A Sneak Peek Under The Covers of AI-Fueled Recruiting, And Lots More

Good morning, everybody. This week I want to talk about talent acquisition and AI, what's going on there and give you a little bit of highlight. But just as a preview, what we've been doing over the last week, in addition to a lot of traveling, is we had a series of meetings with clients going through our AI blueprint, which we decided we are going to publish. It's going to take a little while. So if you're doing any interesting AI projects in HR and you'd like to share them, we're doing a lot of case study interviews right now. So please reach out to us. Cathy or I will talk to you. We did discover that of the 10 or 15 companies we've talked to so far, there's a lot of confusion about where to start. In fact, that's one of the biggest challenges in AI. And I'm going to talk about talent acquisition because talent acquisition is where there's the most mature number of AI vendors and there's the most sort of opportunity to improve things and get your hands dirty with this type of technology. Just in terms of general communications from us. I'm going to keep doing this podcast, which seems to be very successful. We've had more than 4.5 million people listen to the podcast so far, so I'm not going to stop this. I'm also going to be doing some short videos on our YouTube channel, which is a little bit messed up and needs to be cleaned up. We're going to work on that. And then I've created a Substack called Josh Berson Unleashed, which is a little bit sparse, but I'm going to put things on there that I think are a little bit more opinionated and perspectives that may not be general stuff, but things about the market, things about the culture of business and so forth. Okay. So talent acquisition. This is a trillion-dollar spend area with recruiters and talent acquisition tools and interviewers and background screening companies and assessment companies and outsourcing firms and ATSs and many, many things. There are many categories to this. I will be writing a long article about what's going on. Starts with sourcing. Tools like LinkedIn Recruiter, SeekOut, Findem, Recruit, et cetera. Skills inference and talent intelligence: Eightfold, Beemery, Gloat, Fuel 50, Phenom, Lightcast, Techwolf, Skyhive, Retrain, iMocha, others, including the big ERPs who have built a lot of talent intelligence into their systems. Job advertising like Indeed and LinkedIn Jobs. Candidate relationship management, which is a category that's kind of folding into the others, but it's still there. Beamery, Avature, Radiance, Smashfly, which has been acquired, Phenom, iSims, et cetera. Screening, which is tools to reduce the number of candidates. And there's lots of new stuff going on there. Paradox is doing a lot of this, HireVue, Findem, different ways to do screening. Pre-hire assessment, which is a very old market, but it's being reinvented. Hackerrank, SHL, Harvard, Pymetrics, CodeSignal, Plum, Metal Mercer, TestGorilla, a bunch of those. Video interviewing, which has become pretty standard. HireVue, Spark Hire, VidRecruiter, a bunch of others there, many of whom have been acquired. Interview intelligence, which is an interesting market. It's not very big yet, but it's coming where you keep track of interviews and who's good at interviews and what questions work and the relationship between this interview and that interview and the resulting quality of hire. Pillar, Clovers, Right Hire, Metaview, Paradox does this a little bit, Ashby, ESs, which is the foundation of this, which is a ton of companies. Salary benchmarking. By the way, you can do salary benchmarking in Galileo now. In the next release of Galileo, which is coming in March called Mars, we have global salary benchmarks. So you don't have to buy a big complicated benchmarking relationship. You can get it right out of Galileo. Onboarding, bunch of tools there. So this process, which was originally conceived as a supply chain process. By the way, you know, the thing I've realized as I look at the difference between agents and traditional HCM software is traditional HCM software was mostly designed in supply chain workflows, uh, pre-hire to retire, recruiting workflow, learning workflow, career workflow, uh, compensation workflow, et cetera. And that was a mimicking of the way businesses operated for many, many, many years. But now businesses don't quite operate that way. People are changing jobs and roles and companies are reorganizing very, very rapidly. So the agentic approach to things like talent acquisition is much more oriented towards quality of output and not integration of an end-to-end process. But we still need to integrate these tools because there's way, way, way too many of them. And I want to highlight a very important innovation from AMS, Alexander Mann, which is a very successful, large recruiting and RPO firm who we've done a lot of work with. AMS has introduced a product called AMS One, which is basically an orchestration system that stitches together all of your TA tools into one platform. They built it originally to manage their client projects because all of their clients have a lot of tools, but you can use it to see the workflow, the lead flow, candidate flow, candidate process, interviewing results, metrics, analytics of your entire process without having to buy one giant system from a vendor that claims to do all this. You have to be an AMS client, but it's a really, um, sort of spectacular thing that they did. They've been working on it for quite a while, and I've been following them for a number of years, and we're quite good friends with them. So take a look at AMS One if you're trying to pull all these things together. Now, the reason that this area is so interesting to me is the AI that's being applied in talent acquisition. Will eventually be the core AI that will do everything else. Because what happens in talent acquisition is we try to find people or source or characterize people based on their experiences, their skills, their languages they speak, their certifications, their qualifications, and of course, many aspects of their personality and their behaviors. And this is what recruiters do. Human recruiters who are good are really, really savvy at understanding people. And they're really, really savvy at knowing the job domain that they do recruiting in. General recruiters are not very good at specialized, high-skilled recruiting, but if you find a recruiter that knows IT, that knows sales, that knows marketing, that knows product management, that knows oil and gas, they know because they know the technology language, they know how to recruit well. And so what's happening in AI is little by little, these combinations of vendors, mostly in the beginning sections of the ones I just mentioned, are getting smarter and smarter at training their AI to identify these characteristics of people. Now, When you buy a generic tool, Eightfold, you know, the talent intelligence platforms from many vendors, you are sold on the fact that it'll assess skills of anybody. But in reality, it only assesses the skills of people it knows about well. It wouldn't be very good at assessing the skills of a biologic researcher who's studying the human genome if it had never been trained on that. It might come up with words. That describe that person, but they may not be the right words for you. So the process of improving these tools not only involves the vendors working with more and more different industries, but also labeling of the data. And I want to highlight a company we've been working with a lot called Findom. Findom, which you're going to hear a lot more about, is, uh, they just got about $40 million of funding. They're successful, relatively smaller player today. Does data labeling for talent acquisition. So if you looked at a profile of someone like me, for example, because we've tried this, you might think that I'm a CEO type. You might not know that I'm an engineer or that I have a lot of experience in data analytics and then I spent 7 years doing that, or that I've worked for a bunch of startups. Those are inferences you could make perhaps, but it wouldn't be easy because I haven't really put all that detail into my LinkedIn profile. Data labeling allows these AI systems to gain the benefit of experts to tag what the words mean and how the words relate to each other. Has this person worked in a high-growth company before? Has this person led a development team? Has this person experienced a downsizing or a global acquisition, et cetera. Those are things that recruiters just ask and they just know that, but the AI doesn't. So the AI sourcing and assessment world is getting smarter. And what happens is, as that happens, you know, another company that's big in this area is Machi People, is the AI becomes like a highly intelligent, highly trained IO assessment engine. And in the early days of this market, that was all done by hand. It was handcrafted by different vendors going way, way, way back to the ATS job matching where they basically looked at words, words in the resume versus words in the job description. But now it's very sophisticated. And now that we have LLMs from the open vendors, OpenAI and Gemini and others, It's getting even better. For example, Galileo, which we don't really position it as a recruiting tool, is extremely good at comparing people's skills and giving people development plans because it knows a lot about the skills in the job market with data from SHL and Lightcast in it, but also it's so highly trained on different job architectures and career models that it's just very good at giving people development plans. You can take two candidates, and by the way, you can even take their recordings of the meetings they've been in, and I think I mentioned I've done this, and you can assess their skills and capabilities and experiences from their meetings that they've been in because the AI can interpret what they've been doing in the company. So think about how that could be useful for internal, uh, talent mobility. So the big story in AI is not just a video interviewer or a video avatar or a candidate chatbot, all of which are huge opportunities, but it's the underlying ability for the system to give you much better insights into how this person's going to fit or not fit and what role they should be in. The analogy in the automobile self-driving car market is there's some engine in the self-driving car that is aware of the environment and picks up signals about the environment, which include weather, pedestrians, stoplights, and so forth, the smarter that is, the better everything else works. And that's really the same thing here. And there is no best in this space. In the early days of the market, Eightfold was definitely in the lead and they still are one of the leaders here. But all of these companies have tried to build these kinds of AI inference engines pretty well. And they're all getting better. So we'll keep you up to date on all of them. Now Workday just acquired HireScore and Paradox. They are building agentic features into Workday for this. And Salesforce and SAP just acquired SmartRecruiters, which is a very advanced AI-powered system like this. SmartRecruiters has an AI agent called Winston, which does a lot of the same things, maybe not exactly the same, but similar to what Paradox's Olivia does. So those two vendors both have a pretty big extensive range of tools here. In the case of SAP, they are now converting or migrating 5,000 customers as fast as they can from the SuccessFactors Talent Acquisition Suite over to SmartRecruiters. So SmartRecruiters technology, which will eventually be probably renamed, but at least for now is becoming the core stack for SAP SuccessFactors. In the case of Workday, they're doing similar things with agents. Let me mention another piece of technology that relates to TA and other things here too. I talked about AMS-1. Well, In the future, as we have agents and super agents, and this is what the blueprint is all about. By the way, we are going to write up the blueprint and build a whole research document on this. But you really should talk to us. We're going to do more workshops on the blueprint and anybody who wants to come, just reach out. If you think about a bunch of agents doing different things, all of these things I just mentioned in TA, just as an example, each one of them is highly trained. To do the thing that it's really good at. The sourcing thing, the skills inference thing, the interviewing thing, the candidate conversation thing, et cetera. Normally in the past, that would have been one big monolithic application. Now these are more like independent agents grouped together into super agents. Well, instead of buying AMS-1, if you perhaps for some reason don't want to outsource any of your stuff. You can use a tool from Workday called the Workday Agent System of Record, ASOR. And that thing is very interesting. And the reason I'm into it is because we're going to be working much more closely with Workday on Galileo, and Sana is part of the ASOR. About 2 and a half years ago, I was at a Workday analyst meeting and it was in the early days of AI. And I would say Workday was just trying to figure out what was going on and communicating a lot of their strategies to analysts. And I ran into Anil, the CEO, the founder, and a few other people and was just talking to them about it. And the story they were coming back with about AI was, yes, this could be disruptive. By the way, I've discussed this in a video, which I'll link you to. Yes, this is disruptive. Yes, we have to deal with it, but ultimately it's a platform problem. Because we have to manage these agents and we have to give them security. We have to decide who gets access to which one and what privileges each agent has. And then we have to leverage the data that companies already have to inform the agents and train the agents. And then in some cases, we need to store the data from the agents back into the transaction systems for payroll and record keeping. And he was right. That's actually true. The agents don't operate independently. They need to be coordinated. And he said, so we're going to build this layer that manages the agents. And I was like, hmm, sounds like an interesting idea. I thought to myself, you know, there'll be a lot of companies that do this. ServiceNow will do it. Microsoft will do it. Google will do it. A bunch of the IT firms will do it, but maybe Workday could do a good job of it because they know how Workday works. And for the 15,000 or 20,000 customers that use Workday, the agent system of record would be intimately aware of what Workday does and how it all works. And so you could essentially plug in one of these agents into the Workday ASOR, and then you could say to the ASOR, deploy this agent to all our job candidates in Australia and New Zealand, but nowhere else, or deploy it to our hiring managers in this department and that department, but not this department and that department. And if you didn't use the ASOR, that kind of management decision would have to be made in that agent. So all of these agents would have to have their own system to manage user access, authentication, security, and understand the roles in the organization, which, you know, it could be done with APIs, but it's a lot of work. And so this ASOR sits in the middle. And so the ASOR from Workday is an orchestration layer. That's the new buzzword, by the way, orchestration, that allows you to take all these agents and connect them together. And the reason I bring it up is that in talent acquisition, the big trend is to stop buying so many little things because companies don't have unlimited budgets. These tools are not inexpensive. And even though they add a lot of value, you would hope that they replace a lot of existing stuff and they simplify the process. We hope that they're more like super agents and not standalone tools, because each of these standalone tools is expensive, has its own data management and user interface, et cetera. So now the counter to the AS/OR in SAP world is Joule at this point. I would imagine they're going to be adding more capabilities around it. Is that Joule, the infrastructure that Joule runs on is the same type of infrastructure that Workday has announced called ASOR. Just education for you guys. Nothing urgent here to do anything about, but if you're a Workday customer, I think you should think about this. Last point, since I didn't want to go over 20 minutes, is the interesting to me thing about all this technology is it does enable you to rethink the operating model for talent acquisition. The traditional operating model for talent acquisition is the fulfillment center. You open a rec, we fulfill it. We are like the amazon.com fulfillment center for your job hiring. That's the old model. And it's still obviously huge and very important, especially in high volume recruiting, but a company that's been told to stop growing. And your CEO and CFO are pushing for productivity and they're freezing headcount, which is happening all over the place, or they're trying to shrink headcount, or you're going through a merger, or you're reorganizing, which everybody is, you don't want to just fulfill every job directly from a manager. You want to have some process of slowing it down and deciding what is the right growth strategy for that requirement. And maybe it shouldn't be a hire. Maybe it should be an internal candidate. Maybe it should be a retrained training or automation project. And when I talk to TA leaders about this, they're unanimously interested in pushing to that model, which we call the more talent acquisition as a growth engine, not a supply chain fulfillment engine. And I think as these tools become more modular and more interconnected, we can move in that direction. In other words, you could say to your talent acquisition agent, need capacity in sales. We need people who know this kind of stuff and this kind of stuff to go into this location to do these kinds of clients. Go find me some people that can do that in the company and outside the company. And let's compare the cost and benefit of that. And then a job architecture agent, which would be in a different part of the AI blueprint, would look at the jobs in that function and say, by the way, the way the jobs are set up over there is deliberately slowing them down because they're not doing enough cross-functional communication with each other. And if that existed, by the way, those are things that actually do exist, then the person who used to be a recruiter is now truly a talent advisor. And, you know, several companies are building, in fact, including ServiceNow agents. There's a people analytics agent over at ServiceNow they're using. So that this talent advisor could figure this out before they just, oh, you got a wreck. Thank you very much. Let me go try to work on it. Okay. Let me stop there. That's been 20 minutes. I have so many things to talk about. I'm going to keep doing this. I'll send you a link to the video on what's going on at Workday a little bit. And I hope you guys are having a good February. We're going to be in Unleash in March unveiling the new version of Galileo and also another big relationship we're working on. Then we're going to go to Transform and do some cool stuff there. And then I'm going to be in Europe in April at the, in London for a couple of days and then the HR Tech Conference in Amsterdam, if you're coming there. And anyway, if you have any interest in going through this stuff with us, please just reach out and we'll invite you to the next AI Blueprint session that we have for clients. Thanks everybody. Get Galileo. It has everything you need. Talk to you again soon.