Full Transcript for People Analytics and AI with Zach Frank

then, how do we use artificial intelligence tools to help better that 59 00:04:50.000 --> 00:04:56.110 experience that they have and in some cases also improve that decisionmaking, because are 60 00:04:56.230 --> 00:04:59.069 definitely tools that can do that as well. But that's that's really what the 61 00:04:59.149 --> 00:05:01.709 core of the work is about. And then obviously there's all the ones and 62 00:05:01.790 --> 00:05:06.220 Zeros or nuts and bolts of it, which is these meetings and conversations in 63 00:05:06.259 --> 00:05:10.540 the coordination and defining and kind of all those things. But as at its 64 00:05:10.620 --> 00:05:16.740 core, it's really about improving and decision improving decisionmaking and experience around a company's 65 00:05:16.779 --> 00:05:24.250 workforce. Is there a common question that you ask or get asked? What 66 00:05:24.689 --> 00:05:30.329 what do you want to change? You know, that's the there's no right 67 00:05:30.610 --> 00:05:35.839 set of metrics for people analytics. There are some common core areas, right. 68 00:05:36.040 --> 00:05:43.800 So every organization is interested in cruting and talent acquisition. Every organization is 69 00:05:43.879 --> 00:05:48.959 interested in retention and then every organization is to some extent interested in a subset 70 00:05:49.029 --> 00:05:53.750 of retention, which is performance management. Just kind of knowing how we're doing 71 00:05:53.949 --> 00:05:58.589 right. Everybody's concerned about kind of those big three categories. But then you 72 00:05:58.670 --> 00:06:00.310 know, for what exactly do you look at? You know what metrics to 73 00:06:00.470 --> 00:06:03.740 use. How do you look at them? It's really just that that that 74 00:06:03.899 --> 00:06:08.860 most important question of what is it that you want to do, because they're 75 00:06:09.019 --> 00:06:15.779 there is no exactly right metric. There definitely some wrong ones, but the 76 00:06:15.019 --> 00:06:18.329 the metric should really be an answer to a conversation of all right, what 77 00:06:18.490 --> 00:06:23.689 does the company want to achieve? What does it need to change with its 78 00:06:23.769 --> 00:06:29.209 workforce, with its people? Got It? So in the workspace, what 79 00:06:29.410 --> 00:06:35.319 is the importance of correctly sizing teams and is there a good size or the 80 00:06:35.759 --> 00:06:43.759 most efficient size for an any efficient no, you know, I do. 81 00:06:44.160 --> 00:06:46.430 For me, efficiency is really about the processes, like you can be inefficient 82 00:06:46.589 --> 00:06:50.389 with as you be more inefficient with one person than you can with twelve, 83 00:06:50.990 --> 00:06:56.230 and obviously vice versa as well. But I think that question of size, 84 00:06:56.589 --> 00:07:00.629 you know, I think it is. It's important but not necessarily critical, 85 00:07:00.699 --> 00:07:03.819 and some of it goes into that. For me

is I think people generally 86 00:07:03.899 --> 00:07:10.620 approach this question for one of two reasons. One, your team is overwhelmed 87 00:07:10.620 --> 00:07:13.300 and you're wondering if it's just you and like, how big should we be? 88 00:07:13.459 --> 00:07:15.529 I'm freaking out, I've got all this work to do. And then 89 00:07:15.569 --> 00:07:20.850 the second scenario is you're designing or redesigning a team. You're considering that structure 90 00:07:20.889 --> 00:07:24.529 the future of work and you want to know. Okay, am I I 91 00:07:24.649 --> 00:07:27.449 have an opportunity to set this I'm up. Am I doing it right? 92 00:07:28.410 --> 00:07:30.600 So that for that first group, who you're overwhelmed and just wondering if you're 93 00:07:30.600 --> 00:07:34.399 crazy, probably not. I mean, I guess it's possible that you're just 94 00:07:34.480 --> 00:07:38.519 doing a really bad job at delivering products and kind of getting things out, 95 00:07:38.839 --> 00:07:43.360 but probably not. It probably should be bigger and I think while you lobby 96 00:07:43.439 --> 00:07:46.470 for those resources, just try to get better about setting boundaries and being transparent 97 00:07:46.509 --> 00:07:50.189 about the work process so that you're not covering up the warts of what it 98 00:07:50.269 --> 00:07:55.509 takes to do the work. For this second part, the only thing that 99 00:07:55.589 --> 00:07:59.139 I've seen that approaches a short answer is I've seen it a couple places. 100 00:07:59.180 --> 00:08:03.459 People say three to seven percent. I would say that, if you're going 101 00:08:03.500 --> 00:08:07.300 to go for just a quick grab answer, five to eight percent is probably 102 00:08:07.660 --> 00:08:11.459 more accurate. But that's a big gap and I think there are a lot 103 00:08:11.500 --> 00:08:16.009 of dependencies. I mean really the it really depends, but I have I 104 00:08:16.370 --> 00:08:20.490 I've been fortunate enough to feel like I've discovered a set of questions that you 105 00:08:20.610 --> 00:08:24.089 can walk through to help figure out, all right, am I on the 106 00:08:24.129 --> 00:08:26.920 right track? Am I getting close to it? Won't tell you exactly how 107 00:08:26.920 --> 00:08:31.120 many people, but it at least gets you thinking about the right things to 108 00:08:31.199 --> 00:08:37.840 see if you've diagnosed that for yourself correctly. So when we talk about the 109 00:08:37.000 --> 00:08:41.110 end user, how can you be proactive and and value to them, and 110 00:08:41.389 --> 00:08:48.950 what are some systematic approaches that you can take? Yeah, the so kind 111 00:08:48.990 --> 00:08:52.669 of similar to before. It depends very much on your situation. I think 112 00:08:52.750 --> 00:08:58.779 the the biggest thing about going proactive is deciding that you're going to do it 113 00:09:00.100 --> 00:09:03.659 right. You know, that's just making that switch of Oh, we're not 114 00:09:03.820 --> 00:09:07.779 just going to fill the requests that come to us and and you know, 115 00:09:07.820 --> 00:09:11.049 we're not just going to be responsive to that. We are going to look 116 00:09:11.090 --> 00:09:16.490 for opportunities and take charge of the story of how we tell data stories in 117 00:09:16.570 --> 00:09:20.289 the organization. So for us, a few things that have been helpful is 118 00:09:20.289 --> 00:09:24.759 one is give more than what is asked. So if somebody asks for a 119 00:09:24.879 --> 00:09:30.600 basic data dump or even just a set of numbers for I mean we'll get 120 00:09:30.639 --> 00:09:35.360 this one a lot with a request for proposal or if responses, and they'll 121 00:09:35.399 --> 00:09:39.110 say something like all right, what was your internal versus external hiring rate? 122 00:09:39.309 --> 00:09:41.230 Right, and you could just give those numbers over and say, Oh, 123 00:09:41.389 --> 00:09:43.830 here they are. So what we try to do is, okay, if 124 00:09:43.870 --> 00:09:48.590 you're going to give something like that over, give them something else and in 125 00:09:48.710 --> 00:09:50.990 a for instance like that, it could be, oh well, here's what 126 00:09:50.110 --> 00:09:54.220 it was for last year, but that trend has increased over the last three 127 00:09:54.259 --> 00:09:56.259 years. So we want to give some sort of addition, a whole piece 128 00:09:56.259 --> 00:10:01.460 of information. That would answer an unasked follow up question. I think another 129 00:10:01.500 --> 00:10:03.940 thing you can do is talk directly to your audiences. You make sure that 130 00:10:05.019 --> 00:10:11.169 you're interfacing with the people who are actually using your insights so that you're not 131 00:10:11.330 --> 00:10:16.809 kind of interfacing with another layer that's missing, that translation. And then, 132 00:10:16.850 --> 00:10:20.289 lastly, you know, don't be afraid to push back on ownership. Part 133 00:10:20.330 --> 00:10:26.240 of being proactive is making sure you stake out your territory really well. One 134 00:10:26.240 --> 00:10:30.200 of the things that we run into in the HR space is a you can 135 00:10:30.200 --> 00:10:35.590 run into a lack of clarity about what is communication and what is reporting or 136 00:10:35.629 --> 00:10:39.149 an insight, and so that's an area where we have to be really careful 137 00:10:39.149 --> 00:10:43.990 about saying, hey, we can provide the data, we can help you 138 00:10:43.110 --> 00:10:46.590 make sense of the data, we can talk to about trends, we can 139 00:10:46.669 --> 00:10:48.750 try to provide you know, broader market context and a lot of those things, 140 00:10:50.139 --> 00:10:52.899 but if somebody's asking about a situation or a specific account, there's got 141 00:10:52.940 --> 00:10:58.779 to be a conversation that takes place about okay, our insights feed into your 142 00:10:58.899 --> 00:11:05.330

communication to your stakeholders about what's going on. Okay, you mentioned HR. 143 00:11:05.210 --> 00:11:11.970 What is HR analytics and how is it different from other analytics? What's your 144 00:11:11.049 --> 00:11:16.769 take? Yeah, in some ways it's not. You know, at it's 145 00:11:16.809 --> 00:11:22.440 basics, people in Lys HR analytics is about using data to help inform people 146 00:11:22.519 --> 00:11:26.600 in policy decisions. Right, obviously, specifically people related policy decisions. Again 147 00:11:26.639 --> 00:11:31.600 with the the tongue twisting. So you know, in terms of what's different, 148 00:11:33.750 --> 00:11:39.149 everything is protected information, like when you're working with people analytics that you 149 00:11:39.350 --> 00:11:43.309 there's no piece of information that you're touching that's not hat, that doesn't have 150 00:11:43.470 --> 00:11:46.580 some kind of protected data that goes along with it. So you've always got 151 00:11:46.659 --> 00:11:48.700 to have that track kind of playing in your mind. You know, you 152 00:11:48.860 --> 00:11:54.580 never dealing with an obscured or, you know, d identified set or something 153 00:11:54.659 --> 00:12:01.019 like that. I think, you know, in some ways it's easier than 154 00:12:01.179 --> 00:12:05.210 other types of data that I've worked in. Your demand complexity is typically a 155 00:12:05.250 --> 00:12:09.610 little it's not necessarily lower, it has a gentler curve. So a lot 156 00:12:09.610 --> 00:12:11.769 of your people organizations are going to be starting at a lower point on an 157 00:12:11.769 --> 00:12:18.600 analytics maturity scale and they're going to more slowly ramp up into more complex demand. 158 00:12:20.759 --> 00:12:22.440 I think another thing is, you know, there's kind of a natural 159 00:12:22.559 --> 00:12:26.559 meaning to people analytics work. You don't have to ask about the soul of 160 00:12:26.679 --> 00:12:31.470 what you're doing. You're trying to improve people's working conditions and people spend a 161 00:12:31.509 --> 00:12:35.870 lot of time at work, so there's a lot of good inherent value. 162 00:12:35.909 --> 00:12:39.149 It's also really easily transferable. Yes, there can be differences, but people 163 00:12:39.190 --> 00:12:43.620 analytics at one organization is going to be more similar to people analytics at another 164 00:12:43.659 --> 00:12:50.179 organization versus, you know, industrial control system analytics versus call center analytics or 165 00:12:50.299 --> 00:12:54.820 something like that. So I think those are some things. I think some 166 00:12:54.980 --> 00:13:00.850 things that make it harder are the tech is always hard. I should say 167 00:13:00.850 --> 00:13:05.009 almost always hard. There are a couple of organizations that have really good integrated 168 00:13:05.049 --> 00:13:09.529 hr text x, but then you also walk into places like a previous employer 169 00:13:09.610 --> 00:13:15.000 of mine. When I started we had ninety different instances of HR software in 170 00:13:15.039 --> 00:13:18.720 the United States alone. And when I say H I should say HR systems, 171 00:13:18.759 --> 00:13:22.600 not software, because some of these were post it notes getting typed into 172 00:13:22.679 --> 00:13:26.509 excel right like. It was a very generous definition for a system. So 173 00:13:26.750 --> 00:13:31.429 the tech is always hard. Your literacy tends to be pretty low in terms 174 00:13:31.429 --> 00:13:35.309 of your audiences. And then just back to that lack of clarity and kind 175 00:13:35.309 --> 00:13:39.230 of the intermixing of what's a communication and what's an insight, because this is 176 00:13:39.350 --> 00:13:46.419 such every case matters, right. So it's not just the statistics of how 177 00:13:46.419 --> 00:13:50.299 many investigations you're doing, but also what's going on with this one particular important 178 00:13:50.340 --> 00:13:54.259 investigation. So there's always kind of this two stage hand off of internal stakeholders 179 00:13:54.379 --> 00:14:01.289 to add context to external stakeholders. Do you have an example of a data 180 00:14:01.330 --> 00:14:05.570 set that you work with? Yeah, so, I mean one that we 181 00:14:05.690 --> 00:14:11.039 do pretty routinely is candidate information, and even that you can that's kind of 182 00:14:11.080 --> 00:14:16.039 going back to the tech question. Almost every element of HR data sets can 183 00:14:16.080 --> 00:14:18.720 be broken down into smaller and smaller sets. So I can even talk about 184 00:14:18.720 --> 00:14:24.629 candidates right. Well, there is the candidate, the records of the candidates 185 00:14:24.669 --> 00:14:28.269 as a, you know, user in our systems. There's their specific applications 186 00:14:28.470 --> 00:14:33.950 and how those applications are applied to jobs. There's their background check information, 187 00:14:33.110 --> 00:14:37.429 their offer information. So it's even your most, you know, relatively simple 188 00:14:37.940 --> 00:14:43.019 data set has all these sub SCHEMAS and data sets that go along with it. 189 00:14:43.179 --> 00:14:48.139 That it's just it's a never ending onion of fun. Is the never 190 00:14:48.299 --> 00:14:52.929 ending onion of fun? Yeah, that's not what I'm going to coin. 191 00:14:54.009 --> 00:15:00.649 I don't think so. Do you use AI in your work? Like, 192 00:15:00.769 --> 00:15:05.289 have you incorporated ai or machine learning into what you're doing? Yeah, aim 193 00:15:05.370 --> 00:15:09.320 more so than machine learning, and I think you know probably one of the 194 00:15:09.360 --> 00:15:15.039 things that I'd say as a big point of understanding for how we use AI 195 00:15:15.480 --> 00:15:20.600 is that we are very much a vendor led ai organization. So our focus 196 00:15:20.639 --> 00:15:26.470 is much more on integration and oversight for

AI and and I think you know 197 00:15:26.549 --> 00:15:31.070 some of the reasons for that are that AI and HR is fraught with difficulty, 198 00:15:31.549 --> 00:15:35.539 and so I would say that in less ai is something that your organization 199 00:15:35.820 --> 00:15:39.379 does well in another part as part of it's as part of its kind of 200 00:15:39.419 --> 00:15:46.340 natural operating of business. HR is a very difficult place to launch into it 201 00:15:46.580 --> 00:15:48.659 right. You know, even if you're doing it in other parts of centralized 202 00:15:48.700 --> 00:15:54.809 functions. that a couple of reasons. Typically you don't have a lot of 203 00:15:54.889 --> 00:15:58.490 the talent to do it. It's not something that naturally develops in the HR 204 00:15:58.570 --> 00:16:03.529 space. I mean even people analytics is a relatively new field. So you 205 00:16:03.610 --> 00:16:06.159 know, even if you have the talent and house now, recruiting for it 206 00:16:06.240 --> 00:16:08.120 in the future is going to be difficult and you could create a system that's 207 00:16:08.159 --> 00:16:11.840 not sustainable or maintainable. You know, one of the biggest things that we 208 00:16:11.879 --> 00:16:18.629 run into is we don't have pipelines to deliver the insights. So we've created. 209 00:16:18.950 --> 00:16:25.070 We've got a couple predictive models that we've created that have degrees of accuracy 210 00:16:25.190 --> 00:16:26.950 that are good enough to be put out in there. It put out into 211 00:16:26.990 --> 00:16:30.990 the world right and to be operating in our systems, but there's no way 212 00:16:32.190 --> 00:16:37.620 to insert those outcomes into the natural workflows where end users can look at them. 213 00:16:37.700 --> 00:16:41.820 Kind of going back to the the technical problems before your your HR systems 214 00:16:41.860 --> 00:16:48.090 aren't generally good receivers of this and with HR being such a for most of 215 00:16:48.169 --> 00:16:53.529 your users, hr is going to be a subset of a subset of a 216 00:16:53.649 --> 00:16:57.370 thing that they deal with right, because your main audience is hiring managers and 217 00:16:57.490 --> 00:17:04.079 employees, and so trying to deliver insights to them in a third or sixth 218 00:17:04.359 --> 00:17:07.400 system that they have to go to just isn't something that's going to work. 219 00:17:08.079 --> 00:17:12.079 I mean, the other thing that can happen is then for that insight to 220 00:17:12.160 --> 00:17:17.910 be meaningful. So one of the one of the predictive models that we built 221 00:17:17.950 --> 00:17:21.349 that has four jobs that we've seen before. It had. I want to 222 00:17:21.349 --> 00:17:26.150 say, at ninety six percent accuracy was estimating how long it would take us 223 00:17:26.190 --> 00:17:30.180 to fill the job. Right. That's great high degree of accuracy. What 224 00:17:30.259 --> 00:17:33.500 it felt good to playing that, but didn't have a way to get that 225 00:17:33.539 --> 00:17:40.180 information out to recruiters, hiring managers, HR business partners. And if we 226 00:17:40.339 --> 00:17:44.930 did, we had to think about, okay, what value is that actually 227 00:17:45.490 --> 00:17:51.730 creating? It's in some cases that's actually creating more of a dangerous expectation that 228 00:17:51.849 --> 00:17:56.650 doesn't conform to the reality of what's going on on the ground for that user. 229 00:17:57.089 --> 00:18:02.480 So I you know some of those very practical difficulties and then I think 230 00:18:02.720 --> 00:18:06.119 you know not having balanced data. So another reason that vendors are so much 231 00:18:06.119 --> 00:18:11.400 better at Hra I is look at trying to evaluate candidates. Well, you 232 00:18:11.680 --> 00:18:15.950 build an internal AI to evaluate candidates. The most that you can ever do 233 00:18:15.349 --> 00:18:19.670 is look at who has already been successful with your organization, and so what 234 00:18:19.829 --> 00:18:25.869 you can do is create a very biased and and, you know, unfortunate 235 00:18:26.470 --> 00:18:29.299 model that's kind of making all these decisions that you don't want it to be 236 00:18:29.420 --> 00:18:33.420 making. And so vendors have more ability to pull from different employers and and 237 00:18:34.019 --> 00:18:37.500 create a more holistic and a more fair picture and so those are just a 238 00:18:37.579 --> 00:18:44.569 couple of the reasons that we have gone with a vendor let approach in terms 239 00:18:44.569 --> 00:18:47.890 of our HR AI. Now that being said, there are a number of 240 00:18:47.930 --> 00:18:51.809 advantages if you can find good partners. There a number of advantages that can 241 00:18:51.890 --> 00:18:56.920 come from using that kind of AI, like faster hiring of better candidates, 242 00:18:56.839 --> 00:19:03.279 you know, lowering your turnover and, inversely, increasing retention and engagement and 243 00:19:03.359 --> 00:19:06.960 things like that. So there definitely are benefits, but it's very much a 244 00:19:07.279 --> 00:19:11.390 space where I highly recommend being an integrator and an owner rather than a developer. 245 00:19:14.109 --> 00:19:18.869 What excites you the most about HR data analytics? I can because I 246 00:19:18.910 --> 00:19:23.150 can feel the energy from your last as yeah, I do. I love 247 00:19:23.190 --> 00:19:27.740 it. It is the most fun data type that I've worked with so far, 248 00:19:27.779 --> 00:19:33.619 aside from geographic data, because I that's just a special nerdy kind of 249 00:19:33.660 --> 00:19:37.660 fun for me that I don't understand. You know, I think, going 250 00:19:37.700 --> 00:19:40.329 back to what I was saying before about what's different about I think there is 251 00:19:40.569 --> 00:19:47.650 inherent meaning in working with HR data. Right you are affecting people's lives, 252 00:19:48.410 --> 00:19:51.529 their livelihood in terms of how they get their money, their experience of that, 253 00:19:52.089 --> 00:19:55.880 because people spend a signific pig amount of time at work, and so 254 00:19:56.079 --> 00:20:02.720 everything that we do is kind of directtly contributing to that picture of how do 255 00:20:02.839 --> 00:20:07.470 we treat our people well, how to crewy craft policies that, you know, 256 00:20:07.670 --> 00:20:12.109 better enable and empower and take care of our people that make up our 257 00:20:12.190 --> 00:20:15.990 workforce? I think that's one of the reasons, probably the main reason, 258 00:20:17.069 --> 00:20:18.789 that I love it. Also, it's, you know, kind of going 259 00:20:18.869 --> 00:20:26.660 back to what I said before about the the relatively low slope in terms of 260 00:20:26.740 --> 00:20:33.299 demand or the easy slope in terms of demand. It it can be easy 261 00:20:33.539 --> 00:20:37.890 to show up and provide value. I remember and organization that I joined. 262 00:20:38.769 --> 00:20:42.250 We were looking at a number, NHR number, at one point and I'd 263 00:20:42.369 --> 00:20:45.970 ask a question that I thought was going to be relatively easy to answer, 264 00:20:45.970 --> 00:20:48.009 which is, okay, we this number is here. Is this number good 265 00:20:48.170 --> 00:20:52.880 or bad? And in the room of ten other people, nobody knew the 266 00:20:53.039 --> 00:20:57.079 answer to that question. So like all right, I can deliver value to 267 00:20:57.160 --> 00:21:03.079 this organization pretty quickly. Yeah, yeah, well, just a question. 268 00:21:03.309 --> 00:21:07.869 That's right. Yeah, Yep. Well, so, for those that are 269 00:21:07.950 --> 00:21:12.230 listening, how can they reach out to you? Yeah, I am on 270 00:21:12.549 --> 00:21:21.019 Linkedin. I believe my slash is Zach l Frank I'M gonna say. And 271 00:21:21.220 --> 00:21:26.220 also I'm on twitter, although I don't twitter a lot. I'm more of 272 00:21:26.259 --> 00:21:30.140 a lurker on twitter. Zack the frank at Zach the Frank. Those are 273 00:21:30.299 --> 00:21:34.130 two pretty good ways to get a hold of me. Sounds Great. Well, 274 00:21:34.210 --> 00:21:38.450 for more information on the data standard, you can find us at www 275 00:21:38.849 --> 00:21:45.890 dot data standard dio. This episode was also sponsored by Pandio and they're innovating 276 00:21:45.930 --> 00:21:48.720 the text base. You can learn more about their work at PANDOCOM. But 277 00:21:48.799 --> 00:21:53.240 Zadah, you again for being here and it was a pleasure speaking with you. 278 00:21:53.880 --> 00:22:42.609 Thanks, FANNA WO