going to look like that. But these interfaces are opening up because since each of the transactional and conversational interfaces has deep domain knowledge in its own unique way, what we really want these things is to talk to each other. I think of Galileo as a really, really smart HR consultant, analyst, researcher and data person, as if you had somebody like me or one of our analysts available anytime to ask any question. I think of Juul as the SAP guru who will run SAP for you. You sort of go to Juul and say, hey, I'd like you to do so and so. And Juul goes out and finds the stuff it needs to do in SAP. I think of Paradox as the recruiter and candidates assistant to help the recruiter and the candidate find the information they need to locate a job, apply for a job, get assessed for a job, track the process for a job and so forth. And these are very deep domain experts. They're deep and you know, and I think there's a depth issue in these chatbots that a horizontal I can do anything chatbot has a very different level of utility from a deep domain chatbot. So then you look at the Microsoft copilot. What is the Microsoft copilot? The Microsoft Copilot is an AI stack system designed initially to operate in the Microsoft graph. The Microsoft graph is the data you have in your Microsoft system. It's your outlook stuff, your emails, your documents, videos that you've stored, your teams meetings, your team schedules, things like that. And the enterprise or corporate or premier version of the co pilot can access all that stuff. So I've already talked to customers who've told me that if they have the Microsoft copilot they can find emails and stuff just as fast as they can find it going into Outlook. It's not an HR expert, it's not an expert of SAP. It's not an expert on workday or anything else. So what Microsoft has done is create an interface for software vendors to create plugins or different versions of what are called declarative plugins to connect to the Microsoft tool. And SAP announced that they are working on this. We are working on this with Microsoft. Eightfold is working on this with Microsoft. ServiceNow is working on this with Microsoft. I believe workday is working on this with Microsoft. And what we are all trying to figure out how to do is make it easier for you as a customer and the user to use the Copilot in a way to access these other specialized systems. Now in the case of SAP and to some degree us, SAP is Microsoft's HCM. So those two companies have a good reason to work very closely together and Microsoft and SAP are actually building a version of the copilot that will be sort of the dual version of the copilot that'll be deeply integrated with SAP. So what you're going to see from SAP is a very, very deep level of integration where the SAP copilot can access the Microsoft cloud or graph, and the Microsoft Copilot can access SAP. I'm not going to give you the details on that. I'll let them explain that as it rolls out. In our case, what we're going to do, and we have not done this yet, but we're starting to scope it, is make it possible for you to get information from Galileo in the Microsoft Copilot and merge that information with other information that we have in Galileo. Galileo has the capability of storing your HR documentation and policy information, too. So if you're not a SAP customer and you don't necessarily want to use Joule for this, you can do all that stuff with our platform, and we'll figure out what our interoperability works with Microsoft, too. Workday is probably going to do the same thing. Servicenow has been doing demos on this. They're working on it also. I think, though, the architectural issue for you, and I've now had maybe a dozen conversations with customers about this, is how good is each of these systems as its standalone identity? And together, I think we have some time yet, maybe six months to a year or longer before the interoperability between these systems becomes easy and mature. It's quite complex for these chatbots to talk to each other. Not only are there architectural interfaces that are not quite built yet and really defined, but some of these things do actions. So the Joule system obviously does actions. In SAP, our system will do some actions based on the things we decide to do, and then the copilot will do some actions of its own. So there'll have to be some negotiation between these chatbots on who's going to take what action, and those of you that are Microsoft shops will probably connect the copilot to your own internal systems, which are different from these vendor systems. So that's kind of architecturally what's going on in the case of SAP in general. I just want to sort of mention in this podcast that after spending several hours now multiple briefings with them, I think they're really ahead of the curve. I think they're six months to a year, perhaps ahead of the other erps, and they do tend to take an architectural approach to this because they're so big and they have so much technology under the covers. So they have not only manifested all these transactions into Joule, but they've created 75 or 100 different AI interfaces in the core system to create documents, give you insights into your career or feedback. They have a writing assistant, all sorts of really interesting things there. So if you're in the sort of process of shopping for software, I think it would be interesting to look at what successfactors SAP has done to give you some vision from where these things are going. Now, the second thing I want to talk about for a couple of minutes, because I know I've already spent 15 minutes here, is L and D. We are now deeply involved in a pretty significant research program to build the high impact learning organization research one more time in L and D. I've done this exact study at least five times in my career, and I do believe that the disruption that AI is going to produce for L and D is going to be the most disruptive we've seen since the birth of the Internet in elearning, which was in 1998 and 1999. So that was 24 years ago. And so what we're now doing is interviewing chief talent officers, chief learning officers Chros. If any of you listening to this would like to talk to us about your AI strategies in L and D, what you're doing relative to content or generation or consumption of content, just send us an email, we'll set up a call. I would say that by the end of this year or early next year, we should be able to show you a lot of details in terms of a maturity model, best practices, all sorts of things we're finding. But I will tell you that my hypothesis so far is that we're going to have to really rethink L and D in a significant way. Number one, the way we generate content has to change the traditional instructional design model where we do months of performance consulting and then build, essentially, courses will be automated. We're going to launch a course next week on AI prompting that we generated and built in partnership with Arist, which is actually a really cool course. You're going to be able to get that for free. And that course was generated in about three days with Arpip and Arist support. So content generation, content creation is going to be very, very different, much more dynamic, much more personalized. We already have tools like Sauna up limit tools from Arist, tools from Tochebo that can do that. Second, the integration of existing legacy content.
[00:14:25] I am fairly sure that there is a trillion dollars of existing legacy content out there in companies. Compliance documentation, training, onboarding policy documents that will be converted into learning experiences. We are going to take the JBA, our academy, and put a lot of those artifacts into Galileo. So you'll be able to go into Galileo, answer a question and then watch a video or take a course if you decide that's what you want to do. That's a massive issue in L and D that crosses into knowledge management. And so if you're in a big company and somebody's trying to build a capability academy on safety or a customer service or leadership, you're not going to just have to build a bunch of courses and teach a bunch of people stuff. You're going to have a knowledge system or an agent or an assistant that will do that. Lnd is going to be deeply involved in that. And I think l and D might be the group that runs that project. So that's a big sort of future direction. And then there's the issue of consumption of content. The learning experience platform, which is the sort of latest version of how we go find and consume training, which to some degree sits on top of an LMS.
[00:15:36] Most lmss have faded into the background, is going to be disrupted again. The degreed is now working on a new identity and a new strategy to do this. But imagine you had a chat GPT system like Galileo and you just asked it a question and it showed you courses or content, or maybe just answered the question and let you click in for more information instead of browsing through a bunch of topics and courses. And it could recommend to you. Would you like a quick answer? Would you like a video? Would you like a short course? Would you like a long course? Would you like a certificate course? Where would you like to start? What topic would you like to dig into? Right? And then the system would just, you know, kind of generate or take you to the content you need. That whole interface is going to cross between L and D and employee experience. So I think L and D is in for quite a ride here. You know, my experience working in L and D, because I spent ten years in L and D before I got into HR, is that L and D people are very technically savvy, they're very creative, they're comfortable with new tools. They tend to be creators and innovators at heart. They have an instructional mindset. And so I think we're going to see some very exciting changes. It does raise the issue of what is the job of the chief learning officer? Does the chief learning officer worry totally about training? Do they own knowledge management? What is the relationship between the CLO and the skills projects going on in companies? My initial take on the skills stuff is that I think companies are spending too much time on the skills architecture and not enough on the content architecture. I know we need to come up with skills models for recruiting and skills models for internal mobility, etcetera. But if you don't know how to use that interface or that metadata to deliver content to the right people at the right time, it may not be as useful as you think it is. So we're going to put a lot of time into this. We're going to have a couple of CLo roundtables. So I'll tell you everything I can on the podcast and in articles, but we would love to engage with you. We are starting another big reset session in two weeks. So those of you that have been invited or would like to join, let us know. L and D is going to be one of the panels we're going to work on for six weeks. We'll get a lot of information on that. And I'm excited about it because I found the L and D market to be just fascinating in the early two thousands and became rather boring the last few years because there wasn't much going on. I think it's really going to be exciting, very exciting for the next couple of years in that area. So that's kind of it for today. The chatbot world is really interesting. Galileo is out there. You can buy it from us. We're going to have a single user version before the end of the year. Right now you have to have at least five people to buy it, but we're happy to make it available to you in a team version and check out SAP and Joule. It's really state of the art stuff. And then sort of hang on to your hats relative to Microsoft. Microsoft has not only a lot of engineering team working on this, but they're working on this in the Microsoft HR function. So they are not only trying to figure out how to let people like us get involved in this, but they're also trying to do it for themselves, which makes it even more valuable to all of us because we know that the stuff we build with Microsoft will be tested by Microsoft for actual HR users. Okay, have a great weekend, everybody. We'll keep in touch and talk to you next week.
