Vollständiges Transkript von Applying Artificial Intelligence in the HR Space with Zach Frank, Senior Manager of People Analytics at the Freeman Company

reach there and then send me all the barcodes of the freight that they scanned and do that 12, 15 times a night. And so it went through a couple of tests, went through a couple of iterations.  But I fell in love with it. I fell in love with technology, the transfer of data, the ability to then analyze and say, “Oh hey, we're on time for you 97% of the time. The fact that we are 30 minutes late today, not a big deal. There is a big wreck on the interstate or something like that.” And that was really my start in the discovery of all these new things that I could be doing that were really interesting to me and got me started on the pathway. Dayle Hall: Interesting. That may have been how you started. But what I see as success within this space around automation, using AI integration is if you start with a business problem, start with something you're trying to solve and then figure out are there elements of process, tools and people. How do I solve it? More often now, I hear from customers that that's how they become successful, and that's how the initiatives are successful. I think that's just a perfect description is whether you knew 20 minutes in that this was just not going to work for you, but you were solving a specific issue. Zach Frank: Yeah, very much so. A lot of the machine learning practitioners that I follow and listen to right now, there's this phrase that's been developing around. It's not machine learning, which is another way of saying it's not an AI problem, until you can prove otherwise. So first treat it just as identifying what your pain point is, what is the friction. And then you've got a range of options in terms of how you go about it. And AI could be one, but there could be simpler forms of automation. I worked for Bridgestone before I worked for Aramark. One of the biggest improvements we saw in the HR space was just in reduction, so simplifying policies, removing parts of a process where unnecessary.  We saw bigger gains from that than we did from anything. But that came from truly understanding what is the business problem that we're trying to solve and then what is the range of options in our tool belt. Of which, automation is one. Of which, AI is a specific kind of automation. You can get deep into that too. It's a tool that we have to solve problems. And the better we understand those problems, the better we can choose, well, what's the right tool for this situation and our environmental context. Dayle Hall: I love that concept, not everything is an AI problem, more simple automation as you get going. So hopefully, we'll dig into that a little bit more. Talk to me a little bit about what should people understand about your role of people analytics. What does that actually encompass? And how do you actually help the business with that? Zach Frank: People analytics, in a technical or functional sense, in its most simple, is answering questions like how many people work for our company? In its most high-powered and really intelligent is how much does it cost us to have an open position. And in five years, what do we think are the areas of skilling that we're going to be in the most critical need of, where do we have people who are going to retire? In terms of business development, what's going to be really important for us in terms of the skill sets that we don't have right now?  A simple example that I think of in the HR space is back a few years ago when I started as a program manager for Bridgestone, overseeing their digital transformation and HR. There were very few HR-specific project managers. And there were very few people in HR that had any project management experience. Well, now if you look across that, there are people who specialize just in HCMS implementations. So it's that kind of thinking of where we're going to be, but that's really at the high end. There are very few companies whose people analytics chops are in that kind of a state. Dayle Hall: I can imagine though being in a B2B tech-type company in software, it's very Silicon Valley focused. You walk down the street, you're run into three VCs. But I imagine in companies that are potentially like Aramark or heavy manufacturing or where you have a lot of service workers that are moving around, understanding, “Hey, are we going to have a problem in three years because 10% of our workforce may retire?” I would have thought that really helps business planning and actually getting ahead of some of those potentially big impacts on the business. Zach Frank: Everywhere that I've done or been involved with people analytics has been very much of that. Bridgestone, which has manufacturing and also operates 20,000 retail locations; Aramark, which has tens of thousands of food service and janitorial services; and now Freeman, which does live event support, so people on the ground, getting things set up and all that. You do see a lot of more short-term focus because you've got a lot more hourly roles and roles that are high-manual labor, not necessarily low skill at all, very, very high skill. But you tend to see shorter tenures and things like that. And so you're looking at a smaller time frame.  So to your point about, say, three or five years out, actually for software companies, that's a really good stretch of the space of, hey, you've got an engineer who's approaching their senior timeline. Do they understand their pathway for internal development? Or do they feel their best way to get their 30% pay hike or higher depending on where the market is, is to jump ship because the senior bump internally isn't going to be enough for them? Dayle Hall: I have to tell you and I'm sure most people that I talked to or you would talk to today is recruiting and retention is one of the hardest things to manage. I don't care what company you're in these days. It's just really hard to make sure that you can find the right people at the right time and give them the right opportunities. Zach Frank: Absolutely. For

months now, we've been in a space where the unemployment rate is as low as it was pre-pandemic, but the labor force participation rate is also lower and the job openings have been up over 40%. So we've been in the toughest market that pretty much anybody has ever seen. Everybody is just going through that, “All right, yeah, how do we get the new people in?” when TA and HR talent is in such demand that people are getting picked off there, much less the roles for actually doing the things in the company that we need to do. Yeah, it's a very interesting time to be in this space for sure. Dayle Hall: So that brings us on to our first main topic, which is the space of the people, human resources, and there's clearly a lot of benefit in investing in some kind of the analytics-type tools or AI or different types of automation or data management. How do you see the investments that have been made in this area? Has it gotten more intense recently? Are people really looking at this area? Or is there still a reticence to invest in this type of technology for HR? Do people really feel it's a hot space? Or is it still trying to push that boulder up the hill? Zach Frank: I would say both of those things are true. So in general, the HR tech space is expanding and growing significantly. There's a couple of different studies that have gone out in terms of HR tech spending that’s expected to increase fourfold over the next five years. There's a lot going on in terms of companies that have used distributed systems moving to globally integrated human capital management systems, investing in new tech. The space is growing significantly. At the same time, HR tech has been so underinvested in over the years that the ground that has to be made up, means that, yes, there are a lot of new players who are coming in who are offering more advanced services in the HR space, but they tend to be implemented in half measures.  And companies take steps, dipping their toes in things and seeing what's out there. Because there's so much ground to be covered that's taking up money in terms of the basics of, hey, do we have all our employee records in one system. Much less, well, can we see how well matched somebody is for a job based on what they did and didn't put in their resume? So really, you're seeing both, I would say. Dayle Hall: To me, as we just talked about, solving a business problem is usually the best place to start when you look at what you want to implement in terms of new process or in people, particularly with tools. And we just talked about there are a lot more growing issues around retention, around recruitment, around really managing the people side of the business, yet it does feel like, whilst there are a lot more people coming into this space with the technology, they’re still a little bit reticent. I've seen in the last few companies a little bit of a reticence to invest in that area. And I think that the work that you're doing helps to show that there is a lot of benefit to put these technologies in. Zach Frank: Absolutely. Something that's interesting for me is I see the most investment in recruiting in terms of AI technology. That's where you see the most players. That's where you see the most diversity in terms of what parts of the process are people trying to solve, where is the money going, what companies are people talking about. What's interesting about that is I think that's absolutely necessary. I think it's a spot where there's a lot of help that can be offered. There's a lot of problems to be solved. But still, what I think most companies struggle with at a more fundamental level is that your money spent on retention is far more valuable, maybe not as urgent or necessary as recruiting, but your money spent on retention is more important.  And there are far fewer players operating in that space in terms of AI, in part, because it is in some ways a more difficult problem to solve and one that companies may not even be operationally bought into. They may pay the lip service of saying, yes, we recognize that that's important, but we're going to put money into how do we keep candidate outreach to make sure we have low drop-off between higher completion of background checking and the start date or something like that. So you don't necessarily see it in terms of where the investment is.  Dayle Hall: I could definitely see that happening. I actually talked to a company about a year ago called Betterworks, and they do something called OKRs, which is objectives and key results. And I think some people don't necessarily see that as that actually helps retention because it helps employees to understand how their work rolls up to the big company goals. And if you can actually show people how they're helping to drive the business and the top level initiatives, they feel they want to stay. They feel more connected to the business. But that's still few and far between about how people see those technologies is helping on, as you pointed out, a very critical area, which is retention.  When you go through these processes and you're deciding whether to implement a technology, there's this sense sometimes internally within organizations, whether it's generally within the IT organization around whether you buy something or you build it in-house. If you were advising a company now, and I'm sure Freeman will be taking advantage of your experience here. If someone asked you this question, how do you decide whether it's something we need to buy, something we need to build? Or is it actually really necessary? How do you start that process within the HR space?  Zach Frank: If you've answered all of your questions that lead up to this decision, we can't simplify the policy. We can't correspondingly eliminate the work. We can't make process changes or things like that. If you've reached the point where, okay, we need automation and potentially, specifically AI, I would say basically never build in-house, and that

may not be totally true. There could be instances. If you've got really great capability, if it's a really simple solution. But I think your two real options in the HR space are deploy or configure internally. So you may have a tool or resource that you're not using internally, where you could say, oh, we just need to set up these workflows in our HCMS because we never operationalize this workflow or whatever it is.  If that's not the case, if it's not something we already have those tools on a shelf in-house or being used in another department, buy is almost always going to be the better choice. There are just too many issues that come in with building AI solutions internally, specifically in the HR space. There are way too many issues. It's way too fraught with things, and there are so many good players who are out there who are relatively inexpensive. Even the companies that we would think of as saying, oh, yes, they absolutely could do this. Chip manufacturers, high-tech companies that have big data science teams and all this capability, they still outsource this stuff because it's easier and smarter to do so. Dayle Hall: And that brings us on to the next subject that I wanted to hit, a very delicate topic around AI. So the two things that I hear a lot around putting in AI, specifically around these kinds of companies and the impact on people. One that I hear a lot is, well, AI is going to remove people's jobs. My principles have always been AI and automation should make their job easier, should take away the mundane. But it really does then free those people up to do more interesting, fulfilling, potentially more business impacting work. I do feel that's slightly shifting, and I don't know if you still hear that. But the most important thing for AI in this arena that you're in, I think, is the concept of bias and how do you monitor and manage or be cautious of AI and potential bias, particularly in the HR space. So I'd love to get how you think about that and what you've seen. Zach Frank: The other thing that I'd offer too is probably one of the biggest barriers is just data availability, which I think is tied to bias. But if you think of something, okay, who are the right people to hire? There's an element of bias in terms of your data that you're generally going to get back the kind of people that you have hired before, which, to your point, if you're a software tech company, yeah, maybe it's a lot of white males. And so you'll over-privilege their resumes as opposed to women and people of color historically who underreport their capabilities and skills on their resumes. Having solutions that have that thought of bias in mind but then also have the scope of data.  The biggest employers in the United States hire maybe 100,000, maybe 150,000 people a year or something like that. Well, across the millions and millions of hires that take place, the availability of data that you have versus what an external company who's able to pull this data

from multiple sources, it's never going to be as good. And it is more likely to have that bias. The other thing I'll say is just because you go with a vendor, it doesn't mean that they're not going to be biased. I will never tell who this is, but there was a vendor that we were talking to and they were doing candidate matching services. So okay, is this candidate right for this job? So I asked about it. Okay, well, how do you protect against bias? And they said, oh, well, we don't consume demographic information. So there's no way for our system to be biased because we don't know what race or gender or disability somebody is. You just don't know how big of a problem you have. You're burying your head in the sand. Dayle Hall: Exactly, wow. But I think that's an example of something that has to get solved. I can appreciate why they had that initial impression. But it's really good that as they start talking to people like you, you can help them understand why that is not avoiding bias. In fact, it's probably perpetuating it. Zach Frank: Yes, absolutely. You'd see that if you're looking at any of your D&I or hiring splices on D&I through your dashboards after you went through an implementation. It is just taking in blind data and isn't doing any inference based on important differences in how people behave in regards to that demographic information. It's going to exacerbate the problem, not solve the problem. So you compare that with a company that we looked at who was actively accounting for that difference that I was talking about earlier in terms of how women treat their experience on their resumes and underrepresenting themselves and actively correcting for that. And the difference becomes immediately obvious in terms of how that ends up impacting your hiring and your ability to be fair and equitable. Dayle Hall: That's a great story. So as you're going through this process, your area, your field in particular, has to be a lot more cautious of that. So when you're going through a vendor selection, Some of the things are obvious. When they tell you those, this is how we do it. You’re like, no, hang on. That doesn't make sense. Let's rewind that one for me. But how do you go through an assessment then? Is there something specific that you look at when you're looking at different vendors or when you're implementing these systems that says we have to build in -- I don't want to say bulletproof because I don't think anything is 100% bulletproof. But what do you build into your process on vendor assessment? What do you build into your process when you implement? Are there some specifics that you say we have to do these things when we go through the process? Zach Frank: Some things that are helpful. Having a conversation like the conversation I have with those two different vendors, ask them the questions, have it on your mind, care about it. Getting that in front of them and having a conversation is another big one. Another one that I always do is I ask for any published scientific research that they've done. So there's a vendor that presented us a technology that seems impossible to me, even as I did it. You just click on these little pictures and it did a personality assessment just based on how you clicked on these pictures. And there's just pictures of pieces of color, not even any real pictures or anything like that. So I just asked to see what research they had, and they sent me the layers of research of, okay, here's the undergirding color research. Here's how it's specifically researched in terms of how it's applied in our solution and the ability to make determinations based on that underlying color research. I don't read everybody's academic papers that they send me or things like that. But if they're willing to send it to me, that's already a good indication. If they're hemming and hawing on there like, oh, trade secrets. We’re like, okay, best-case scenario, you haven't built it yet and you're just trying to get me to buy it to fund you, which could be fine, but let's be honest about where things are. So caring about it is a big one, asking those questions and just seeing how willing they are to engage with you and showing you how the sausage gets made. Dayle Hall: Yes, that's a great example of we as technology buyers. I think I'm open to purchasing software, purchasing tools even if they say this isn't fully baked. I like the fact that they would let you see under the hood. I think if any vendor says, well, that's a best practice or that's IP or that's under the hood. If they're not willing to show you, you should probably have some concern. Zach Frank: Absolutely. And a simple one that I run into all the time of people I noticed is like, okay, where is the connector download for us to get our data back? Just our data, I'm not asking for anybody else's data. And it's surprising the number of times that they'll say, oh, well, you get 30% of the fields or something like that. But yeah, it's a problem that is not uncommon. And it's really surprising to me. It's our data in the best of circumstances. And then the further down the road you go, maybe it's like, okay, well, if you can't even explain to us the basics of how your algorithm works or whatever you want to call the secret sauce that you're working on, how can we trust it? How can we trust it on an operational level? How can we trust it on an ethical level? The openness around these things is a big indicator in and of itself.  Dayle Hall: Slightly off topic. I was at a company and we acquired a company called Klout. I don't know if you remember that.  Zach Frank: I've heard the name, yeah. Dayle Hall: So that was an online social scoring of the things you posted and shared. And the principal at the time when social media was becoming big, it was massive. But as you dug into it, it just became clear that it was very hit and miss as to how things were being assessed and whether that was real. And again, it was a good idea at

the time, but that's a good example of it can't just be we'll tell you a score of who you are, but we can't tell you how it works. You've got to have more details on that. And the founder and CEO, who is a great guy, he started a bunch of companies since, if he ever listens to this, I'm sure he's going to respond and say, no, this is how it works. That's a little bit off the topic.  Anyway, let's move on from the AI and the unbiased and the controls and making those assessments. Let's look at how you then use the people analytics, the things that you create, the output to actually improve things. So obviously, one of the things that people want to do with their data, okay, they capture it. The thing that annoys me the most personally is when someone tells me we're going to create all this data, and I have no idea what to do with it. It just sits there in their tool, are they going to send me a bunch of reports and i.e., they don't have the knowledge, the time or the people to actually do something with it. It's not even analysis paralysis.  I just get the data and say I don't know what to do with this. I don't know what to improve. What should I change? So I think that's a big problem overall. For people management and with people analytics, I can imagine you could get lost in what you're being shown. So how do you manage it? How do you actually then take the data that you're getting and use it? Zach Frank: It is very situational because you have lots of hurdles that you have to overcome in the people analytics space, which I will say I love it. I hope to never have to do anything else for the rest of my career. I also recognize there are a number of difficulties to doing people analytics. If you think about people analytics versus marketing or product analytics or something like that, you tend to have an audience that is less technically experienced. They're very smart. They're very good at what they do, but you have to do a lot more translation. You can't speak at a purely technical level almost at all.  Another thing that you see are a lot of practitioners in the HR space who know how to do the functional elements of their job well but they don't necessarily think about it in a larger process orientation. So thinking of recruiting as, okay, yes, there are definitive steps if you're struggling with a requisition. Here are the three things that you can go and do that have the most provided value. For a lot of companies that I've talked to, there ends up not even being truly a holistic picture for what does HR do. Does it just solve responsive questions of the business? Or is there actually a focus on, no, here is what good management and enablement of our people looks like?  Not a lot of technical experience. You can often see difficulty in your users, your audiences, your HR members not necessarily having a defined picture of what good looks like for them. So you're having to have more fundamental conversations about, okay, well, how would you

know if a recruiter was not doing well? And the other thing that you see is going back to what I was saying before about underinvestment in tech. I've probably talked to a couple of hundred different companies and gotten some level of insight into their people analytics space. And I would say even now, less than 20% or so that have a data warehouse, maybe they've got a data lake house if they're really lucky, where they're getting these externally dumped files and then you connect those files to a BI service or can run some custom code on them if you're really lucky and have a practitioner who can do that. It's got a long way to make up in terms of technical sense.  You've got these big hurdles that you have to work with. Now all that doesn't really answer your question. It just sets the frame for why is it difficult and why is it so important. I think something else that we do as people analytics practitioners and analytic practitioners in general that shoots us in the foot is we get overly reticent about providing recommendations. So in this lack of clarity, in this environment of difficulty, we get more hands-off and say, okay, well, here's the data. You figured it out. You're the localized expert. I think some of that is wise.  But I think there's also just an element of fear there where we either haven't done the due diligence to understand, okay, how does our company make money? And then how does our people picture serve that? How does it serve our shareholders, our stakeholders, our business executives, our customers? The more you understand those things and then can be willing to just take a few steps forward and say I'm going to make some recommendations. And I'm going to stick my neck out there a little bit. Because otherwise, what I'm doing, you get pushed further and further back until you just become a reporting and records organization. It's how efficient can we make the active employee list or something like that.  I hate to come back to this because I know I said it before, but some of it is just caring about it, recognizing that the problem is there and being willing to step into the fray and say, all right, I'm going to engage with these challenges, and I'm going to be willing to stick my neck out a little bit and say it looks like this is what the data is telling us we should do. Dayle Hall: From my experience, again, different to yours, but from other tech in this space and from working with a lot of HR leaders over the last 10, 15 years, what I feel we need is -- I don't want to say case studies or best practices, but people who are using the tech, who are using the data and have meaningful business impacts. I think the more that comes out, the more others can see it, the more people will say, hey, we can do that for our business. That impacts the bottom line because we're spending less on recruiters, and we're spending less on random pay raises or whatever it is. But I think this space, the HR, edutech, fintech, healthcare tech, this is going to be one of the biggest growing spaces, I think, over the next 10 years because the implications, the positive outcomes, if you do it in the right way, I think could be massive. Zach Frank: Absolutely. There was a company here in town. I didn't work with them directly, but I was advising the person who was. And they're just looking at turnover in an hourly employee population that they had and figure out, all right, if we could get it down just 10%, we'd be saving $5 million a year straight, not even like implied savings, direct savings of $5 million a year because we wouldn't be seeing this turnover. And we're able to then track that down to, all right, well, if a person has two unexcused or unplanned absences within a certain period of time, they're at 80% risk of leaving if they're in one of our hourly positions, higher liability of identification, big impact and the ability to do that. And they didn’t even have to turn that into an operationally deployed model.  They were just able to change the policy and say, okay, here's a better employee assistance program. Here's a more generously engaged time-off policy around family emergencies and the use of the employee assistance program. Great, I think it ended up being $7 million a year or something like that, that they overshot the target and ended up with a 13% reduction in turnover, which is just phenomenal. Dayle Hall: And that's crazy. And I think that comes back to the point you made earlier, which was you don't have to implement an AI-driven system in this function to have a meaningful outcome. It can be a more simple, data-driven model using that. And I don't know anyone that doesn't want to save $7 million from their business. Zach Frank: Oh my gosh, yeah to what you're saying. You don't have to have this operationally deployed AI model that goes in and selects these people and individually identifies them. You can run this analysis every six months and say, all right, do we still have the right value set in terms of employee assistance numbers, retention bonuses, things like that? Great, okay. So you guys in six months. Dayle Hall: Yeah. This is really interesting. I want to close it out with we just talked about some of the opportunities and the benefits you can get. So imagine there's an HR leader out there or the teams and they're exploring whether they want to look at certain types of automations, full-on AI, if necessary. And you've given them some very good advice, things to be cautious about, how to think about it. But if you sat down with that team now and said, look, this is where I would start. These are the things you have to really assess for your own company to be successful. What are those things? What are those couple of pieces of advice you would give that team to make sure that they can set off on the right foot. Zach Frank: I say this for automation, I say this for people analytics. It should be highly tailored to your specific business. You're

always going to see a lot of the same categories, but in terms of specific metrics or things like that, it needs to be responsive to where your business is. It needs to be a part of the conversation. That said, the broad guidance that I'd love to give is this is possible for you. This is not something that's really far off. You don't have to hear AI and think about something that's becoming sentient or doing all these crazy things. There is some process that has been frustrating you for weeks.  And there is probably some solution that is available to you, either in terms of a simplified policy or process or some automation that could be brought in more easily and less expensively than you think, that could help you sleep better at night. These solutions are there. You understand your business well. You understand what the friction and pain points are. There are really great vendors. They are really great experts out there who even just a couple of quick conversations can help you get guidance into, oh, yes, there is a solution out there for this, and it's not something that's only available to the tech giants. It's available to your 500-person company to be able to bring this solution in and start seeing immediate gains.  One that I saw that I absolutely love for us is our applicant tracking system didn't talk directly to our HCMS. So we're just able to plug those things together and say, all right, take the data from over there and put it over there. And that freed up 13 jobs, people that we were able to find, other employees within the company. So again, we're not laying anybody off. We're not putting anybody out for this. But all of a sudden, I had $300,000 that I could then go create my own people analytics team and start working on implementing these things. It was right there. It was available the whole time. It was technology that we had in-house, costs us virtually nothing and just really started this ball rolling in terms of, yes, seeing more and more gains getting further down the road, seeing more and more time freed up, more people who are able to do other things and just compounding gains the whole way. Dayle Hall: I love that piece of advice. I love the fact it doesn't really matter what size of company and you don't have to be a 10,000-person organization. You can take advantage of these things today. Last question, what are you most excited about within the potential of AI within the HR people analytics space? It doesn't have to be 12 months. It could be 5 years, 10 years. What's the most exciting thing you're looking forward to seeing? Zach Frank: Really from an operational perspective, I'd say the advances that are coming in terms of identifying and addressing retention. That's one of the biggest strategic opportunities that exist in terms of HR and serving the business better. The problems from a data standpoint, from just being able to run a Cox regression and see when people are likely to leave and why are not that hard. So I think there are going to be vendors who are going to be stepping into that space and offering really good turnkey solutions. I'm very, very excited about that.  As I look even further down the road and I think about things like reskilling, which I know is a really big conversation that happens around the technology and transformation space, but I have yet to see anybody offer a really, really good solution there. So being able to look down that road and say, all right, going back to that skills conversation, what skills are going to be the most important for us in five years? Who is closest to being able to do that internally? And then how do we most easily convert them over to being able to grab that skill? So that we get better longevity, we get better institutional knowledge and they get a better working position and career.  Everybody wins in this scenario. But I think being able to draw out that trail, we're at the very, very early stages of each of those pieces, much less the whole thing being able to work in some comprehensive manner. But I'm very, very excited about things like that. Dayle Hall: Yeah, two very key areas within HR-type functions, reskilling and retention. Zach, it’s been a great conversation. I feel sorry for our Aramark colleagues who have to lose your expertise. But congrats to the Freeman Company for getting you on their books. Thank you very much for your time today. Zach Frank: Thank you. I enjoyed it quite a bit. Dayle Hall: Great. Thank you, everyone, for listening to this podcast. We'll see you on the next episode of Automating the Enterprise. I'm Dayle Hall. We'll see you on the next one.