Andy Doyle: Let 15,000 agents bloom

Most tech organizations would say, “We’ve turned it on. We’ve deployed it in the ecosystem. It’s there. You’ve got Copilot.

” But if you don’t know how to use it, if you don’t know how to experiment with it, or you’re scared of building an agent, those are the skills that you need to develop. And people are scared of this technology. They don’t like getting things wrong. They, you know - they’re scared to fail, And actually, the thing I’ve learned about using agents and building agents is you’re just iterating.

You’re not failing. You haven’t succeeded yet, but you’ve got to take some time to use it to make that work. Welcome to Work Lab, the podcast for Microsoft. I’m your host, Molly Wood.

Our guest today is Andy Doyle, Chief People and Agent Officer at the marketing data and analytics company, Kantar. Andy has been leading one of the most ambitious enterprise AI transformations in the non-tech world, and rethinking what the people function looks like when agents become a core part of how a company operates. Andy, welcome to WorkLab. Thanks, Molly.

Nice to be here. Let’s start with your title, because it recently changed from Chief People Officer to Chief People and Agent Officer, which is a strong signal that not only is AI and agentic work important, but also that you’re combining these two functions, right? I think that’s right. I mean, one of the things that we were thinking about is what does the workforce of the future look like?

And I think it’s a mixture of people and agents. When you separate those two things, you kind of optimize for the technology or the people, and we were starting to think about, well, what does that mean if you put those things together? How do people and agents work together? And we’ve started that journey.

We don’t have all the answers for it yet, and we’re discovering lots of things about it, but we’re on the journey now. People, of course, are a big part of this conversation in a lot of different ways, and one of the things that this title does is sort of put you right at the center of the conversation about maybe some of the anxiety that employees have about how their jobs might change and whether they will still have the same jobs. And I wonder how you think leaders should be talking about that with their teams and how you think about that.

Well, I think people are very sensible, and inherently they see the world of work changing around them. They see their behavior as consumers changing around them. And I think leaders’ job is not to try and minimize that anxiety, but face into it and to be honest. And I don’t think any of us know the answers to all of this.

What we do know is that work is changing and that change is not evenly distributed. So there are some jobs that will change. I mean, just as there were where there were typing pools, and receptionists, and switchboard operators, those jobs changed. And I think that’s true in the world of work today.

We just don’t quite know how or when, but what we’re trying to say to our teams and our colleagues is, how do we help equip you and give you the best skillset to be able to adapt and use the skills that you have in this new world? Before we get further into the transformation at Kantar, when you think about being the chief people and the chief agent officer, you know, sort of again, going back to those signals that those two things are really interrelated, but also that agents are now employees, do you think of yourself now as the person who’s in charge of all of the employees, whether human or digital?

We do, but it’s a bit of a mind shift for us all, -Yeah. That’s something that, if you’d have asked me a year ago, I’d have gone, “That’s weird.” Today, I think we’re still trying to work that out, and the technology is catching up with that thinking and giving us better opportunities and better way to monitor and create visibility of agents across our organization. Talk to me about this transformation at Kantar.

We’ll get into some of the details, but, you know, I mentioned at the beginning of this interview that this is this massive transformation in the non-tech world, specifically. Can you just describe, in brief, what has happened and where you are now? As you say, we’re not a tech company. We’re a market research company.

We’ve got researchers, our culture is based on order and caution, and we do longitudinal research to help companies around the world make the best decisions they can about their products, their future, how their brand grows. In that context, trying to say to people, “We want you to deploy technology, embrace new technology, change technology,” It’s like, “Wow.” It’s like our people went, “Are you sure?” And we rolled Copilot out.

About 18 months ago, we gave everybody a Copilot license and we said, “Let’s experiment, learn, fail.” When we went through that process, something woke in the organization about the capability of the tools, and what we found were kind of mavericks across the organization going, “Look what I can do!” And we probably leant into that and went, “Okay, how do we amplify that?” And that’s where this all came from.

It wasn’t some grand plan. It wasn’t like, “Oh, I know what we’re going to do. We’re going to have a strategy of do this, and we’re going to combine people and agents.” It’s kind of evolved over the last 18 months, and as we continue to learn, we’re continuing to change the way we think about agents and the way we think about the interaction between people and agents.

I love the boots on the ground perspective, because we sort of talk about what the playbook could be for companies and where to start and what to do. Did you have a business goal in mind? Did you have a line of business you were trying to cannibalize? Were you thinking this is the ideal outcome, or were you just like, “Let’s go?

” We, like many firms, had a few Copilot licenses to play with, and we put them into one of our lhistoric parts of our organization where we’ve been in that office for 50 years. It was about 200 people in that office, and we gave everybody in that office a Copilot license, and we gave a hundred of our senior executives a Copilot license, and we said, “Go play.” At that point, we had no expectation as to what this was going to be. It’s amazing that it was only 20 months ago.

I mean, the world has moved so much in that time, but out of those 200 people who were in core research, in business development, in client relationship management, in analytics, in HR, payroll team were based there, they just started coming up with ideas. They just started experimenting, and out of that, you know, we just thought, “There’s something here.” We don’t know what that something is, but we kind of took a bit of a leap and expanded that Copilot trial and then went to everybody 15 months ago.

And it was the point we went to everybody, it becomes a way of working, a different way of interacting. And I think that was the piece. And when I talk to colleagues in other firms that have got 10% of their population with a set of tools, it doesn’t work. You stay in pilot mode.

Until your organization is all embraced and all in on whatever tool you’re working with, that’s when you start getting the breakthroughs. That’s when you start getting the change, and that’s where we are today. Kantar reported an 85% daily Copilot use rate across your rollout, which is an outlier among all kinds of firms. What does daily usage look like?

What are those 85% doing? I think across the 12,000 people that we have, there’s no answer that’s the same. I mean, people are using it to suit themselves and their job and their productivity. -Yeah.

And sometimes that’s very simple stuff, like just recording meetings, and getting actions, and following meetings that you don’t, you know, we’ve been encouraging people not to go to meetings. You know, meetings are the curse of corporate life. I love Kantar. Hire me!

We just felt we spend so much time in meetings. Let Copilot do the work for you, and it sends you any actions that you might have out of the meeting, or particularly something that you might be tangentially involved in, or going for ten minutes of. Some usages like that all the way through to big heavy lifting agents that we’ve built, we have more agents than people in the organization. That comes comes with good things and bad.

Right. But, we’ve got 15,000 odd agents across the organization today. Some of those agents are just very individual for someone. Some of those, about 150 of them, are big enterprise agents that we’ve built to do big corporate tasks.

So we’re all points in between that thinking. Okay, let’s talk about the 15,000 agents. It sounds like people really, like you said, leaned in, started building all these agents, which then led you to think, “Okay, actually, we have to centralize and organize the agents.” Talk to me about the Kantar agent factory.

The agent factory kind of came about really, as a, you know, we let many flowers bloom across the organization. It was brilliant. There’s a lot of duplication. There’s a lot of repetition.

There’s a lot of very diverse quality in those agents, and many of those agents work brilliantly for an individual, but as you try and scale them to somebody else, they have been written with someone very specific in mind, and not everybody works that way. So what we started doing was spotting the agents that more people were using, and a lot of the activities we’ve had are by word of mouth. It’s not some big corporate program, it’s, you know, I’ve built something, I show it to my colleague, they go, “Oh, that’s really cool.

Can I have that? And you give me access to it?” And then they show someone and it builds, and you start seeing across our ecosystem agents that are being used by multiple people, and then we gather those up and say, “We’ve got the IP right on those, we’ve got the right governance on those, we’ve got the right quality control on those,” and then we kind of package those up and deploy them to the organization because they’re agents that are useful. And then does that, sort of, do you find, encourage the creation of more agents to go into the store?

Does it - factory can be a word that suggests the end of some innovation, maybe. And I think that’s the balance. I mean, what we’ve tried to do is to say, “We’ve got really three types of agents that we build, like no code, low code and pro code,” and how we build the pro code agents that handle some of our really big tasks is fundamentally different to how I might create a no code agent. And you can spin those up in 30 or 40 minutes, and we don’t want to lose the innovation.

What we have learned, as we stood those kind of three work streams up in the agent factory, is agents move between no to low code, or up to pro code, and then equally they move down because sometimes you think, “Oh, this is a really complex build and it’s going to be really difficult.” And actually, when you get into it, you go, “Oh, it’s not. It’s actually much easier to do,” and we’ll push that to the team that maybe doing some of the lower code work. And so this is a journey for us all.

We didn’t have we didn’t have a technology function that was organized that way, you know, six months ago. We’re learning how to do this, and I think that’s one of the things that I was talking to my head of the agent factory, who’s again, not from technology, she was a leader in the business. We’ve probably reinvented this every 2 or 3 months. It’s not a static department.

It’s a very agile department that’s evolving because as the technology evolves from prompting, to agents, to skills, you need to respond differently in those environments, and that’s what we’re learning. You’ve got to be very agile in the way you think about your organization and how you bring your people with you in that journey. How long ago, remind us, did you come to Kantar? I’ve been at Kantar five years.

So like when you got there five years ago, would you have thought to yourself, “This is the organization that is going to be a frontier firm, this is the organization that’s going to implement this, be willing to adapt, roll with the changes, create an entirely new infrastructure internally so that we can adopt the latest and greatest technology and move forward that way?” Never in a million years. Yeah. I mean, I’m certainly not trying to cast aspersions, but you know, when you think the hard thing about companies is to make this mindset shift that you’re describing happening so naturally and organically, it’s like the dream.

It sounds better than the reality on the ground. I’d say that. I mean, like we, you know, every day, I mean, as humans, we like certainty, we like familiarity, we don’t like change. We did a survey recently, and I think there was about 85% of our people said, “There’s too much change going on.

” And we all want change to slow down in the world, but at the same time that we want that, the change is speeding up. And how do you help people through this? And how do you think about what work means for people, and how do you think about people’s willingness to constantly reinvent themself? Because it’s exhausting for people, and that’s one of the things that we find in adoption.

That’s the new frontier of this. You can spin up the ideas more quickly than you can drive the adoption into an organization, and that’s where we are today. We’re trying to work out how do we take adoption to the the same levels as we’ve done with Copilot on the use of agents, which we’re further behind on. Not to ask you to give away any secrets, but can you give us some examples of what are these, what these bigger, more complex agents might be doing?

I think it’s still very abstract for people, and the more concrete we are, the more we realize it almost like unlocks people’s ideas. Well, we’ve created a people agent, which deals with HR queries. Our aim by the end of this year is that 95% of all our HR queries will be dealt with without human intervention. We’re about 40% today.

We were at 0% in February. We think by the end of the summer we’ll be at 70 to 75%. So what does that what does that mean? That’s not one agent.

That’s about 30 agents working together, providing kind of an orchestration layer around them, and that’s very simple queries, like, “I want to book some holiday” to “What’s the policy on a life event? I need to take some time off. I need to get an employment verification letter.” Actually, that’s one of my favorite things.

It’s a no value add process for a company, but people do need employment verification for mortgages, for acting as their guarantor for maybe one of your children going to university, or renting a flat or apartment, that used to take us three days to do. In a service level, and you’d fill in a form, and it would go off to the service center, it’s now 120 seconds, and it’s in your inbox as soon as you’re finished. The nice thing is - yeah, I’m laughing, my CEO had just had to do something like this.

He phoned me the other night and said, ‘I’ve just used that service.” You know, it’s like, “Okay, I’ve cracked the CEOs using self-serve to do employment verification letter for one of his - you know, like, my work here is done! -Right. Yeah, exactly.

You’ve described the goal around agents as fewer, bigger, better, and so as you kind of created this factory and sort of pruned from all of he things that had been created, how do you decide what’s worth investing in? What you might need? I mean, are you now at the point where you can put RFPs around this too? I mean, I think we’re approaching it in the same way as like, if you’re going to put enterprise resource against something, what problem are you solving?

How big a problem is it and how many people need to use that? And, therefore, what’s the value creation opportunity and how difficult is it to solve? And I think the temptation is always to try and solve really big problems, and I certainly found in our organization that, if I go back throughout my career, big problems, big tech, monolithic programs, most non-tech organizations don’t have great track records of delivering big enterprise tech programs on time, on budget, to spec.

They always take a bit longer, they always deliver slightly less promise, and they normally cost you more. What I think this technology has done is democratize how we think about technology. The people agent I’ve just referenced was not built by anybody in tech. It was built by people who have never worked in tech, who worked close to the business problem, and are the service delivery team, and the payroll team, and the people that do the work.

And I think that’s a big shift that we’re seeing with the technology that let people who understand the business problem or the business process reimagine that, freed of the need to be a out of the box, non-configured solution from an enterprise platform. Well, and I love that too, because the people who do the work are the people who understand the problem. Someone has struggled constantly with having to provide the employee verification letter, and having impatient employees who say, “My mortgage is on the line” or this or that, and so they know what problems to solve, and it seems to me that that gets to what you were saying before, which is you don’t want to get stuck in a single department pilot.

It’s that diversity of usage that causes you to be able to solve real problems that add up to, from what I understand, meaningful time saved. That’s right. I mean, we know that across our organization, in the first year of having Copilot, we saved on average two hours per person per week. That’s the equivalent of 600 people for us.

Now, you can’t really get the productivity like that because it’s little bits of everybody’s - you know, two hours of my time and it’s two hours of your time, but it gives you an opportunity to think about what you do with that time, and for some people, that’s doing better work. For some people, that’s taking some pressure out. Maybe they’re kind of capacity already, and you’ve created a bit more capacity for people. So having that flexibility I think is where it’s really interesting, and as we start moving to our dream of the 95% of HR queries being sold, for example, it means that we’re creating new opportunities in the HR function because we go, “Well, we need less of those sorts of skills, we need more of these sorts of skills,” and sometimes that’s about having a smaller team, sometimes that’s about a team doing new things that we can’t do because we’ve been resource constrained, but what’s really interesting for me is the quality of work that people do has gone up, and engagement is going up as well.

I’ll use my talent acquisition team. That team recruits the same number of people as they did last year. They recruit them with half the number of people today, and that team’s engagement has gone up enormously because they’re doing the thing about the job that it really enjoy, which is interviewing people, most of the time. They didn’t like the administration, they didn’t like the non-value added work, so I think you can find balances here where you can really improve the quality of work at the same time as using these new tools and agents.

How are you measuring this? How do you know if the agent factory is working? I mean, clearly you have some metrics about engagement and time saved, and what are the other - I think a big question for people is: How do I know if there’s ROI here? I’m not sure you do.

I think there is a bit of a leap of faith. I work in a private equity owned business. Leaps of faith are not normally where private equity investors like to be. They like data.

So we do have data. We know we’ve got productivity inside the organization, we’ve got engagement metrics inside the organization, we’ve got quality metrics inside the organization, because in any big organization, not everything you do is at the highest quality. There will be pockets at times where you go, “Oh, is that presentation our best work for a client?” And what we’re starting to see is you can really raise the bar on performance but also kind of raise the floor.

So you’ve kind of got a quality threshold where you might be able to get work to a client more quickly. So I think you’ve got to develop your metrics in a sensible way. One of the things that we are seeing is - and grappling with, frankly, is: How do you think about tech cost? How do you think about labor cost?

How do you think about token cost? and how do you triangulate those three in a way that makes sense? Because the challenge, I think, for an enterprise is not to think about the individual budgets, but to think about: Well, what does this cost us overall? And where do you deploy more technology solutions or super agents to help you?

Where do you put people? Where do you make that choice? And up until now, that choice has been forced on you by the availability of technology and the skillsets of the people you have. Now, I think, you have some choices to make about how do you deploy people in the best way?

I mean, it’s so interesting because as we are talking, it seems so obvious that you would combine these functions, that you would be the chief people and agent officer, because I think that there has been a tendency to think about this as a technology solution, as opposed to like a new set of employees, or - we’ve sort of talked for years now about how it will change people’s jobs, but as you’re describing this, it sort of feels like it all comes down to the skills, and the people, and the people managing the agents, and the agents filling in for what people were doing, like it just - at what point did it seem that obvious to you, I guess, is what I wonder?

Not at the beginning. Yeah. I think that’s been our journey, which is - it kind of sounds credible and sensible now, because you go, “Oh, well, that works.” When we started and we were talking about this kind of concept, I remember talking to our CTO, he’s like, “Are we sure about this?

” and we came to the conclusion that the technology works. And if if you start with, you can make the technology work, there’s nothing that we are doing that we are not a tech company, or a user of technology. So this works, and you can learn how to make it better, you can tune it, you can do all sorts of clever things with it . Actually, what stops that, what changes it, is can you drive the adoption and the human mindset?

And that’s a change journey. And that’s a people journey. Most tech organizations would say, “We’ve turned it on. We’ve deployed it in the ecosystem.

It’s there. You’ve got Copilot.” But if you don’t know how to use it, and you don’t know how to experiment with it, and you’re scared of failing, or you’re scared of building an agent, those are the skills that you need to develop. And people are scared of this technology.

They do kind of go - I don’t like getting things wrong. They, you know - they’re scared to fail, And actually, the thing I’ve learned about using agents and building agents is you’re just iterating. You’re not failing. You’re just you haven’t succeeded yet.

Right. And as you develop new ideas and you develop your thinking and run that query 10, 15, 20 times to kind of refine it, the first answer you got isn’t the answer. Oh, that’s a brilliant piece of work. Answer 20 might be, but you’ve got to take some time to use it to make that work.

Right. It’s really a teamwork journey. I mean, I was going to say, just tell yourself you didn’t fail. It failed.

Yeah. Exactly. And you’re just - you’re just raising a toddler here. Yep.

Back to that anxiety piece. When you talk about, from an HR perspective, you’re managing the change, you’re managing people’s anxiety, and then you’re also saying, “I want 95% of this work to be done by agents.” What happens to the rest of the team? So we’ve been very transparent with the team about this objective, and it was hard.

I mean, I remember when I first the first town hall, I talked about this, it was like, “Are you announcing a reduction in force here?” I went, like, “No, no, no, we’re going to try and work this out over the next couple of years, And as a general rule, we’re not going to replace people that leave, and we’re going to see if we can pick up the slack with the use of agents.” It’s largely what we’ve done. We’ve also found that because of the mindset of the people we’ve got in the the people team now, the people team have the highest usage of Copilot in the organization.

We have the highest number of master users, and we categorize a master user that’s using kind of like a power user. About 45% of the HR team are power users of Copilot. They just work this way, and that’s let them be really valuable across the organization, so many of them have left the HR function and gone to other parts of the organization. One of my recruiters said to me, “I’ve been offered another job,” and I was like, “I’m really pleased for you.

,” you know, “it’s sad to see you go.” “But I can’t go.” “They don’t have Copilot.” “I can’t imagine going back to an organization that works in an old fashioned way.

I’m staying here.” So I think there’s a big way that this - when you start creating the environment, and the culture, and the tools that help people be their best version of themselves, they don’t want to go somewhere where they can’t have those tools. Wow. That’s a win.

That must have felt great. It did. Talk to me more about the people who were the evangelists, because you have talked in other interviews about using this almost influencer model, that there were people who you didn’t expect to step up and start creating agents, and really demonstrating this for other people within the organization. How did that evolve?

When we came to this, I probably looked at a fairly traditional change model that says we need some change ambassadors, we need to do kind of - we’re going to do change to the organization. -Yeah. And we started like that, and it didn’t - it wasn’t going very well in the early pilots, and I think we were trying to adopt just the program mentality to get in, unfreeze the organization, introduce the change, refreeze the organization, the job is done. It just doesn’t work.

And if I look at social media, I look at where brands are advertising, they’re increasingly turning to influencers, and they’re trading a little bit of control for the power of influence, and it’s like: Could we try that internally? And so we opened up and said, “Who wants to be a Copilot influencer?” And you could just apply. And we picked some people.

They were not the people I would have picked from a change journey experience, where you’re thinking about development of careers, and all this would be good for their career development. They were a pretty random bunch of colleagues. They’re a great bunch of colleagues, and we learned a lot by not all influencers were equal, and some were brilliant, some we had to help a bit more, sometimes they would post things like the hack of the week, and you were going, “Oh, it’s a bit...

” I’m a control freak at heart, and that made me a little bit uncomfortable, but we found they got a following, they were the people that were in offices showing how to build an agent, how to build and use prompts. They just got into it and because they were from the business, in the business, they’ve just got much more credibility than the change ambassador. Yeah. So that model of: use people across the organization, and it’s been completely unpredictable who those best influencers were.

They’re not all that younger generation of our workforce. There are some of our change influencers are in their 50s and 60s who are deep experts in what they do, and have found a real passion about, “Hey, I can expand my skill set, but I can also, inside the organization, broadcast my capabilities and show you how to be better at your job and give you the value of my experience using an agent, or a prompt, or whatever it might be.” I feel like everyone can benefit from understanding how you did this.

It sounds like you announced it as a fairly formal thing, and then were people just - internally, were there Teams channels specifically for this? Were there workshops that people could attend? How did they influence? We brought the influencers together and we gave them kind of a bit more training on Copilot, because when we did this, Copilot was new.

I mean there were - one thing we found is that people that join our organization today, if they’ve never used Copilot in a work environment, you’ve got to train them. You’ve got to help. We don’t train anyone on how to use Excel, or PowerPoint, or Word because we all just assume any colleague who’s going to turn up will know how to use that. That’s not true with Copilot.

We train everybody every week. Every single week, we do training on a Tuesday about what we’ve learned this week, what’s new, what’s different. That’s now led sometimes by someone from the agent factory, someone who’s built an agent, one of the influencers. So it’s a mixture of very formal, very informal sort of work, but kind of creating the channel so that people are always learning is a really important bit, because the tech is moving so fast that what you might have learned three months ago, actually a new feature has arrived, a new way of thinking has arrived.

So how do you keep the agility? And that’s where influencers working around the company are actually really helpful. Are you bringing in new influencers and rotating some out? Is it sort of - I mean, it sounds quite fluid.

Yeah, and I think that’s the thing. It is fluid, and people join, people leave the organization, new influencers emerge, Last week, in London, we brought 20 of our best agent builders, and we called them citizen agent builders because none of them were technology people, but they were the people that had built the most agents inside our environment. They were our top agent builders, and we brought them together or a two day workshop, and the power of their knowledge, because they were all getting stuck at different parts of the journey.

But collectively, you put those 20 people together in a room and they would go, “Oh no. When you get there, this is how you, in Copilot Studio, do that.” That community we’ve created is a really powerful community, and they are the people that build agents. They are the people that are solving problems for us.

They’re not in the technology function. They’re in the business, they’re in teams, they’re in legal, they’re in finance, they’re in HR, they’re in our research departments. Those people are making a huge difference and creating really interesting skills for themselves in the future. What would you say to leaders who might be in your position or close to it?

You’ve alluded to like: it sounds great now but it was a little - it’s a journey of change, which is always tricky. -Yeah. What would you say? What piece of advice would you give people who want to start this kind of transformation now?

I’d go all in. It’s not going to go backwards, this technology. The technology is moving ever faster. Now, I’m at one level, I look at the last 18 months of our journey and I go, “We’ve learned a lot.

We’ve done a lot.” And then at another level I go, “If I were starting today, how cool would it be?” Because I look at the technology that exists today that didn’t exist 18 months ago, and I look at the kind of way that people can track and monitor and manage agents that didn’t really exist six months ago. That was really tough when you’re trying to manage that governance now, there are products and solutions that let you do that.

I’d be really excited about starting today, and I would go, “I’m pretty sure you’ll be able to accelerate more quickly, but you’ve got to bring your people with you, and you’ve got to kind of focus on the human centered change in this.” The technology is amazing, but don’t forget about the people. Don’t forget about the anxiety that people have. Don’t forget about the huge capability that your colleagues have across the organization.

No one comes to work to go, “I’m waiting to be found out and not do a good job.” Everyone turns up in the office in the morning to try and do good work, and what you’ve got to try and do is create the conditions for letting them to do even better work. Andrew Doyle, Chief People and Agent Officer at Kantar, thanks so much for the time. Thank you, Molly.

Thanks again to Andy Doyle, Chief People and Agent Officer at Kantar. For more conversations from the frontier of AI at work, follow the show and listen to past episodes wherever you get your podcasts. I’m Molly Wood. Thanks for listening.