Helena Almeida - AI Governance as a differentiator for HR and your Organization

[00:00:03] Welcome to the HR Data Labs Podcast, now part of the WorkDefined Podcast Network. Join us as we explore the vital role of compensation, strategy, data, and people analytics in navigating today's complex business world. With the resources of WorkDefined, we're now bringing you deeper insights and actionable ideas from top experts. Now, here is your host, David Turetsky.

[00:00:25] Hey everyone, welcome to the HR Data Labs Podcast. I am your host, David Turetsky. And we always try and find the best and the brightest inside and outside the world of human resources to talk to us about what's going on today. Well, today we have a very big treat for you from the mothership ADP, Elena Almeida.

[00:00:49] She's the VP of ethics basically, or managing counsel and legal officer at ADP. I should actually say that instead. Elena, how are you? I'm doing really well. Thank you. It's really great to be here. It's good to have you because one of the fun things we get to do, especially when we talk to our friends from ADP, we learn a lot, especially because of all the different business problems that clients of ADP are facing.

[00:01:18] But we're going to talk about some really fun stuff today that are kind of next level things, especially from a legal perspective. So I can't wait to hear about them. But why don't you give us a little bit about your background and what you do at ADP? Sure. So I am a recovering litigator. So I came to ADP about eight years ago with 20 plus years of doing commercial litigation.

[00:01:40] And then about five years ago, I moved over from litigation to product. The idea of how do we avoid having problems that bring us or our clients into litigation? So I work closely with the product teams to kind of at that intersection of technology, product ethics and laws to really figure out and help them develop products that are compliant, help our clients comply and really taking advantage of the latest technology. That's excellent.

[00:02:10] So I think nothing's changing at all in my job. Oh, no, no. It's so boring, I imagine. I mean, things don't ever happen. Yeah, it just feels like we're watching paint dry or watching a golf tournament. Yeah. So before we get into our topic, which will be really fun, what's one fun thing that no one knows about you?

[00:02:35] So usually when someone asks me for a fun thing about me, I would say that my husband's name is Michael Jordan, which has been a whole lot of fun and challenges over the years. Like getting restaurant reservations is a challenge. But you said that no one knows about me. So I'll tell you that I used to make up fun facts. So when I was my daughter was little years ago and my hobbies involved, you know, trying to figure out when I could sleep or do laundry and things that I didn't have hobbies. Right. You know, I was like, I'm not going to tell people that I like to do crossword puzzles.

[00:03:05] So I would make up hobbies that I had. So I used to tell people that I was an archer that I would get, you know, so that I would. That's what I did on the weekends, you know, brought out my bow and arrow and did archery. And it was, you know, always fun to see who believed me or who didn't. Did you really? No. No. Wait a minute. I was the mother of a little girl. I was just trying to get to PTA meetings and work. But I didn't want to be the person who had like crossword puzzles and, you know, reading as they're happy. So I was just making things. But archery is so much fun.

[00:03:35] It is. It is. And actually my town has an archery hitch. Yeah. I don't know what they call it. See, I didn't. That was the problem. It was never really that believable because I didn't know any of the words associated with being. But you should try. Because it was actually a fun sport. I think it would be great. I think it'd be really fun. I'm going to. And now that she's older, maybe now that this year, this year. How old is she? She's 16 now. Yeah. Well, definitely 16. So I can put archery on my roadmap now.

[00:04:05] Exactly. For this year. Yeah. Exactly. Yeah. I mean, you know, I have a lot of bucket list items. Archery is not one of them. I used to do it back at Penn State. So I know my limitations there. But, you know, I don't know whether it's called a pitch or a field or whatever either. So I couldn't tell you. But I have a feeling you have some great hobbies. Well, you know, other than building businesses and kids and chickens now, you know. Yeah. Well, I don't have the chickens yet. I have to get approval for them.

[00:04:34] I got to hear more about the chickens. I got to hear more about the chickens later. But. Okay. Yeah. We'll cover that afterwards. But today, we're going to be talking about something really cool, which is AI governance on a global scale. And this will be a lot of fun for a lot of us because while a lot of us are kind of dipping our toes or diving in with both feet, it would be really cool to hear a legal perspective. But we'll do that right after this.

[00:05:13] Hey, everybody. I'm Lori Rudiman. What are you doing? Working? About work. And that barely counts. So while you're at it, check out my show, Punk Rock HR, now on the Work Defined Network. We chat with smart people about work, power, politics, and money. Are we succeeding? Are we fixing work? Probably not. Work still sucks. But tune in for some fun, a little nonsense, and a fresh take on how to fix work once and for all.

[00:05:49] So, Elena, let me ask you the first question, which is a really cool one. When we first start talking about AI, typically we start talking about the technology. But when we're thinking about the workplace settings, what really are the things that we should be focusing on first? And not just the technology, I imagine. Yeah. I mean, I think, don't get me wrong, right? Organizations definitely have to care about those things like data quality, you know,

[00:06:18] is the AI being tested, privacy, security, monitoring. All of those things are extremely important. And that's something that, you know, vendors and developers are spending a lot of time thinking about. But what I find really interesting is in workplace settings, when we're trying to introduce AI into what, you know, what we're doing all the time in HR, you really have to be thinking

[00:06:44] about, like, what happens when the human interacts with the AI inside, like, an actual workflow, right? Not that the data quality and the testing and all that doesn't matter, but I think the more interesting question is really what happens when humans and AI meet on the day-to-day, on the average day. I don't remember what's that saying about, you know, no plan, no military plan survives first contact, right? It's sort of like that.

[00:07:10] You know, obviously, practitioners aren't thinking about model documentation or the technical controls when they're sitting and trying to get payroll out or when they're sitting and going through a recruiting process. They experience it through interaction. And that, that's a challenge, as you know, because, you know, AI could sometimes bring out the, some of the worst of humans, the best of humans. And really, it's how do you govern that?

[00:07:35] Well, it can also hallucinate too, which is one of the really wild things about, about AI is that, you know, trust is important. And we live in a world now where there exist things called alternative facts and now there are HR, sorry, AI hallucinations. So I guess the question I wanted to ask around this is, it's not even just that the technology

[00:07:59] is floating new things in front of practitioners as they're involved in workflows like you were mentioning, but do they need to be skeptical of not just recommendations, but also the data they're seeing, ensuring that they're on target for completing that process? Or should they, should they really be on their toes as they're doing this?

[00:08:25] I think it's, it's really interesting because when I first started working a lot with AI years ago, I think what we were worried about is, can we build AI that people will trust enough to use? I still think that's a really important question. You know, we worried about adoption. Are people going to be comfortable bringing AI into their workflows? And I think more and more now I'm thinking about, are we going to trust it too much, right?

[00:08:52] Because like you said, hallucinations, there's, there's errors, there's mistakes. I don't think we haven't gotten to the point where humans aren't necessary in, in, in these processes, especially in what you and I are talking about, because you and I are talking about AI in situations where it impacts who gets hired, who gets fired, pay decisions, these really crucial decisions to people. So sort of right is, is not, is not good enough, right? We really need accuracy and we need people to be, to be paying attention.

[00:09:21] So I have definitely shifted my, my thinking to, are we going to be able to create something that is, is, is good enough for people to trust it? To how do we calibrate the trust? That's the goal really, right? It's not, you know, blanket distrust of AI and it's not blind trust of anything AI tells you. It's how do we design a system where we're encouraging people to really engage and think about it and not, not be on one of those extremes.

[00:09:46] Remember back to the days when IVR wasn't necessarily new, but it was the hottest thing in the HR world. We were saying, oh, you know, you can call into this IVR and get your number of vacation days or you can get answers to your questions. Well, now we have the, I guess you could say the feedback, immediate feedback loop of having that chat bot where you say, hey, how many vacation days do I have left?

[00:10:15] And it has to do that math and it has to know exactly what to do. And it also needs some of the rules that are specific to you, to your municipality, to your, to your state. Are you on FMLA? What are the issues that are going on with you and not hallucinate? Because the answer that you get back, couldn't that lead to litigious situations? Because if it says you have 30 days left, but you've exhausted it all, isn't that the problem as well?

[00:10:41] Is that we are darn sure that those direct reach outs are, they're pristine. Yeah, there's a lot of thought that goes into and a lot of testing that goes into how can we make those, how can we make sure that the answers that we're providing on those key questions are really accurate? It's, it, there's a tension between, you know, a lot of clients want to hear, just kind of make it as simple as possible, automate.

[00:11:08] And we always talk about it too, reduce friction by having a service that will, that will answer employee questions like that and maybe take some of that work off of, off of HR's plate. Getting it right is, is extremely important, takes a lot of testing. And that's why we really, you know, do emphasize making sure that we're able to do that, but in the context where there's still meaningful human engagement around it.

[00:11:38] Because you're right, you know, we don't, you know, we, we do what we can to make sure that our clients are complying with their, with their obligations. And, you know, the clients trust us though, to, to helping them with tools that are, you know, are they, are going to help them and not hurt them in that context. So there's definitely a balance and attention between being able to be as out there as we can and providing answers and providing advice and, and, and calculating it, but doing in

[00:12:02] a way that's still has that meaningful human connection because, you know, our, our, our different clients have different rules and different complexities. And we want to make sure that we're providing something that, that, that deals with them. It's, it's, I mean, you remember from, you know, when the pay transparency law started coming out, right. And you had a law in this jurisdiction or that jurisdiction or that state and trying to manage across all of the different rules that, that people could be interacting with is a challenge.

[00:12:32] And we're, we're doing that here with AI and trying to make sure that we're striking the right balance and, and dealing with all the clients' different asks and needs. Yeah. The problem though, with pay transparency is there's a spirit of the law and then there's the letter of the law, right? I mean, I'm telling a lawyer this, but, but, but, but in terms of the spirit, they want, and, and, and the transparency laws, even the one in Maine and New Jersey and the others

[00:12:59] that have been enacted not too long ago, they're going a little bit next step. Whereas the original ones were all about just, you know, don't ask about the current pay range or don't ask about the current pay. Don't ask what they want and tell them what the range is on the advertisement. Well, now it's going beyond that to asking what's the requisite pay for the experience,

[00:13:27] as well as what are the benefits offered within the first year? So the next step on pay transparency, that double click into it is even more complicated because now when you're posting a job, you have to be at more accurate. It's not just give the whole pay range. It's give the requisite range for the level of experience that we're trying to hire for.

[00:13:51] And I think that's missing from a lot of the conversations around transparency. People are like, ah, I'll just give them the range. You can't do that realistically because you're not going to pay somebody at the max. No, it's been, I mean, and we've seen, you know, we've seen different, you know, since Colorado passed their first law, we've seen kind of successes and, and kind of approaches to, to work around this. And, you know, when you think about it from, from ADP's perspective, when we have, you know,

[00:14:20] 75 plus years of, of, of helping our clients in these situations, all our clients have different needs, like, which is why we're building, we're focusing really on building with flexibility, right? Making sure that people have the data that they need, right? And, you know, you know this as well as I do in terms of kind of robust data that allows for benchmarking that's actually relevant, that's current to help with those kinds of pay transparency decisions.

[00:14:48] But it's really clients, clients are kind of have a lot of different needs and we have to really be flexible to be able to make those. And, and that's what ADP has been doing for a very long time. I mean, we, when I worked at ADP, we always took the approach of trying to be as configurable as possible because of this, the, the richness in the differences within a company, not just between companies. There's always differences between companies, especially in compensation and benefits.

[00:15:14] But the richness even within a company, whether someone's a union employee, whether they're an executive, whether they're on secondment, whether they're an expat, you know, there are so many differences in complex business environments. ADP tries to deal with all of those things at once. So there's no, you know, person left standing over there. We can't deal with them. One of the issues that brings up, especially in the context of artificial intelligence is

[00:15:43] trying to figure out how all of those things figure into not only the, what you say to whom, but also the decisions that you're mentioning before, who gets hired, who gets fired, who gets an increase, who gets, all of those things become that Rubik's cube. Yeah. So the complexity becomes almost unbelievable. I was recently talking to a client who was, you know, um, you know, ADP has, has ADP assist,

[00:16:12] which is clients, client facing AI that, that can answer these questions. And he was saying, well, we have three different unions in six different states. And the, you know, the, the answer depends on this kind of permutation of different things. And can, you know, can AI really answer all of that? First of all, my question was, I don't know, can a human answer that? I mean, the, the, the complexity of that is... The one person probably can. Right.

[00:16:42] One person who can. And you can be in a really long time. And hold on to that person, right? Right, exactly. But the challenge is really, how do you do that with AI across all the difference, as you said, within a company? Well, you know, we're looking at this a lot through personas because, you know, if you look at the sort of, what does a practitioner need? What does an employee need? What does a manager need? What does a recruiter need? Because we found that approach is really helpful to trying to get at the problem we're trying to solve.

[00:17:09] I think, you know, AI is a whole world of what can you do? You can automate, you can summarize, you can recommend, you can do this. For us, it's really more about focusing on, which helps me as an attorney, it's focusing on the problem we're trying to solve and really directing. How can we do that in a way that doesn't kind of cause additional issues, as you mentioned, but really gives our clients at the practitioner, employee, manager level, the information they need when they need it to make the decision?

[00:17:39] I mean, that's what it comes down to, whether you're using technology or whether you're, you know, you're answering something on the phone. You said something fascinating just a little while ago that ties into this, which is getting the right answer. And there may be only one person at that organization who has that information anyways, and who understands what that right answer is across those differences, whether it's the six states or the three unions or whatever. And even that person can be wrong. That person tries not to be.

[00:18:08] Right. But that person can be wrong. We also have data in our HRMSs. We have data in payroll systems. We have data in the benefit system that might not be right. And so everything has to be taken with that grain of salt, Elena. 100%. And by the way, a new regulation comes out tomorrow. Everything we have in there is garbage.

[00:18:30] So, you know, I have been said, so, so one of the things I think is fascinating is that when we started looking at use cases at ADP and we started thinking about sort of what, as you know, so the laws, the laws don't have a standard, right? It has to be accurate at this percentage according to this test. And so a lot of this is best practices and, you know, listening to industry guidance about how to do this. So when we first started looking at it, we were trying to think about, you know, accuracy.

[00:18:59] Is there one number where, you know, this has to give the right answer 99.9% of the time? Very hard to do. But I kept saying, well, how accurate are our people when they're answering the phone and addressing these questions? And we don't have that. We don't have that metric, right? We have a lot of training that goes into helping our service associates, supporting them with, you know, experience and the information that they need in order to do it.

[00:19:23] But I know in my life, I pick up the phone and I call, you know, I call a vendor, you know, service provider for me. And the human might not always be accurate. I understand that. I understand that we also have a different expectation for AI. And that's a little bit of the trick, right? People are extremely willing to believe what AI tells them, that whole automation bias thing. So on the one hand, we're holding AI to a different standard than humans.

[00:19:53] And we should, to be fair, we should. But, you know, but we're also trying to make sure that people aren't, you know, forgetting their common sense at the door when they're dealing with AI and questioning what they should question. So, so, so, but in that realm, and this gets us to the second question, when you're developing things and when you're talking to ADP about developing things and, and, you know, you're thinking about how it fits into everything.

[00:20:20] And there may even be ways you design human in the loop. How do we get to the concept of responsible artificial intelligence? Is there, is there a design parameter like you were talking about before? What's the standard for the percentage of, of correct answers? What is responsible AI in that, in that scenario? You know, it's, it's, okay.

[00:20:47] Tip one, number one, if you're a lawyer and you're, and you're going to talk to, to people about AI is try not to say it depends, but really that's, you know, it depends. You know, I, you know, I, we were talking a little bit before about, I have a daughter who's 16 and she's going through this recruiting, you know, she's, she's an athlete and she's going through this recruiting process. And, you know, we, we, we were talking to, talking to an AI about an AI that focuses on recruiting about different schools she was looking at.

[00:21:14] And it spat out some just wrong answer, right? It was just wrong. I knew it was wrong. You know, my daughter knew it was wrong, but you know, that, that's just sort of, okay. It was annoying, but it wasn't, it wasn't kind of a huge decision. It was right. So that's low stakes, right? High stakes is when you're using AI to do something really important in the workplace. And that's, there are different standards.

[00:21:37] There have to be different standards for, for accuracy, for, you know, for the humans to trust it and for the humans to question it. And I think we do have an obligation to make sure that we're as accurate as possible on these vital questions and that we're, you know, the concept of explainability, which we're seeing in a lot, a lot of laws that are being passed and best practices.

[00:22:01] Explaining to users, you know, this is what this AI is, is, is trained to do, trained to do well. It is still incumbent on you to make sure that you're, you're thinking about this particular show. I don't know if you remember when we first started playing around with ChatTVT and everybody realized it couldn't do math, right? Or couldn't do math very well. And every, I mean, you know, things are getting better and everything, but it was important for us all to know the number that gets spit out. Think about it.

[00:22:31] Does it make sense to you? And that's what we're trying to do is really have a risk tiered approach to, to essentially the requirements that we hold our AI to so that we can continue to innovate quickly, to be able to address the things that our clients want and need, but also make sure that we're, we're holding high standards for those things that really matter.

[00:22:54] You know, when I, when I think about what you just said to me, there's a, there's a kind of going back to the old days of when we started thinking about outsourcing HR and outsourcing pieces of our, our jobs and our administrivia to other people. And the thing that we did was we held them to SLAs for how many times they were right or wrong and, and what was their response times and things like that.

[00:23:21] And I think that we give AI a little bit of a, we give them a lot of breathing room because their response times are really quick, but the answers might be really wrong. And unless you're knowledgeable, like unless you know math, you don't know that their math is wrong. Right. If you don't have the counting when they're next to you, you don't know. And we give them a lot of room because we say, well, wait a minute. They have a lot of information at their disposal to be able to make this decision.

[00:23:52] And I'll give you an instance where this is really bad when you get the wrong answer. Like I was asking it to give me the names of the administrators for certain counties in a state because I was reaching out to them. And I looked at the answers and I knew that they were all wrong. How? Because I'm working with some of those counties. And I looked at the words and I looked at the names they were giving me and I'm like, that's not the person from Hancock County. I know that person.

[00:24:21] That's not that person's name. And so I asked, pull blank, I said, is this accurate information? And they said, no. I guess I lied.

[00:24:34] And so part of our problem is one of the things, if we're looking at responsible AI, we need to make sure that the people who are developing these algorithms, these tools for us, is asking more questions and designing it in a way that it's, I don't want to say foolproof because we're not fools. Unless we actually use it and just buy into it 100%. You know what I mean? There needs to be that bias put into the beginning of doubt so that we know we're not getting crap back.

[00:25:04] I 100% agree with you. I think it's, you need knowledgeable, smart people designing the system in a way that will be useful to knowledgeable and smart people. I think I have this conversation with our head of the human experience, kind of our UX, our design team. Like just because you have AI doesn't mean you're a lawyer. Just because you have AI doesn't mean you're a designer, a UX designer, right?

[00:25:31] Maybe you can be better at dealing with some of those things than you might be without it. But that's, I think you kind of nailed it because you need smart people who know what they're doing to design the system. But we're also still going to need smart people who know what we're doing to be looking at the output. And, you know, I think a lot about, you know, we both have teenagers.

[00:25:55] And I think a lot about making sure that they have, how do we make sure that they still have the skills that make them have the judgment, the common sense that they're going to need in their workplace, whatever they end up choosing to do. Great. Well, guess what? Mom still doesn't know how to tie his shoes. Well, if your kid can find his shoes, then, you know, the money can step ahead. We should put the, what do they call the air tags in the shoes so that they could say, Hey, what are my shoes?

[00:26:24] And the phone's out with Siri, don't answer. Don't answer. I know where my shoes are. She's about to answer my phone. But she said, okay. But to your point though, one of the concerns I have is that we're losing so many people because we're trying to automate so much now that those smart people you were talking about, they may not be around because we're trying to say AI is good enough or not we,

[00:26:52] but certain leaders and certain organizations are saying AI is good enough that I can lose those people. Yeah. I mean, I, you know, I think we all sort of see stories about that in the news, but I, I don't know. I don't know, David. I feel like we're all so busy that AI is really helping us achieve all the things that we want to, that we're, you know, that we're achieving as part of our current jobs. Right.

[00:27:20] I think there's a lot of room for AI to just support people and actually not feel like they have a to-do list that they can never get through. So, I mean, there's still a lot of tasks and a lot of work that could make me, can make you better at our jobs. The company still needs us. Right. But, you know, but I think there's, you know, we're not there yet. So, we'll see.

[00:27:47] I know that's a lot of concerns and you're hearing in the news and college students and everything worry about it, but there's still so much that companies rely on for other people. Well, I mean, college students right now are freaking out because a lot of those jobs that they used to take that were, you know, the beginner jobs, you know, they're, they're not a lot there. And there are, you know, hundreds of people who are applying for each one of them. And so, yes, I, I've seen this happen.

[00:28:16] In fact, my kid just graduated, my oldest kid just graduated college in May, early May. And they're freaking out because, you know, they're in the cartoon world and artificial intelligence is really good at creating cartoons. I mean, there's still always going to be a need for someone who draws. Yes. Maybe.

[00:28:36] But, but you have to weigh that return on investment for that career, for that insight, for what you're developing those skills for. Or if they, if they're still around or if they're still needed. And I pray they are because I just spent a lot of money on that decree and I love my kids. Right. You know, there's, there's, there's so much that goes into every job.

[00:29:05] We're doing research here about kind of, you know, splitting jobs into tasks and really thinking about, thinking about each job on a task level. When you kind of, when you start doing that and breaking it down, there's still so much that people need to be at the forefront for, which means making sure that they have the experience and the, and the common sense and the judgment to be able to, to do that, to do those parts. Common sense. Common sense. Very interesting words these days. So let's transition to the last question.

[00:29:34] I think this is going to bring a little home for a lot of people because this asks a really important question that business leaders believe that they need a technical background to be able to talk about AI and really think about AI governance. How do they actually, how do they actually, how do business leaders who don't have that technology background make an impact in this conversation at this point?

[00:29:57] Yeah, this is, this is a great question because I mean, the short answer is you don't need to be an engineer to be involved in, in responsible AI or governance or really be thinking about how to bring it to your organization in a responsible and thoughtful way. I think some of the most AI, the most important governance questions that I think about on a daily basis when it comes to AI are actually really human questions.

[00:30:22] I always joke with, with a friend of mine here at work that I've never said the word human as much in my life as I have in the last couple of years when you're talking about technology. But, but it's really comes down to those questions about, you know, you know, you mentioned accountability when, when, when AI is wrong, but it's also leadership questions about, do I feel that my, my workforce understands, understands what AI is doing well, what AI is not doing well? Do they understand, is it transparent enough to them?

[00:30:49] Do they have, is it designed in a way that they'll feel empowered and encouraged to challenge output if it doesn't, you know, if common sense is telling them there's something wrong there. All of those questions are really human questions. I think they're HR questions. I really, I encourage, you know, people in HR to really lean in and not feel like you need to understand how the model works in order to have an impact on how AI is rolled out in your organization.

[00:31:16] The only comment I'll make to that is that to the extent at which it does get into the policies around making decisions. I don't think you need a technical background necessarily, but you may need to make sure that the process supports whether it's the technology or whether it's the human in the loop. And one of the ways I'll, I'll highlight this is the things going on with BQAM and Workday right now, where BQAM there's a, I think it was, no, it's Eightfold. I'm sorry, not BQAM.

[00:31:45] Eightfold has the issue going on with the FCRA where there's some schooling being done on employees and they weren't notified of that. You know, we all know that there are legal issues and regulation issues that, because we're in HR, that are the regulatory or legal issues that come up every day when we're talking about people. You need to be good at what you do and you need to know your stuff.

[00:32:12] And I'm talking about you, I'm talking about in general, the general you. And I will lean on your non-technical skills, the things you know and you do every day when you're going into these technical conversations around AI, because we need you there. We need to know how this technology stuff applies to the people side, because there are issues that are going to come up that the technology people don't know those rules. And we need to be there to help them with that. Now, I love that.

[00:32:40] I love that because, you know, just so there's a whole body of AI laws and regulations that are emerging. That doesn't mean the existing laws that HR professionals know and enforce and think about all the time go away. Right. So, you know, we, we, we, when we're developing products, when clients are taking in products, they think about kind of, you know, I say just because you can doesn't mean you should.

[00:33:07] Right. So we think a lot about like, when should AI, when AI can do a lot of things. Right. And what we need is really HR professionals to be thinking about, okay, yeah, I can, but, but should it, is this right for our organization? Is this right for this particular workflow or, or, or task? And, and that's, that's, that's an HR question. That's a human question for the leaders to be thinking about. And a lot of processes that we use in HR haven't been looked at in decades.

[00:33:33] And, you know, we go back to these personnel change forms, the PCFs or the PCRs or whatever they're called these days. And a lot of times those are just put into a technology like an enterprise or a vantage or a workforce now, or whatever your system of record is. And now with artificial intelligence, we may need to actually look at that. We may need to look at the workflow and in, in, you know, impose a human loop.

[00:34:01] If there wasn't one already there to make sure that decisions getting made and you don't need to, you don't need a technology degree for that. You need an HR person for that. Right. I mean, a couple of important things that you just said. One, I think we can't just stick AI on top of old processes that nobody's looked at for a while and expect a good result. Right. We, you know, as hard as it is for some of us kind of, of our generation to be thinking about, we, we've always done it that way. We really need to rethink the things that we've done.

[00:34:30] But I'd also say that, you know, the, the, sometimes the workflow, if the goal is automating and if the goal is kind of removing friction, friction is the feature sometimes. Right. So it's not minimum friction, maximum friction. And I think an HR person is really going to be very important for figuring out where do we need to design or where do we need that pause for somebody knowledgeable to think about what AI is recommending or suggesting or, or, or saying.

[00:34:59] And, and that, you don't need a technical degree for that. What you need is, you know, the experience and knowledge of, of the HR practices and, and laws and, and what's going on at your company. I love the friction example. Cause to me, when we see headlines that say HR is the problem, and we've seen that lately with certain CEOs saying HR is my problem. I'm going to get rid of all my HR.

[00:35:24] I don't think they realize that HR is created to not only cause friction, but to end friction. And there's a lot of healthy friction in HR because we are the controllers of how not to get sued as well as to have a healthy workplace. And I don't mean that in just the health and welfare side. I mean, to have a good, healthy, you know, group of people working together, you need people to help with that. Yeah.

[00:35:53] And, and I, I mean, friction is the feature sometimes, and it's not that we want to make things as complicated as possible, but I think responsible AI and, you know, HR are both. Let's be asking the right questions. Let's make sure that we're, we're thinking about the right things. We can, we can automate as, as, as much of the other stuff. What did you call it before? Administrivia? We can automate as much of that as possible, but sometimes the friction is good and we need a pause and we need to be thinking.

[00:36:24] And, and, you know, I think that's a, that's a great example of another way that HR can really jump in and, and make sure that, that AI is being ruled out responsibly. Well, I don't think I could have said it any better. And I certainly couldn't have from the legal perspective that you bring and the brilliance that you are. So, Elena, I think we're going to end it there.

[00:36:44] And I think I might ask you back again, because there's so many things we could have talked about in the context of this, but I think that's good for another conversation. All right. I'd be happy to come back. All right. Cool. Thank you so much for being here. Thank you. Thank you all for listening. Take care and stay safe. Thank you for listening to the HR Data Labs podcast. Don't forget to hit subscribe and share it with your network.

[00:37:11] You can also check out the recordings on Spotify or the HR channel now on Roku and Fire TV. Thank you. Take care and stay safe.