Episode Transcript
[00:00:03] Last week I talked about the two approaches to implementing AI in hr, the application model versus the experimentation model. And this week I had the chance to spend a lot of time with LinkedIn at the LinkedIn Talent Connect conference and talk to people about the Hiring assistant, which is an application that LinkedIn developed. And many, many people have asked me, tell me, what are some of the applications that we should be doing in hr? It comes up over and over again because people know what the AI tools look like, but they're not sure exactly what they're going to do with them. And let me make a few observations about that today. First of all, the level of depth and number of applications of AI is limited only by your imagination. I know that sounds strange, but I absolutely am seeing this bear out in reality. And the reason is these generative AI systems basically can collect and analyze and assess information from vast amounts of data and generate vast amounts of outputs. So the number of applications to be developed is huge. And I don't think we're thinking big enough. And so what I'm doing with the team here is we're building a book on 100 use cases for Galileo. We've got about almost 70 of them written up and then we'll get to 100 pretty quick, mainly to expand your imagination. And let me sort of give you a couple of examples. Now, one of the simple use cases that keeps getting demonstrated over and over again is this boring use case of building a job description. Theoretically, what you do is you take a bunch of information about a job that was written down and you ask the machine to write it up. It's a writing assignment. But actually building a job description is not a writing assignment, it's a thinking assignment. Because what you really want to do if you're looking for a job candidate is think about all of the work. Not the job, the work, the tasks, the activities that people do in this role and the types of skills and capabilities and experiences and cultural behaviors that make that successful and put that into some document that will help a recruiter or a candidate decide who's appropriate for this job. That is not a simple problem because I've talked to hundreds of companies about recruiting for many years and they found things that proved to them that the competency based assessment or the skills based assessment or the credential based assessment didn't work because there were other factors that were more important. And that's why we have human recruiters is human recruiters know, know this in their heart of hearts because they've done it a lot and they don't really know how to write it down. So imagine a new form of job description creator that interviewed or asked through email all of the employees in that role to tell the computer what they do on a daily basis in a narrative and what they believe is the most successful part of their job and the thing that makes them the most successful and the thing that is the most difficult or the things that are the most difficult about the job that make it unsuccessful and let them talk and take those narrative discussions, maybe only interview the high performers, maybe only interview the managers, but you can interview everybody and put it into the model and let the model crunch on it and then generate the job description. That is a new somewhat what I call super worker use case. No recruiter is going to do that. They don't have time, nor would they even be able to assess all that information. They learn it through their years as a recruiter, of course, and it goes into their mind, but they don't do it every time somebody opens a new job. The same exact type of collection process is used in training, in onboarding, in performance management, in leadership assessment, in leadership development, in coaching. Many, many of the things we do in HR are basically personality or skills or job or role or task related things that we try to use some form of behavioral science or data to make better decisions. Well, those are filled with opportunities for AI and these are use cases that I don't think you guys are being imaginative enough about. So when you read the hundred use cases, it's really going to expand your mind. And they're everything from hiring, assessing, developing, growing coaching, deciding how much to pay somebody, deciding when somebody needs a promotion, putting together a development plan, creating a new career architecture for someone, and on and on and on. And let me say that this simplistic idea that we're going to build a skills model and do this based on skills is just ridiculous. Now I'm not against skills based hiring and skills based promotion, et cetera, but as most of you know, and I've been in the working world for 50 years, skills are not the key to success. You develop skills in a job as needed. You, your ability to learn and to develop yourself is a key to success. And you develop enough of a skill to make yourself competent, but you don't over skill yourself and become better at your job with deeper skills. There's a saturation point. So if you think clicking on a skill and using that as a tool to make all these decisions is the answer, that's not at all true. And so all the AI engineers and vendors listening to this, and I think you guys all know this, have to be cognizant of the fact that your models have to be more subtle and more essentially nuanced in their predictions. I remember when I talked to Ashutosh at Eightfold about this many years ago and he made comments to me like, we know what you're good at because of who you worked with, where you worked, and when you worked in a particular job or company. Because we know what tools, technologies, problems, solutions and other people were there at that time. Recruiters know that simple skills models may not. So these AI systems can do things that are far superior perhaps to humans in many of these applications. Now that gets to another topic that I think is important to think about, and that is this idea of productivity. Most of the use cases that Microsoft and vendors talk about are, I saved 35% of my week on meetings because the co pilot scheduled them for me or took notes or whatever it may be. And then you look at the data about what did people do with the other 35% of their time? And I have a study on this and I'm going to be writing about it. And what you find is they did more work.
[00:06:41] So they took their machine orientation to work and they simply cranked it up a little faster to get more work done. And the challenge that a lot of researchers are asking is, do we as people become more human when the machines automate things that we thought were routine, but we're wasting our time? And the answer so far is it's not clear. It depends on the person, it depends on the job, it depends on the company. I know in my particular case, because I run a relatively modest sized company and I'm an analyst by nature, I really treasure the opportunity to think about things. And when I have to sit down and analyze a spreadsheet, I get a bit of a kick out of it. But what I really want to do is understand the answer and understand the concept and where it fits into everything else I'm working on. And I can't do that if I have to spend an hour finding the data. So my mind is shifting from operational tactical work to thinking back and forth. And it's very frustrating, of course, when you can't do the thinking because you're stuck on the doing. Well, some of the studies show that that isn't quite happening yet it will. Just like as I said last week, the dishwasher replaced the washing of dishes and speeded up the time after dinner. What did you do with your extra time when you turned on the dishwasher, did you spend more time with your kids? Did you spend more time with your spouse? Or did you just turn on the TV and go to bed earlier?
[00:08:13] I mean, who knows? But productivity isn't necessarily the key here. So these studies that show that somebody saved 35% of their time is. They're hugely valuable because they're massive numbers. I mean, we've had people using Galileo that told me they saved a day, a week or two days a week on their work. But what excites me more is when somebody says to me, and actually the head of L and D@ LinkedIn told me this is she's using Galileo to create an innovation group on corporate training and come up with new models and ideas and approaches to building and deploying content within LinkedIn, because they have this tool that allows them to do new things. So on the flip side of productivity is the imagination idea that we're going to do different things. We're going to do more, not more of the same, we're going to do different things. I am convinced that's where this is going. Another, you know, sort of great example of a sort of tactical use of AI would be supply chain. Let's suppose you're a supply chain manager. I know a lot of you are not, but just imagine it, and you have a supplier that delivers some part that your company needs, maybe it's a really important part, kind of tricky to manufacture and they go out of business and you have to find an alternative. That is not an easy problem. You could go down the yellow pages and look for the next company in line and call them up. Or you could look at all the vendors that have been employed by your company and sort through which ones delivered the most stuff and look there. But actually, as you can imagine, just like with the recruiting example, there's a lot of subtle data that would lead you to figure out who the replacement supplier would be. AI would be really good for that, right? Or you have a contract from a client and you're about to sign it and you scratch your head and say, this looks familiar. Some of these terms and conditions I've seen before, should we accept them or not? The lawyers may know whether they're acceptable or not, but the AI could comb through all the similar contracts with similar vendors and say, this one's out of line. We shouldn't accept it because all of our other providers didn't ask for these terms, which is, by the way, a little bit what my life is Like, I'll tell you some funny stories about that sometime. Well, these are use cases of imagining a better company, imagining a better future, imagining a better business. And they're not going to be obvious to you until you think about them. So let me go back to LinkedIn for a minute. So the LinkedIn hiring assistant, which is still in its early days, is a meticulously designed productivity and innovation system. Because what it does is it really rethinks recruiting in some very subtle ways. And in some ways it's completely different from Paradox, which is also a recruiting agent that automates lots and lots of things in the recruiting process in the LinkedIn model. Imagine one of the way you sort of get started as a recruiter is you say, find me a person like Joe, and you point the system to Joe and the system looks at Joe, presumably analyzes Joe's job experience and what it can find out about them in LinkedIn, which may not be that much, but imagine that was happening inside of your company where you have Joe's performance history, job progression, salary history, performance ratings, and so forth. And it generates the beginnings of a description of a position. It then asks you if you'd like to tweak it. It allows you to put information in about your company, your company's culture, and then generate a better and better job description until you're ready. Then it will go out and look for candidates and it'll come back to you in a few minutes or a few hours. It will give you a list of candidates, and in each of the candidates profile, it will show you why it selected this candidate versus another. And you can literally scroll through them and see what criteria were used to select the different candidates. You can go back and highlight the criteria and make them better and sort through the candidates to get to the ones that are worth spending more time with. And it goes through other steps in the process. I won't bother going through this. You can read the article I wrote about it. But you can see even in the beginnings of this process that I just described, it's doing things that a supercomputer or a super recruiter could do, but maybe a casual recruiter may never do. So these are use cases that I know Hari, who runs engineering at LinkedIn, spent a lot of time on that. LinkedIn had to spend several years thinking about to really build out in a complete way. So when you think about these AI systems like Galileo or LinkedIn hiring assistant or Paradox or whichever ones you may end up using, the implications are much deeper than you realize. And again, if you can get your hands dirty with them as soon as possible, the faster you're going to learn. And let me sort of conclude with a third use case that I think is really fascinating and talks about imagination.
[00:13:07] So there's a bunch of companies working on coaching bots for some reason, I don't know why this is so popular, but there's this belief that we can create an AI coach that will be like a management and leadership coach.
[00:13:20] And of course, we all know we need more help with people issues and management issues and change issues and all that stuff. So what these vendors are doing is they're using models and they're training the system to have a dialogue with you about a problem. And I'm not going to mention vendor names. Some of them are domain specific, like Galileo, that knows everything to know about hr, sales, marketing, whatever. Some of them are more psychological coaches. Like a psychological coach. And the belief is that we're all going to have one of these things to help us. And I've been, you know, a little bit skeptical about this because I know in my career I've never really had a coach. I've had a few coaches here and there when I needed help with something, but I don't feel like I need a coach all day. Well, I think I told you the story about Joel Hellemark, who used sana, the platform that's underlying Galileo, and loaded every video and document he could find from Steve Jobs and uses his virtual Steve Jobs as a executive coach for him to make decisions at sana. So this idea of having a coach is out there. And I talked with a vendor I know very well, I will mention who they are later. And they have built one of these based on their own psychological coaching model. A lot of experience they have in the coaching industry. They were showing it to me. And the way it works is you ask it questions and it comes back to you and gives you advice and then it asks you as a user if you'd like to do a role play or some sort of a scenario. And it'll take you through a scenario and coach you on your responses to various situations. So it's a development coach, as if somebody said to you, what would you do if an employee did such and such? How would you react? And then it gives you feedback. And I was watching it and I asked the vendor, is this information stored? And they said, yes. The system is keeping track of your results and your responses to these dialogues and getting to know who you are and what you're good at. And what you're not good at. And presumably it has enough intelligence to make sense of all that. And I said, well, imagine now that everybody in the company, all the employees, had this thing, this virtual coach that you've created, and they used it for a few years or maybe even for a few months, but for some period of time. And we had these digital assistants getting to know who we were.
[00:15:44] And because this thing is used for a lot of soft skills questions, it's getting to know our personality and our nature and perhaps a little bit about our technical skills. And honestly, if they're trained correctly, they could get to know your technical skills. Because when you ask an agent like Galileo a technical question, presumably it knows that you're interested in that question and that maybe you already know a lot about it. I said, well, now this isn't a coach anymore. This is essentially a digital assistant of you.
[00:16:14] And it can be used for development planning, for succession management, for leadership assessment, maybe even for performance management and a whole variety of other issues. Because if these systems did, and I'm not sure that they do, but presumably they do keep track of all of the interactions you had with them. They become very, very knowledgeable about who you are, your strengths, your weaknesses, your situations, and to some degree, your capabilities and skills and aspirations. Now, if they're available to a leader or an HR department, the HR department could ask the digital coach, is this person ready for a promotion? Please develop a development plan for this person. What are some of the areas of improvement this person could make? Or suppose we're doing a massive M and A or redeployment or change in the company and completely reorganizing a bunch of people. We could talk to these digital coaches as a group and say, okay, guys, everybody in this department is about to go through a job change because we're going to do this new thing. Tell your partners, your coaches, that they're going to have to get ready for this and give them some preparation. Now, I'm not kidding.
[00:17:26] I actually think this is likely to start to happen next year. Now, the technology may not be mature enough to do this, and there's a massive privacy issue here of who owns this data. Because if my coach knows a lot about my issues or my situations at work, the International Coaching foundation essentially says that information is private and the only situations where that can be disclosed is if it's an illegal behavior or safety problem. When in the case of the AI, we may find a situation where a company has many one or many hundreds of digital coaches and Somebody in HR or in some other parts of the company wants to know what's going on. Now, this particular vendor told me that in order to run that kind of a query, you would get aggregate data. So you would see a summary of the problems or issues or topics that the digital coaches are dealing with without going into individuals. But you can see where this is going. And I'm not going to disclose who this company is or when they're going to market or anything yet, because that's coming. But you can see that the use cases here are much more creative and expansive than you might imagine. So let me leave you with that. We are very excited about all of this because of the potential. Galileo Pro is in the market now. We have thousands of users. I encourage you to get it. It's very inexpensive. It's your chance to really get to know how this all works. We have some new announcements coming. There's a new content being added and some new capabilities. You're going to be able to add an AI meeting assistant to take notes, a note taker to take notes in meetings. There is going to be a personality tool to tell Galileo what personalities you'd like to create. So you can create a Galileo for recruiters, a Galileo for L and D, a Galileo for business partners, a Galileo for leaders, things like that and some other capabilities coming soon. And you know, as an engineer by training and an HR person, I cannot tell you how exciting this stuff is. Hope this was kind of an interesting thing to get your mind thinking over the weekend. Hope you all had a happy Halloween. For those of you who celebrate Halloween and I'll talk to you next week.
