Episode Transcript
Speaker 1 00:00:07 Hello everyone. Today I want to talk about our learnings about generative AI in the workforce and in business. And the title of the podcast is Generative AI is one massive global process of playing around because that's what's going on. So we just finished five weeks of interactive discussions with approximately 200 to 300 companies discussing their experiences, their fears, their worries, their uncertainties, their successes, their expectations in generative ai. And the main conclusion I come to, which we will be writing up in a research report, is that people are playing around because the technology is new, it's raw, it's unfinished, it's undisciplined, it's not focused on problems yet. And we've all experienced generative AI in our consumer lives. And it was pretty exciting at first, and then it became a little dull. And then as you learn to prompt it, it got a little better. So we all thought, or we are thinking and we expect that it will do the same in business.
Speaker 1 00:01:18 And so we've really had a lot of fascinating conversations about this. And let me share with you what we've learned. First thing is there are many, many separate use cases for generative ai, and they're all quite different. The use of AI to generate videos or images or audio for marketing, for advertising, for creating a course for your own artwork is one class of applications. And companies in the media industry and the advertising industry are starting to use it for that. There's the use of AI to generate documents, emails, text that can be used to generate emails, to job candidates, to write job descriptions, to write documents, to just do a better job of writing whatever it is. You're crafting lots and lots of tools for that. There is generative AI for search. Our generative ai, the HR co-pilot is actually a extremely good tool to find research, to ask the research questions that may be hard to find in a document and get interpretive content from the research.
Speaker 1 00:02:29 And so if you're building process documentation, process training, onboarding, training, training on how to use the system, training on how to do a process, training on how to work at the company, et cetera, et cetera, all of that can be put into generative ai. In fact, there's, I'm gonna talk about that in a few minutes. That's a big use case. There's the use case of analyzing data. Most of us have spreadsheets with tables and tabs and pivot tables and lots of different things we throw in there, and we have to manipulate the spreadsheet and work on it and try to get it to do what we want it to do. Generative AI is pretty good at that. You can ask it English questions about large amounts of data and it will find things pretty well actually. And, and then there's applications in training for chatbots that can be used for teaching assistance.
Speaker 1 00:03:16 Chatbots that can be used for transactional processing, chatbots that can be used for customer service chat bots that can be used for sales <laugh> chatbots that can used for lead generation and on and on and on. And guess what, what we're discovering is that every one of these is different and the vendors who are selling more or less generic tools haven't specialized very much yet. A few of them have, obviously paradox is specialized in, uh, recruiting candidate experience, but uh, most of them have not. So what you're starting to see, and what we're starting to see is generative AI add-ons to almost everything you already have. So U K G, the big payroll company just announced or previewed a generative AI tool to build job descriptions and emails to job candidates. Eightfold has announced something like that. Beamery has announced something like that. Phenom has announced something like that.
Speaker 1 00:04:11 I'm sure Workday's gonna announce something like that. SuccessFactors is doing something like this. So these generative AI plugins are appearing all over enterprise applications. Now, when you ask companies what they're doing inside of their companies with this, you hear lots of stories. And for the most part you would say there's three categories of HR awareness of this. One is people that still aren't really sure what it is and they wanna learn more. So we shared a lot of information on resources and tools and training and so forth. And one of the companies we were with this last sprint was a large pharma company that, you know, and they started a company-wide education on generative ai. They built a bunch of courses and they're asking everybody in the company to take these courses and start playing around and learning how to do prompt engineering. Because if you're not aware of this, one of the tricks to, to using generative AI is prompting the system to give you the information you want.
Speaker 1 00:05:11 That can be programmed, of course, but the user has a lot to do with that. So that's about maybe a third to a half of the companies. Another maybe 20 to 25% of the companies are not doing anything and they're intimidated and they're not sure what to do. And then another 15 or 20% of companies are doing pilots, they're experimenting. In fact, one of the big C P G companies in the group is doing a prompt of hon and what they're doing, it's really good idea. They came up with 22 use cases that they believe generative AI can be used for in HR and some of the things like I just talked about. And they have 10 to 15 people working on each use case, and they're gonna play around with, uh, a chat B P T implementation, and they're gonna come back as a team and they're going to share what they've learned.
Speaker 1 00:05:58 By the way, this is what we are doing in our copilot. We have people now testing it, and we're gonna do more testing before we offer it to everybody to see where it's strong and where it's weak. Now, the reason I call this playing around is because we don't really know what this is capable of, nor do the vendors for that matter, <laugh>. I mean, if you listen to Satya Nadela or the folks at Google, of course Google has a very defined strategy because they wanna use it for their search process and their search engine. Google, uh, Microsoft's different. Microsoft doesn't really have a big search business, so they're expecting their co-pilots to be productivity tools. Well, they don't really know how it's gonna be used for onboarding. They don't know how it's gonna be used for training. They, they haven't built all these use cases out, and each one of them is different.
Speaker 1 00:06:45 So the vendors in a sense are te selling raw technology. I've used this analogy before, but lemme go back in time when I was, I guess it was the 19 early in 1980s when I left i b m 1990s, excuse me, I went to work for Sybase. And at that period of time, there was this massive growth in the relational database industry that just started. There were two dozen, uh, R D B M S vendors. They were building products on Unix and on different, uh, platforms. Oracle was pretty big, but Oracle was kind of a clunky company at the time, and nobody really knew what these things were gonna use for. It turned out that some of the relational databases were used for transaction processing. The company I worked for Sybase was good for that. Some of them were used for data warehousing, Teradata and Oracle was good for that.
Speaker 1 00:07:30 Some of them were used for object-oriented applications. That was ingress. Some of them were used for very high volume transaction processing that was tandem and informix. And so the vendors within about two years figured out what part of the market, what part of the use cases would be best for them, and then they refine their products to meet those particular areas. Now, two decades later, the relational database industry is much, much more mature. And the products do you know, mostly everything. And they're much more generic. And the, the number of vendors has shrunk because it is a very complex domain. But the gen AI is way, way back at the beginning of this, where none of these products are specialized yet. So you're basically buying raw technology that the vendor doesn't know what it's gonna do for you yet, and you're gonna have to experiment with it.
Speaker 1 00:08:21 And that is what's going on in these companies. Lots and lots of experimentation. Now, that is actually both good and bad. The good is that you're gonna learn a lot and you're gonna discover use cases that are great. So for example, the typical use case in HR is the HR self-service chatbot. Locate my uh, number of vacation days in Workday and please enter my vacation for next week. Look up my 4 0 1 K and send me the balance. Where's my badge? I lost my, you know, card badge. Can somebody send me a new one? Who's the specialist in this area? Well, you know, it starts to blend because it's just like the employee experience platforms. Some of those are transactional applications, some of those are informational applications, and some of those are knowledge based applications. So you're gonna have to experiment yourself and look for vendors that specialize in different areas.
Speaker 1 00:09:13 And I would be willing to guarantee that in October when we all go to the HR tech conference, the one in Europe and the one in the us, we're gonna see a lot of this stuff and they're gonna look great, and the vendors are gonna have beautiful demos, but I would be willing to guarantee that 99% of them are very, very early in their product maturity and they're gonna be counting on you to test it and use it in the way that will help them refine the product. Now, just to give you an example of what's going with us, so I'm gonna give you a little peek under the covers here. So we've been working on our copilot for just a couple months, not very long. We've put all of our research into it, all of our blogs into it. We haven't put all the J B A stuff into it yet, but that's coming in.
Speaker 1 00:09:57 We've been asking questions, and there's some things it's exceptionally good at. I mean, it's really good at answering questions about vendors and asking questions about business practices in HR and looking for case studies and finding maturity models. It's actually really good for asking it, what are the best practices in this? What's the implementation plan for that? What it wasn't very good at yet, and we're fixing it, is asking it questions like, show me a list of all the reports that the Berson company wrote on blah, blah, blah. It doesn't know how to do that because we didn't teach it what a report is. We just gave it this big corpus of knowledge. And so my point is that as an IT department, as an HR department are gonna have to refine these things, and the vendors are learning as fast as they can. One of the funny stories that came up in the, in the discussions was one of the companies said, the IT department is always coming into these meetings scared because they're afraid we're gonna ask them a bunch of questions about how this works.
Speaker 1 00:10:54 And they don't know either <laugh>. So there aren't that many specialists out there. In fact, I'm not sure there are any, everybody is good in understanding parts of this, but I have yet to find someone who really understands the whole big picture. There will be specialists that'll really get it in different domains. Let me give you another example of a specialized implementation that exemplifies this immaturity of the market. So a couple months ago, there was an announcement by SuccessFactors that they were partnering with I B M to build a co-pilot like chatbot for SuccessFactors. They're doing a lot of stuff with Microsoft, I know that. And Microsoft uses SuccessFactor. So I would anticipate you're gonna see some, you know, really nice implementations of the Microsoft co-pilot with SuccessFactors, but they're also doing something with I B M. Well, I B M has rebirthed the Watson brand.
Speaker 1 00:11:45 It's now called Watson X to kind of give it some new life. And one of the things they've built is a product called Watson Assistant, and another one called Watson Orchestrate. The Orchestrate product is a process manager, more along the lines of ServiceNow. Other tools, like first up, do this. Some of the other employee experience platforms have process management tools. Process management means you can just create a series of steps, not as a programmer, but more as a visual designer and say, I want you to log in here, do this, log in here, do this, do this, take this data, move from here to there and do that. Well. So they have bolted together, the Watson Assistant, which is the English language front end to Watson Orchestrate, and built out a solution that understands all of the business processes in SS A P, not just success factors.
Speaker 1 00:12:37 Now, I don't think this is out yet, but if you've ever read SS a P documentation, it's like getting a PhD in business <laugh>. I mean, every business process that anybody's ever invented, SAP's tried to automate and they wrote it all down. They're just sort of amazing engineers. So working with I B M, they're building a series of orchestration products to sit on top of different SS a P applications. So let's suppose you're an SS A P customer and you go, I would like to file my expense account and send it to my manager. Click, click, click chat bot says, entry expensive cloud here. What is the date? Et cetera. Sends it off to Concur. Concur sends it to your manager. Your manager approves it, it comes back, you get, you know the check and all of that is gonna be orchestrated. And the way this works is the chat bot now has to have essentially three features in it.
Speaker 1 00:13:28 And this is really my big learning about this stuff over the last couple of weeks. First is the search. They have to figure out what you asked the chat bot and where to go for that information. What application system, what database, what content. Generative AI is pretty good at that, but much of the engineering that's going on in generative AI is based on search. That's what binging is all about. That's what Bard is all about. And our engineering guys have found that there are eight or nine specialized new search tools based on generative ai. Once it does the search, it has to give you an answer. So then there's the English language generative response. How do I find the document that you're looking for? Understand what you asked about it and give you a narrative answer. That's the second piece. Then the third piece is the orchestration.
Speaker 1 00:14:24 Now, what do you wanna do? What is the transaction you want to execute? Do you want to enter information? Do you want to click on a button or do you just wanna open a document and read it? So these things that look like simple English language analysis tools are gonna move into these three domains. They're gonna be doing search, they're gonna be doing generative content, and they're going to be doing transactions. And so they're gonna become complex applications. So don't let me disappoint you here, but we're getting into basic IT stuff. We're gonna have to do data management. We're gonna not have to look at security. Who has access to this? Who has rights to what? Do the system know who the user is? And is the system gonna give the user the appropriate level of access to information? If I ask the system about the CEO's salary, it shouldn't tell me that unless I'm the CEO's assistant.
Speaker 1 00:15:16 It's going to have to understand its relationship to the other, you know, applications inside of the company and all of these security implications of that. It's gonna have to pass data across those applications so you don't have to log back in and tell it who you are again. And then it's gonna have to have some form of user experience to give you a relatively small amount of information to respond. And all of that is IT stuff. This is what IT departments have been doing for decades and decades and decades in integrating different systems. So these are not just tinker toys for kind of asking silly questions and seeing what's going on on the internet, but they're actually gonna be very sophisticated application systems. And we are only six to nine months into this. I mean, it's now a middle of the summer and the chat G P T was only launched last fall, so we aren't even one year into this evolution.
Speaker 1 00:16:08 So what was inspiring about all of these conversations is that companies are really experimenting a lot. They're playing around in a very positive way. And I would warn you a little bit about the vendor market that if you haven't been playing around and you're not willing to play around and you don't have somebody to help you, you could get goaded into buying something that's not mature or not appropriate. It's a little bit of a dangerous time to be on going out and buying new tools, but I know we're gonna have to do some shakeout anyway. And in any one of these new technology cycles, there are some companies, many companies that will buy great stuff, they will learn a lot, and then they will replace it two years later with something that's more mature, and that's perfectly fine too. The final thing I'll just mention that comes up in these conversations with companies is the education and skills of you, of your users, of your IT department and so forth.
Speaker 1 00:17:03 I would encourage you to read our AI research report getting under the covers of ai. You really have to keep up on this. I mean, I encourage you to read articles, listen to podcasts, pay attention to the vendor market because it's changing very, very fast. And your incumbent vendors are going to be introducing things that you're gonna probably have to try because they're gonna be built into the systems you have and you want to be able to ask them good questions. Because if the vendor you have, say it's Workday or Oracle, or whoever says, oh, we have a new generative AI feature, and here it is, and here's how much it costs, you may decide to use it just because it's there, but you may not know that there's a much better best of breed solution in the market being developed by a more specialized vendor.
Speaker 1 00:17:45 So just like all of the other features of HR Tech that have evolved over time, the bigger vendors wanna do everything of course, but there's so much specialization and learning to go on. It is not clear, at least to me yet, if any vendor is gonna be good at all of this. You know, and the final thing I would sort of just mention is if you read the article, you know the book that we wrote on getting into the covers, we categorize the market into three parts. There's add-ons, generative AI add-ons, there's features to current systems, machine learning features to core E R P systems, and then there's full-fledged neural network based talent intelligence systems. I think you kind of have to think about the underlying architecture when you look at these things. What is the L L M that they're using? How has it been trained?
Speaker 1 00:18:34 If they have an L L M, how will it be maintained? Is it, if it's open source, is it one of the open source tools that's been proven to be useful in the application you're using? It's hard to get under the covers of these things, but ask those questions and let the vendors tell you, and you'll get smarter much, much faster than you think. The final thing I'll say before I wrap up today is yesterday I posted the keynote speech for irresistible 2023. It's about a 40 minute YouTube. It's on the front page of the website, and it really details the post-industrial economy and the implications on business and HR and many, many other things. And I encourage you to watch it this fall in about six weeks. We're gonna have an event at the HR Tech Show in Vegas and another one at the event at the LE Conference in Paris, and we're gonna unleash more details about that research and what it means.
Speaker 1 00:19:30 But I wanted to make that available to everybody who didn't come to the conference to just give a sense, get a sense of why that topic is so important, and we'll be continuing to talk about it over the next couple of weeks. The superclass and org design is now live in the J B A and it is unbelievable. It is just one of the best courses we have ever built. So for those of you that are not yet J B A members, this one class justifies the whole price, which is only four $95 for all of the content, all of the courses, all of the activity for the next year. I encourage you to check it out. There's about 500 people in it as of today, and you'll be running for about six to eight weeks, and then there'll be another version of it that'll start later in the fall. So I just wanted to point that out to you too. Have a great weekend and I'll talk to you guys all next week.
