The New HR Playbook: Leading Change in an AI World

Welcome to HR Disrupted with me, Lucy Adams. Each episode will explore innovative approaches for leaders and HR professionals and challenge the status quo with inspiring but practical people strategies. So if you're looking for fresh ideas, tips, and our take on the latest HR trends, subscribe wherever you get your podcasts. So AI is in the news every single day, and it's kind of depending on what you read. It's either going to take all of our jobs or it's going to create entirely new ones. Maybe a bit of both. The pace of change just feels unbelievable. And yet behind all of these headlines, a lot of the organizations that I speak to are still figuring out where to even start with AI. But there are some companies who are just quietly getting on with it and doing amazing things, and Kantar is one of those. And so what's interesting about today's episode is that this isn't just a kind of tech story. This is HR right at the heart of something genuinely transformational. And so today I want to go kind of beyond the hype and really understand what it takes to make this stuff real. And I am delighted to be joined by Andy Doyle, CPO at Kantar. Welcome, Andy. Hi, Lucy. How are you? I'm good, I'm good. Now, we go back a little bit, don't we? We do indeed. We have history. We have history. What about for people who don't know? And why would you. I was at the BBC for a few years and Andy was at ITV whilst I was there. And we, I don't remember. We would occasionally meet for dinner and it would be you and I. And then there was somebody from Channel 4 that was there. The other television. Big television. Yeah, Diane. The other big television broadcaster in the uk. And there used to be this thing where whoever had had the worst six months didn't have to pay for their dinner. And I never paid for dinner once, which I think kind of just shows that my job was so much harder than yours, Andy. So, so much. But the ITV is. Is just one element of a really interesting HR career. So just for a bit of background, just give us a snapshot of your HR journey and then go on to tell us a bit about your role at Kantar. Okay, look, as I say, we. We worked together back in the media days. Good. Fifteen years ago now. Oh, my God. Officer at itv. And then I went and joined the world of private equity and became chief people officer of company called Worldpay, who was the largest PE float in London Stock Exchange back in 2015, I think that was. And subsequently that I became Chief people Officer at National Grid. That's a very different. And then I joined Santa. That's real. That's a really different. So that's kind of big utilities, heavy engineering regulated sector. And then I came to Qatar for you. Not really you, is it, Andy? It's not. I did three years. I did three years. You did a bit. I know it was a. And I learned a lot. You know, I think like you. You also. Sometimes it's when you're in the harder parts of your career that you learn more. And then I came back to private equity and back to Kantar three, four years ago now. And I'm now chief people at Agent officer. Yeah. So yeah, I should have clarified that. So it's CPO and Agent Officer. Yeah, cpao. You know, there you go. It's a whole new acronym for us all. It's quite interesting that, isn't it, that it's. That the whole. And that's the agent responsibility for the whole of Cantor, not just for hr. That's the whole. And what was the thinking behind putting it with hr? I suppose there's two bits for us. One is the future of the workforce is a mixture of people and agents. And we don't quite know how that's going to play out, but we know that those two things will come together in how we do work. And secondly, the technology is not the issue. The technology works. You can make the technology work. And the issue is adoption and getting people comfortable with using tooling and using the environment that we can create for people. And that is a human challenge. That's a human change challenge. And we didn't want to put it in technology. So it feels like this is a technology piece. It's very much a human change agenda that really belongs in HR is my view. I'm a big advocate of that. And one of the things I say to HR leaders when I talk to them is if you are letting your team technology function, take the lead on agentic deployment and the whole AI agenda, you are missing a beat and you will not be successful. Because actually it's not the technology that's the issue in this case, it is how you use the technology and how you drive adoption and change in the workforce. And that is an HR challenge. Not. Yeah, well, I think that's really progressive of you and Kantar to do that. I've just come back from a big HR tech conference and exhibition in. In. In Amsterdam and, you know, there's all these amazing products and tools and, you know, just blow your Mind. But as you say, every single lecture or presentation I went to, everybody was talking about the human challenge of it. And whether that be the adoption piece or whether it's even things like, you know, if you're a manager and you've got a range of people, you know, underneath you that know more about AI than you do, how lead them? How do you question there's all sorts of things, aren't there about kind of, you know, entry level jobs being kind of disappearing and what does that mean for people's ability to learn how to. Critical thinking and questioning and the kind of fundamentals. So I, I think, I think it's absolutely right that it should be with hr. Well, I would say that, wouldn't I? But we might be biased about that. Yeah. Maybe that I've learned with this is the tooling that now exists is democratized technology in a way that I can't imagine happening a few years ago. So I can now build really quite complex agents myself. I can't code. I make no pretense about my ability to code. As soon as you start talking about Python, I kind of glaze over and I get a bit lost. But I can build really quite complex agents now without any technical background. And I think that is a really interesting evolution for us. All the, you know, and one things that we've all found in Kantar is, you know, we have more agents than we have people. We have more agents in HR than we have headcount in hr. All of those agents have been built by colleagues who work in hr. They are not built by anyone in technology. They are built by, by our people who understand and process and the problem they're trying to solve. And that is a really big change from where we were a few years ago about, you know, you had to go to a workday or an SAP or a big system. You got to standardize your process. Yeah. Try and drive everyone into a standardized process. Then everyone customize that standard process. Yes, of course they did. We all started off, it's going to be vanilla. We're just going to follow the process. No one ever does. You customize one of your big ERPs, you then can't make it work. You customize it again, you can't get the upgrades. It takes a long time. Takes too long. And then you've also spent, you've doubled your budget and you're having to go to finance committees and you're having to rely on capacity within the tech team and. Yeah, not anymore. I've never known any one of those Projects be delivered on time, on budget. Normally all three of those go wrong. But so against that world you're now operating in, a world where you can develop agents in hours, might be very simple, and you can develop complex agents in, in weeks and you can get the time to benefit and the time to adoption is much quicker. And that's a fundamentally different fundamental. Just give us a, Just give it. Because people may not have heard of Kantar, but I mean, obviously they're massive. But just give us a sense of what it is that you do and scale and just give us a sense of the kind of the number of employees and locations and just give us a kind of an overview. Yeah. I mean, at heart, Kantar helps marketing teams understand how their brands are growing, what consumers think about their brands, the data and analytics behind consumers and brands. And if you think about all the noise that exists, all the signals that exists on social media, on, on TV advertising, on online advertising, how all that manifests itself to a brand saying, I need to, you know, I'm spending $10 million, I'm investing $10 million in my brand. How do I get the best return for that, to make the best return for me as a company? That's what we do. We help companies understand all that data through surveys, through signals, through our own IP, and we work with about 98 of the top 100 advertisers in the world. If you work in marketing, you will have heard of Cantor, maybe worked for Kanto at some point. Today we employ 12,000 people in 63 countries around the world and we cover about 30 more. So pretty global, you know, very focused on helping brands grow. Brilliant. Right, let's get into it. I want to try and I suppose break this conversation down into four key parts. The sort of why you made the changes that you did within hr. And when we're talking, looking for the purposes of today, we're not looking at all of the agentic work within Kantar. We're focusing purely on hr, but why you made the changes that you did, what you actually did. Because I, you know, I went and met you for coffee recently, you told me what you did and blew my mind. How you made it happen, including some of the messy bits and the learnings and, you know, where, you know, you would have done things differently, had, if you had your time again. And then I'd like you to kind of give your view on what you feel this means for the future of hr. So we'll get into that in a bit, but let's focus first of all kind of. What was the business challenge that you were aiming to solve with this? Was it just one day you woke up and thought, I think I'm going to build some agents, or what was it what was going on for you in HR and the HR team that made you think, right now's the time? Yeah, it's been a really interesting 18 months and a couple of things came together by happenstance. And I often think it's the sliding doors moments in your career aren't there, that you go, I wonder, if I hadn't gone to that meeting, what would have happened? And I went to a meeting in October 2024 with Microsoft in Seattle, where their headquarters in Redmond, when they were showing, it was the very beginning of kind of copilot deployment and Microsoft copilot deployment. And I kind of sat there and thought, wow, that is, I mean, curious. You know, it was like, it can do all these things and it's amazing. And you go, wow, that's a. We got to use that in hr. I had no idea how to use it. I had no idea what we do. At the time. We had like, many companies had 300 copilot licenses as the special preview. And you found out every company of scale in the UK had 300 copilot. You weren't that special. Best thing ever. But we all felt special. Yeah, we deployed those copilot licenses and said, let's just understand what we can do. We go gave 100 to our senior leaders and 200 to our colleagues in one location. And we said, go play. Just go see what you can do. Go learn. At the same time that happened, Qantar had been on quite a big journey. We'd been private equity owned for five, six years. We were much bigger. We were kind of breaking the business up into component parts and selling the component parts. And we're now down to two big businesses, but we've gone from 30,000 people to 12,000 people as we've divested businesses. At the beginning, the HR team had about 700 people in. We still had about 350, and we'd reduced the cost by about half in that time. I still pretty good about life and the board. One of the things I really like about the board is they gave us a challenge. Yeah, challenge that said, why can't you half, you know, what would you need to believe to half your cost again? And I was kind of like, whoa. And because I was going, that's not, you know, kind of reducing that headcount down by a couple there and Giving. Increasing the ratio of employee employees to HR there. That's fundamental. Yeah. And I think one thing that I really like about working with the RPE owners is that they try and push you to challenge conventional thinking because it wasn't often an aggressive way. It wasn't asked in a. It was, what would you need to believe? What. What would you have to frame? What would you have to believe was true? And I kind of went, I don't know. And actually what we need to believe was we're going to have to blow up the entire HR operating model and reimagine a different HR operating model driven by agents. And being your operating model was what standard? Kind of Ulrich, you know, centers of expertise, business partners, a shared service that. I mean, I suppose I've used that Ulrich model everywhere I've worked for the last 25 years. And I was always like, I didn't think there were enough. There was. Was there ever a time before the Ulrich model? I mean, I didn't, you know, I just. That had been the defining thing. That's how you organize hr. And so we had the agent piece coming along like this looks really interesting at the same time as the, if you want to reduce our cost by 50%, you'd have to do it differently. And that was the genesis of the journey. I mean, it started off with what would we need to do? And at the time, the idea that we could use agents in the way we now use them was fanciful because our first agent foray was complete failure. And so. And actually, I think that's one of the things that we've been really good at as an organization is failing and learning through failure really quickly with our agenti work. When you say, when you say it failed, what had gone wrong? Well, we just went, I know what we'll do. We'll take a copilot because this looks very credible and very capable and we'll put it over our intranet and it can answer all the HR questions. And what is exposed was, yes, it can do that. But the data that was on our intranet was. Now we had 23 versions of the same policy. The policies were written in different ways in different formats. You know, just from a policy point of view, you know, HR functions are not very good at keeping policies up to date. They're written by different people at different times. If you just take the core policy suite that you would have on a. On sickness, on absence, on holidays, on whistleblower, you know, they're fairly. Nothing very special. But the kind of core Components of an employee handbook and a policy framework. We have about 60 policies per market. Some have a little bit more, but when you operate in 63 markets globally, that's about four and a half thousand policies. Yeah. And they've all been written with different, you know, some were guidelines, some were, yeah. Some detailed, some very specific, some user guide, you know, some looking very pretty because we tried to engage colleagues to make it easy to read. Machines don't care about graphics and how to make it easy. And, you know, we found 23 versions of the same policy hidden on various parts of our website because that was a 2009 version that hadn't been archived and Copal had loved them all equally because without guidance goes, oh, you asked me to find a policy, I found it. I've got it here. Here you go on my clever. And the fact that it might be wrong and you've got hallucination, you've got all sorts of problems. You went, okay, so the big idea has just failed at the first hurdle. Yeah. And from that came, we've got to do this differently. And we started by like, let's standardize all our policies, all 4,000, I think it's 4,253 or something like that. Let's standardize and put them into one standard frame. That there's this, you know, put them in one place, in one style, in one turn of voice choice. And we built an agent to do that. And so we're getting into the what. So just to kind of just recap then. So motivation. The why was all around, yeah, cost challenge around headcount reduction, but quite dramatic. And, and I like the way that they framed it. The challenge, it wasn't, you've got to find 50%. But what do you need, what do you need to believe to be true if you were going to radically reduce headcount and, and cost, but also it around the kind of opening your eyes, the exploratory stuff, the connection with Microsoft, the freelance is the. So that kind of mix of. Was there. Was there anything that you were. Were you frustrated with the model that you had? I mean, you mentioned that you'd had it forever. It was just. This is how it's always been. It's working fine. If we, if we hadn't had the provocation from the board, I think we'd still have exactly the same model that we had. And because I went, that's what I know. And it's, you know, the advantage of having been doing this job a long time is you go, I know it works. Why would I change it and change it for a model that I don't know works? I don't know. One of the things that was really helpful from the board at that point was when we went back and said, I think I could do it. It looks like this was. I said, but here's the risk profile. I don't know whether this is going to work. What's your risk tolerance for failure? And I think that's where our board went. The worst thing that can really happen is you make a complete Horlicks of it. You know, you make. It all goes horribly wrong and you have to revert to a different model and it will be embarrassing and it'd be difficult, but in the context of the benefit, we'll take that risk. So you've got that top cover. You'd got a degree of tolerance around risk that gave you the space, which is, you know, not every CPO has. It's kind of, you know, yep, go away, play experiment, but for God's sake, don't let anything go wrong, which is not helpful. So the conditions were good in that sense. You'd got a fantastic challenge, which meant that you couldn't just salami slice and do things, you know, sort of slightly less well or slightly differently. It had to be radical. You'd got these new sort of ideas and tools and you'd got the space to make it happen. So you started to talk then, and apologies for taking you backwards, but you started to talk then about first step was we had to consolidate the policies into something that copilot could understand and presumably prioritize in some way or handle in ways that aren't linear, that there might be certain policies which superseded others or so because it's. We're dealing with human beings, aren't we? So it's not just what's the policy? Here's the answer. It's also, well, who's the manager that's asking this and what's the precedent and what's gone? You know, so it's a complex area. So take us through the consolidation piece and then the. What it kind of ended up looking like. Yeah, I think there's two pieces of our journey. One is you've got to underpin in order to radically shift your model, you've got to create capacity in the model. So the agent work is quite good ways for us to create capacity, to let us reframe the model elsewhere. So when we talk about our agents, the agents that we built in HR are all designed to try and Free up human from work that they do today that is predictable, dull, monotonous. It frees up time. Yeah. And so it frees up capacity. And then you can choose what you do with that capacity that you create. So using the example of the policy, it would have taken, it would have been the work of years for someone to review 4,000 policies, put them in a standard template. I mean, they've lost interest after about policy four, we created an agent that did that and said, look, this is how it's easy for us to create a common language. We created a set of agents. And one of the things that I think we learned in playing around with this at the beginning was the more complex the agent you create, the more risk you carry and the slower it goes. But if you create very simple agents that do just one task, you can make real progress really quickly. So rather than trying to get an agent that did everything, we created a set of agents that said, this agent is designed to take the policy and rewrite it into a standard format that covers all the things that we need and identify if there's not enough information in the current policy to complete that task. It didn't worry about tone of voice. It didn't worry about legal compliance at that point. Just, here's a clean version of your policy in a standard way. We created another agent that said, I am your tone of voice guidelines agent and I will take whatever you've given me in whatever language you operate within and I will put it into a tone of voice that is consistent with our brand guidelines that exist today. There you go. The third policy would do a translation because maybe we would operate the policy in, you know, a local language we don't have. English is not our natural language of the organization. We are inherently quite a local organization. Whilst English is probably the most spoken language, we translate most of our documents into seven major languages and our engagement survey is taken in 29 different languages. And 99% of people take it in their natural. In their preferred. Their preferred language is their home language. Even if they are perfectly fluent in another language, they typically take their home language. So we had a translation agent that said, if you are translating this from in, you know, I, I don't speak any languages. I want to have a look at some policy. Can I read it in both English and German? It will give me a translation side by side. And importantly, not only will it do a translation, it will highlight phrases and terms that don't translate well or mean different things. Like termination of employment in the UK is, is not, you know, in some languages, couldn't be translated as death. That's an unhelpful policy frame from a HR point of view. The agent, the translation agent just handled that, didn't just translate literally, but identified, was sensitive to nuance. And then we had an agent that was a legal compliance checker that said, I go out and look at this policy that you've written relative to the employment guidelines that exist in this market. How high risk is this policy? Are you compliant? Not compliant. Should you go and engage in further review or not? So, and as a result, we translated all 4,000 of those policies into one common framework in a week. In one week. So that is years worth of HR work that is really dull. And was it perfect and tons of legal fees and was done in house in a week. And actually the thing that slowed us down was actually feeding the policies in because we hadn't got an ingestion engine that kind of like did it quickly for us. And I love the fact it's taking you a week and you're saying the things that slowed us down. Yeah. But, you know, this is. This is the sort of stuff with. If we were doing that today, we have an ingestion engine at the front to kind of push the folders into it to automate that. It goes into the machine and you save them in a policy document at the end. We didn't automate that, but that was still humans doing that bit. Yeah, I would just automate that if I was doing it today. So these are just. You've now got all of these amazing. You know, this policy stuff, it's streamlined, it's up to date, it's legal. Legal in the jurisdiction in which it's based. Etc. Then you're now doing the bit which is the human piece, the answering queries. And. And I think when we met for coffee, because this is the bit that I found the most exciting was you talked about different levels. Yeah. I think what we tried to do is it. And I, I, anyone who's listened to this, I would say that the policy agent is very factual. I refer to it as just Dutch. Any Dutch listeners. It has a Dutch directness, which is brutally true. If you say, how many days could I have for this? Or am I entitled? Under our policy, it will go. The answer is five days. It has no nuance, no subtlety. It's accurate. It doesn't care how you are or how you're receiving this message or whether you'll be upset by it. It's just the truth. Yeah. This is the single source of truth. So as we started playing with this, we went, okay, good news that we've got clarity. But the delivery of that matters. Yeah, because if someone is upset or if someone is searching for information about a life event or, you know, you know, they're thinking about, you know, someone is pregnant, they're trying to understand what their rights are, or someone is being, you know, wanting to raise a discrimination claim, or someone is trying to raise an issue about bullying, you can ask the agent about that. But if you don't have subtlety and nuance associated with that at a time when you might be like, if you're just asking, how many holidays or days have I got there? Then you don't care. But there are lots. But as you say, you either got some amazing, exciting life events happening for you or you might be going through some very difficult, challenging times and that might need something a little bit less brutal. And just as we would expect a hr, a really good HR business partner to be able to read the room a little bit. Yeah. And go. Although I have met one or two that are a bit more direct, but generally HR people would have an emotional empathy that says, you know, someone just told you they're pregnant, irrespective. If you're really thinking, how am I going to cover that vacancy, start with congratulations, you know, like. And the agent does that. So the agent tries to replicate some of that. That. So the policy agent is very direct and straightforward. Sitting alongside that is an emotional agent, a sentiment agent, essentially, which goes, I will intervene if some. On certain life events. And we've trained it to do certain things, to be a bit more human, to say thank you or to, to acknowledge that this must be a difficult time. So if someone's asking about compassionate leave, you don't want the answer to be, well, under our policy, you have X days to take off at funeral if the relative is one of these conditions. And this, you know, yeah. You, you want something that says, well, the policy is like this. It must be a very difficult time. Here's another support available. The policy is a guide. It's not an absolute. This is what you should do. Let me help you. I can direct you to other resources. I can do so. It's more helpful. But if you then let the policy and the emotion or agent come together, you, the emotional agent, might get a bit carried away and try and be too helpful. So you need to kind of put a conditional element around it to go. But I only want you to operate in this framework so we have what we would call a boundary agent that says. And we're constantly learning how to tweak the boundary agent up and down as we get more experience of things that are working and things that don't work, that kind of goes. That's the judgment bit, isn't it? Which, as you said, you know, you've got, you know, a great HR business partner will give you the right information. So that's your first one. It will apply emotional intelligence, and that's your emotional agent. But it will also have this wealth of experience that will just go red flag. Not comfortable with this. And that is so contextual. And how on earth do you develop a boundary agent? And then maybe you're always going to have to be flexing this. I don't know. But it's just will be. We'll always be flexing it. And I think you're trying to go that you and I, with lots of experience in hr, just sometimes you come across a situation where you go, oh, yeah. And on the surface it all. Yeah, the questions are all exactly that, but there's just all of these other factors that your Spidey senses just pick up. So, look, we can't create the Spidey sensors yet, but what we can do is we create conditions where we go. We're worried about the boundary. So I'll give you an example. When we started doing this, we didn't want maternity pay calculation done by the agent because we went, right now we're just starting on this journey. Let's just make sure it can do all the warm, fluffy piece of. This is what you're entitled to and this is what you can do. And here's the support and the information. You said you also could get it to, you know, suggest you want me to write the letter to your manager, letting him or her know and super helpful. And it does all that. Really helpful. Yeah. But we didn't want it to calculate maternity pay because we want a human to do that right now whilst we learn. Actually, the agent started calculating. We'd given the agent the ability to do that in our coding or training of it. So it started being helpful. It's like, let us show you. And this was something we built up in the pilot. And the boundary agent wasn't shutting it down. So we learned two things. One, you've got to run this in pilot and you've got to understand what the agent does. So don't release this to your organization until you're really confident. Yeah. You know where your risk profile is. So how Long were you training these three working in tandem? The reality is there's about 40 agents that work in tandem because every interaction requires a different agent and it's in a wrapper. So we've been training them for about six months now. It's live within the HR community. We've been using this down with our service delivery teams to answer queries. So they're using it in. And actually that's an easy way before you release it to the organization, to just release it to your service delivery teams. Because with a global team, you know, if I'm based in India, do I know what the French employment law is on this, I've given that team a really, really good tool. Yeah. To be able to answer with more confidence. So some of the team would go, I knew the answer. And some of the team would go, oh, I didn't, I didn't know anything about French employment law or whatever it might be. Yeah, that team been the biggest advocates of this system because they built it with us and they'd be the people that said, I solved the, you know, when we started, they solved the query in the old way and then pushed the query through the agent to see what it looked like, you know, to give us confidence in that and that that highlighted sometimes we, we had gaps in our own knowledge where we're going. The agent is suggested doing something and it's different to what the agent our human has done. Is it because the agent is wrong or is it because the human is wrong? And sometimes it was the human who is wrong because, you know, we also talk a lot about hallucination and machines. You can train out hallucination and tighten up hallucination, but humans hallucinate too. They just get stuff wrong and if they didn't, and I'm sure in your career as a senior HR leader, you occasionally got things across your desk. You. How did all that get to this? So, yeah, we've developed the agent, we've got it across the people community. We're launching it to the organization actually next week. So it will become the only way to talk to HR from next week where we start bringing it to the organization scale. But that's, and I just, you know, that's one big agent. But we've got loads of other agents across ta, across reward, across comms, across business partnering, all of which are helping people to free up time to do better. And, and it's been, it's, it's meant a big reduction in the number of HR headcount, doesn't it? Yeah. I mean, we probably gone from about 350 to about. About 200 now. I think we don't know where this will take us and we don't know where we're making choices on capacity. And there are some bits of the work where the work is genuinely going away. And our talent acquisition team is quite a good example of that. We've automated and put agents across systems and agents across our talent acquisition team. In the last 12 months, we've doubled the productivity of the team and half the team size. And that team is hiring the same number of people they were hiring 12 months ago. Amazing. Quicker pace and the team are more engaged. So in the engagement results we just had back, that team are happier because what we've done is the jobs that people are doing are more meaningful, they're more interesting. They're doing the stuff that recruiters love doing, which is interviewing people. They're doing the stuff. Less of the stuff about writing up interview notes. That's all done now. That's all automated. So we've made the jobs more valuable and actually better paid. And I think there is a trade here about where do you take jobs of the future? Because I think you might have, in some teams, fewer roles, but they'll be better, more valuable to the organization, and the people doing it will be more valuable. The experience of managers, leaders, employees who've been on the receiving end of advice, guidance, support from the HR team team who are now interacting with agents. What's been the feedback from them? Look, I think one of the big things that we've had to break is the manager, business partner relationship. And this is one that asked me in a year's time. This was the one where I went when you told me. Yeah. So we've done two things. We had about 60 business partners of varying levels of senior Rossi, I'd say about a dozen really, really experienced business partners, Very, very senior, very, very capable. And then about 48. Exactly 48. In fact, I know it wasn't about. It was 48 business partners. You know, they're very good, capable business partners. And we split that team into two. The 48 essentially went into a resourcing group where they all work together and the 12 remain loosely affiliated to parts of the business. The 48 no longer have manager relationships. So they would have all had managers and client groups. And we took all that away and we took their workday access off them, so they can't see anything in client groups anymore. So the. So the manager and leader, because there's always been that Classic thing, hasn't there? You introduce technology or you introduce shared service center or whatever it might be, and yet the business partner, because they care, because they, they don't want the manager to be bothered, they don't want the manager to have to do the admin or whatever it might mean. So they act as the intermediary. And you've basically removed that. Yes. Were you the devil incarnate for a period of time? I think I probably still am. I think two things have happened. One is the HR community have been very understanding and frustrated, but at least understand that we're trying to do something different and we've been very transparent about the HR community. Is everyone happy about that? Absolutely not. When I took workday access off people, that was about 12 weeks ago. Now it's a little less noisy, but it's still noisy. And this isn't that easy for people. No. What we have found, hugely transformational for people and their skills and the role and how they feel that they're adding value and their contribution. And this is such a massive change. I think this is one of the big unlocks, though, for me, because in our old model, the service delivery teams did stuff. They couldn't do everything because they could do anything where the answer was clear, the answer wasn't clear. They needed to escalate to a business partner and the business partners don't want to do that sort of work. So quite a lot of things got stuck. I'm too busy being strategic to do that and I want to be strategic. I said, don't we all? But, you know, like, I still do operational stuff. I always say this, you know, go. I mean, I can actually count on one hand, I think, the number of times my CEO asked me anything truly strategic, you know, it was, most of it was kind of, can you fire my PA and can you recruit me a new one? One, you know, it's like to do that, no matter how senior you get, you still have to do the operational, don't you? Yeah. And I think one things we found is with moving all the 48 business partners away from manag groups, we've unleashed a huge amount of capacity. Absolutely. What we actually found was that you got 48 people really committed, really capable, looking for work. So they meet managers, they do stuff for managers, they help managers, and they navigate lousy processes in your function to make it easy for people in doing so, you never fix your lousy processes because it all works. And because Those people, those 48 people turn up to work every day trying to Do a good job. They keep being busy. So they go, I've got a problem here. I'm going to create a new development program. I'm going to create this new idea. Engagement is not very good in this part of the business. You're going to do this. And managers don't take responsibility for some of that. So we've stripped all that away. Look, it's not perfect yet, and it's a journey, but we stripped that away. It has created huge capacity. And I've had a couple of those business partners say, I just don't have enough work to do. And I said, look, that's the nature of the model right now, because we haven't quite transitioned the workforce planning and the strategy to that group in a way that I'd like. Because what we're trying to create is I've got 50 people who, you know, who are very, very capable at solving problems and don't need supervision. And so what would be saying to the chief people officers is if you know the answer that you want, the output you want is probably a service delivery request. If you know the problem you're trying to solve, it's probably goes to the business part of the community. And that's taking a bit of time for us to land. This is, I mean, this is such a, as I say, it's such a huge shift. I just wanted to touch on the fact that you built this pretty much within your own HR team. And I think there was one person in particular because, I mean, I, you know, you could imagine people listening to this and thinking, well, they've obviously got some super tech whiz kids in their organization and you may have. But wasn't it not your payroll manager who led on this a lot? Yeah, I mean, he was on the very first people go copilot license too, because he happened to be in the, in the office. I mean, it's happened starts again. He happened to be in the office that we gave everyone a copilot license and he ran payroll and service delivery for us and he's now our AI manager. But actually, I think it's a mindset. I don't think it's just about a person. Anybody can build this technology. Right now. There is no one listening to this podcast. There's no one in HR who cannot build an agent. The only thing that's holding you back is your fear of building an agent. So, you know, go on to Claude, go on to Microsoft, go on to OpenAI or ChatGPT, describe the problem, ask for help and Start there and you might be, you know, there's massive amounts of tools, information available. I, I'm proud actually that as a HR function, we are the highest user of copilot in the organization. We have more master users. That's brilliant. Than anywhere else. Yeah, no, that's brilliant. And I think, you know, I'm proud the fact that in HR we've tried to upskill our people to build agents. And if I look at our top 10 agent builders inside the organization, four of them are in HR. So it's not just one person we've upskilled the leadership team. You know, we've you. Anybody can build an agent. I build agents all the time. I've got 30 or 40 agents that I've built to help problems. And some of them are very micro. You know, I, I built an agent for two of my HR directors that didn't understand a reward issue because our reward stuff is quite complex. Yeah, I just built them an agent to help them be able to navigate hiring someone new in our reward schemes. You know, it took me half an hour, 40 minutes to do. Really simple to build. But it just meant everything I know about this topic I could give to a couple of my team in a way that they can ask questions, they can interrogate it. It's not all agents need to be big and, you know, orchestration and interconnected. And going into work day, lots of the agents you build are knowledge agents that you're just transferring knowledge between people. Just reflecting on. I mean, you've kind of done this as we've gone along, but there may be some other bits and pieces as well that will be useful for people to hear. If you were starting again, knowing what you know now, is there anything you would do differently? Not at all. Different order. Is there anything kind of that you think, oh, we move too quickly on that or not fast enough on that? You know, what's the, what's the, the kind of the learnings for you? I would probably double down on training the HR community. I think that one of the things that for us as a community is we're a bit afraid of tech. And there aren't that many people in the HR space that genuinely understand technology. And I think you could do some really good foundational basics in technology that will help. What is a large language model? How does it work? How do you create orders built in? How do you avoid it becoming a black box where you can't see what it does? These are things that I think we've had to learn. Yeah, we've crazy agents are a bit black box thinking. And you go, the answer is 42. And you go, where would you. Yeah, where's your workings out go? Your workings out. How do you remember that in maths you always had to give your workings out? Yeah. And, and actually that's what we've learned is like from an audit trail ability, from governance point of view, from, from a control environment point of view, you need to see the workings. That said, what part of the process, when you know, did it, you know, did you interpret the policy correctly? Tick. Did you then make a recommendation on. Based on the policy? Tick. How did I evaluate whether that answer was correct? Did I keep a record of that? How have I built the records and the systems and governments to do this? I think when we started we were probably, and I blame myself for this, we were trying to drive progress and we've had to kind of come back around some of that. I think if we'd have built that in at the beginning, we wouldn't have had to rework some of the agents that we've done. Because when you can't try and get to production, rather than a proof of concept, all those things matter, particularly if you're a regulated industry or something like that. And you go, well, why did you give them that person that advice? You can create the record role. And we, we kind of came back around that loop later and I think the, the second thing I do is there's a big barrier on here about adoption. The technology works, it's the adoption that matters. And therefore if you can create that capability in HR of a willingness to fail. None of us like failing. But with this approach, you've got to fail. You've got to build and iterate and try and not think of it as failure, but think of it as, I just haven't solved the problem yet. And as you work through that, you build capability in that. In the HR function which transmits to the organization. And one thing I really believe is if you'll be able to show the art possible in HR about your intervention with a manager. About. Here's the job description came out from the agent. It's just a touch point for the manager going, can I do that? Or here's the reconciliation agent that reconciles the finance headcount and the HR headcount every month. A great little example for me of super agent. We built the head of our head of analytics built. And she said to me, I spent eight hours building this agent at the weekend. I was so excited because now I Fixed it and I created the audit trail. This used to take someone in my team a week to do everybody. I remember that, the bun fight with finance every time FTE versus et cetera and trying to reconcile those two figures. And we now do that in two minutes every month. So you know, you go from that. That team is a team of three people. And if one person in that team of three is spending a week, a month on a reconciliation piece and you've now created an agent that does that in two or three minutes, the capacity analog you've given to that team is vast. It's not, you know, it's like at a functional level it matters not a job, but you know, that's 12 and a half, you know, 25% of one person's capacity in the team. So what else? What do you want to do with that time question? Oh, I could talk to you for hours about this, Andy. I think it's just fantastic. You are really well ahead of the game. I appreciate that there'll be lots of people out there who will be already catching up and doing wonderful things. But I think it's what I hear from HR people at the moment is that they're using it personally for productivity rather than on an organizational systemic level. And I suppose that brings me to my final question to you and just to get your thoughts on this. It's a worrying time for HR people. They can see roles starting to change, they can see roles starting to reduce or disappear. And when AI can do what humans can do in certain areas faster and in many cases is better, what does that mean for HR skills and mindsets? Is the function potentially irrelevant, redundant? What does it look like when you think about it? Absolutely not. I think actually the reverse is true. I think the function has never been more important and I've never been more optimistic about the future of HR than I am right now. I just think it'll be a different future. I think it'll be a future where human centricity is going to become more important because the risk of agents is that you can automate everything and you can take the human out of the loop. Now I don't think organizations going to want to do that. I certainly wouldn't want to do that. And I think as you drive an agentic workforce and a different workforce, I think of it as a force multiplier for your people, not at a replacement of fewer people. That the importance of judgment and you know, people still want to buy off people. You know, we're a business to business client Relationship management with expertise of our people. I, I think, you know, as you seller of services to clients, buyer of services, you'll want to be able to do things that only people can do. You won't want to have a drink with an agent, you won't want to go for lunch with an agent. You know, and I think the idea of, you know, my agent talking to your agent to agree something that neither of us want to do, it is a fallacy. Humans are unique and technology is in service of us. It's got put guardrails around it and there'll be some companies that probably get it wrong and will learn from some stuff. But I also see new jobs, new opportunities rising. You know, I've got a whole AI team now in hr. Those jobs did not exist. I the idea that we would have a HR AI manager and HR AI analysts and that we would have agent trainers and orchestrators and people managing and monitoring the agents and thinking about the joiners to leave us process for agents, you know, that that's all stuff of like, whoa, this is new. So I totally agree with you. I think it's so exciting. You know, you and I have been banging on about kind of human HR and judgment and you know, that kind of broader capability not just being transactional, process compliance oriented. And, and it's always felt a little bit out of reach until now. And I think it's going to, that's where the future will be. And therefore how important is it for your people to have skills in mediation, have skills in dispute resolution, have skills in coaching and organizational design and team effectiveness? And how do you drive high performance through people enabled. Enabled by the agents? Not exactly. Yeah. And I think that's, that's. I'm going. This is going to be a, I genuinely think this will be a glorious decade for HR involvement. But if it is, we've got to lean into tape moment. And therefore I encourage every HR director, every HR person I speak is like, get yourself upskilled. Listen, learn, be curious, fail. Look, some guardrails don't, don't go posting your sensitive data into a large language model. You've got to be careful about some of this. But don't be afraid. Don't be afraid. Well, I think that's a really positive note to end on. Andy, thank you so much for taking the time and sharing your experiences and the journey that you're on and other cliches. I just think it's so exciting and I really appreciate you coming, coming and joining this podcast today. If people wanted to connect with you. Is LinkedIn the best way of doing it? Yeah, just drop me a note on LinkedIn. I kind of post there about some. I started posting about some things we're doing as a HR function and also some tips to get started on working on agents yourself. So there'll be a couple of things that are appearing on LinkedIn to try and be helpful and try and take some practical. Right. That is great. And we'll put your LinkedIn URL into the show notes so people can connect with you. Then let's leave it there. Andy, thank you so much. Thanks, Lucy.