Inside most organizations, the pressure has shifted from proving that AI works to showing what it is worth in cost, productivity and headcount. Some workforce plans are now being decided on that basis before anyone has looked closely at how the work is done. The organizations getting the most from AI are treating it as a question of work design rather than technology.
So what does designing work around AI involve, and what does it cost the organizations that skip that step?
My guest this week is Matt Campbell, Managing Director at Alvarez & Marsal, where he leads the Talent, Organization & People practice and advises organizations on workforce strategy and organization design. In our conversation, he explains why most AI pilots produce an inconclusive result, a five-stage model of AI’s impact on work, the ways organizations misread their own people, and why cutting too far, too fast creates a talent debt.
In the interview, we discuss:
- Why most AI pilots are designed to prove the wrong thing
- Who owns AI?
- Why big tech’s approach to AI and headcount is the wrong template for everyone else
- The difference between a job and a role
- The five A’s: avoid, assist, augment, automate and autonomous
- How to put numbers against cost, productivity and capacity
- The four things organizations most often get wrong
- Why unauthorized AI use gives employees the benefit and the organization the risk
- Talent debt and the cost of cutting too far, too fast
- Are you designing work around AI, and what does the future look like?
Follow this podcast on Apple Podcasts.
Follow this podcast on Spotify.
Key takeaways
- Matt Campbell argues that most AI pilots end inconclusively because they are set up to test whether the technology works rather than whether it solves a business problem.
- A job is a bundle of tasks, but people also fill roles within an organization’s social system, and that social system is the part of work AI cannot see.
- The five A’s model used at Alvarez & Marsal assesses AI’s impact on work in stages, starting with the work an organization can avoid altogether and moving through assist, augment, automate and autonomous.
- Employees who bring their own AI tools to work capture the individual benefit while the organization carries the risk without the commercial return, so Matt Campbell’s advice is to engage and scale that use rather than police it.
- Cutting too far, too fast in pursuit of AI savings creates a talent debt, and some of the big tech companies that cut early are now slowly rebuilding the capability they lost.
- Matt Campbell expects the question to shift from what organizations should do with AI to whether they are designing work around it, and the organizations that pull ahead will be the ones that redesign the work, the roles and the social fabric together.
Transcript
Matt Alder 0:00
When companies start talking about AI, conversations often focus on headcount, as if a job was just a list of tasks. Jobs are roles that are filled by people, and roles are all about how they connect to each other and to the organisation, which is precisely the part that AI can’t see. So how do you redesign work around AI without breaking the social system that holds the organisation together? Keep listening to find out.
Advert 0:31
Support for this podcast is provided by Proof, the identity platform built for hiring. You’ve probably heard that by 2028, one in four job applicants is predicted to be fake. In fact, you’ve probably already seen that in your recruiting pipeline. Today, scammers are sending one person to get the offer and a totally different person to start the job. But most identity tools either only screen an application or verify a candidate’s ID. Proof does both. Integrated directly with applicant tracking systems like Lever from the first interview through to onboarding. So you have proof that the person you interview is the same real person you hire. You can learn more by going to proof.com. That’s P-R-O-O-F dot com.
Matt Alder 1:41
Welcome to episode 824 of Recruiting Future with me, Matt Alder. Inside most organisations, the pressure has shifted from proving that AI works to showing what it’s worth in cost, productivity and headcount. Some workforce plans are now being decided on that basis before anyone has looked closely at how the work is actually done. The organisations getting the most from AI are treating it as a question of work design rather than technology. So what does designing work around AI involve and what does it cost the organisations that skip that step? My guest this week is Matt Campbell, Managing Director at Alvarez & Marsal, where he leads the talent organisation and people practice and advises organisations on workforce strategy and organisation design. In our conversation, he explains why most AI pilots produce an inconclusive result, a five-stage model for AI’s impact on work, the way organisations misread their own people when it comes to AI, and why cutting too far too fast creates a talent debt. Hi, Matt, and welcome to the podcast.
Matt Campbell 2:56
Hey, Matt, good to be with you.
Matt Alder 2:58
Pleasure to be talking to you. Please, could you introduce yourself and tell everyone what you do?
Matt Campbell 3:03
Yeah, I’m happy to be with you today, Matt. My name is Matt Campbell. I lead our talent organisation and people solutions at Alvarez & Marsal, which sits within our corporate performance improvement team. For people who don’t know Alvarez & Marsal, we’re a firm that’s founded on the principles of leadership, action, results. So we’re usually working with our clients when they’re under maximum pressure to make things work. And it’s usually happening at speed. And my team within that is very focused in terms of how do we bring that discipline to our clients and how they address performance in their business, organise their workforce, how work gets done, the roles, and increasingly, how they integrate AI into the work that’s happening at the moment. So it’s an exciting time. I think one of the early things I see is how do we get those conversations happening in the same room? Because the workforce conversation is not always happening in the same room as the AI conversation.
Matt Alder 3:57
Yeah, I think that’s a really interesting point. And it’s obviously one of the biggest catalysts of change at the moment. I suppose the other issue is that there is a huge amount of hype and speculation and marketing noise and all kinds of things going on around the AI conversation. What are you actually seeing? So beyond all that hype, what are you seeing happening inside organisations at the moment, the clients that you deal with when it comes to AI?
Matt Campbell 4:21
Yeah, I think we are starting to see people become more discerning about the hype. We’ve moved from leaders who had experienced a free version of a chat AI tool 18 months ago and thought everything’s a hallucination and doesn’t work. The technology has moved on dramatically since then and people are becoming more mature in terms of how they’re starting to think about the problem or the disruption that it can create for their business, both positive and also negative in terms of how they might respond or not respond to it. I think one of the first things we’re seeing with clients is this idea of pilots. Pilots for the most part are usually not designed well in the world of AI because they’re looking to prove that the technology either does or doesn’t work rather than does it actually solve a business problem. So when we’re working with clients we lean very much towards what’s the business problem we can solve here using AI as a pilot rather than treating it as a technology in and of itself, and that gets you to an inconclusive result because it didn’t actually create anything meaningful. The other thing that sits underneath that is most organisations are spending a lot of time still trying to work out who owns AI as if it could be owned. Is it IT? Is it a centre of excellence? Is it up to every business unit? And that debate can run for a lot of time rather than people actually getting clear around that this is something that cuts across all of us and how do we do something. So I think that’s probably the big things that we’re seeing, the failure modes that come out of that. At one end, you’ve got that arm’s length view of shallow commitment, inconsequential scope that leads to nothing. And at the other end, we’ve got teams who are over-engineering trying to get the perfect answer, which is hard because the technology is moving so quickly, which means by the time you come up with the perfect answer, it’s probably irrelevant for the generation of technology that’s actually shipped now.
Matt Alder 6:25
Yeah. And I think that’s such a big challenge because coming back to what you were saying there about the models that people have used, the capabilities have advanced so much, even in the last five weeks. I mean, how do people keep up with that?
Matt Campbell 6:35
I think there is a lot of scenario planning around where you think the technology might go. If we just step back and look at the large language models, which is what most people associate with the term AI today, they’re moving to a world of generalised intelligence. So that’s their objectives to get there. So AI will be in everything and theoretically can do anything where there’s data to draw from. So that’s where the technology ultimately is going. I think for organisations to be able to keep up with that, they need to understand what the conversations are, understand that all AI is not just one big lump of large language models. There’s lots of different technologies that get wrapped into the conversation around AI. So getting educated as to what those different types of technologies are is probably the first step in getting involved in terms of what is progressing and what are the implications of how that might impact us.
Matt Alder 7:36
One of the things that’s been happening a lot is when companies are talking about AI, the conversation seems to jump straight to jobs and headcount or headcount reductions. And I think a lot of that has been driven by some of the things that the big tech companies have done over the last year or so. Is that the right starting point or should companies be looking at something else first when they’re looking at sort of AI efficiencies?
Matt Campbell 8:01
You’ve hit a really important point there, Matt, in that the big tech companies have driven a lot of the perception as to what this means for everybody else. The big tech companies have different business models and different ways of getting things done than most organisations in the marketplace. And so it creates this sense of existential crisis, right? Because people are looking for answers. They’re worried about what might happen. And so you end up with this, between social media, folklore, the hype cycle, people jump to, well, how many jobs is this? And what does that mean? But if you actually slow down and start to look at what actually sits underneath, how work happens in an organisation and where they might start to look first, a job, and we do a lot of org design, so I say this with a degree of humility but also confidence, a job is really a bundle of tasks, and that’s the work system. The other part that often gets neglected is people also fill roles and that’s a part that someone plays and that’s actually usually more the social system in an organisation and how people connect to each other, how they matter to the organisation, how they matter to each other. And that is the piece that AI misses because it’s a rational system that can target tasks and can lean in very quickly in terms of what is the work that should be impacted by AI, but it doesn’t sort of address that fear cycle that comes out around headcount and what that looks like. So we’ve got to recognise that a social system is getting disrupted here. There’s plenty of work that sits in organisations and we call it work and create tasks out of it. But it’s really busy work because it’s compensating for something in the social system of an organisation that’s not being addressed. If we look at recent learnings, the pandemic was a pretty good example of that because everybody was concerned that productivity was dropping and that they were looking for why are we getting tasks done? Are we not getting tasks done? And leaders were genuinely confused — if we’re completing the tasks, but something doesn’t feel right. And that’s because the whole social rhythm of the organisation was being disrupted. So it’s kind of a proof point for how AI is now changing, again, what that social system is and what those new rhythms are, how people connect, and what that looks like.
Matt Alder 10:10
And I suppose to dive into this a little bit deeper, when we’re talking about kind of looking at someone’s job or looking at the tasks that they do, I think there’s always kind of a move to talk about AI being automation and that they’re the same thing. But actually, it’s far more complex than that, isn’t it? It’s not that the work disappears or even gets automated. Sometimes it changes. We talk a lot on the podcast about the hiring process actually changing rather than just being automated as it is. I mean, how do you go about this? How do you break down the different ways that AI might actually affect a role within the organisation?
Matt Campbell 10:50
Yeah, I think if it was as simple as automation, we would have solved that a decade ago, right? When we had the big hype around RPA or robotic process automation. And what we’ve tried to do is take some lessons from that in terms of what goes beyond automation. I think one thing that’s unique at A&M is we usually start with what’s the work that doesn’t need to happen. And so we’ve built this model that we call the five A’s in terms of how we think about the impact that AI may have on tasks in an organisation. And the first place that we start, which is kind of counterintuitive, is what’s the work we can avoid doing in the future because it’s no longer needed. So back to some of that, what’s the work that actually matters in the organisation to be completed, what can we actually avoid so we’re not just pushing it into technology but we’re taking a critical look as to whether the work matters in a future state. Then we move to the idea of automation, which is really where you’re doing the work faster. But as we move up that, I guess, the cycle of A’s, there is the avoiding the work to start with. There’s assisting. And so where AI is acting as a co-pilot, which is really taking friction out of the work. It may not have a lot of financial benefit, but it may make the employee life easier. It may make leadership decisions faster. It may make customer experience a little easier because we’ve taken friction out of things. After that is where you then start to get to some of the commercial impact because that’s where you’re looking at the augmentation of work and an individual now having the ability to do, you know, three to five times more work as they have the technology driving more of the work for them. After that is where you actually get to automation, and so automation is where, for us, where AI is actually running routine work end to end, humans handling the exceptions. You get the human in the loop, which is a phrase that gets used a lot, to make sure that the routine work is covered and working well. And then autonomous is the fifth A there, which is a newer space for most people to even wrap their heads around, where AI is actually operating all of the work and the humans governing the process, the outcomes, to make sure that the right sensory information is going in so that the AI can generate that. So that’s kind of how we’ve been looking at it. But when you get back to your point around a leadership team saying we’re automating, it’s often a language barrier because that automating and AI really means you need to step back and understand what are the different technologies at play? What does that mean? And in terms of how we’re going to apply that to work in the organisation, then how does it actually change jobs and roles in an organisation? And it gets served up as a technology problem, but it’s more often than not a vocabulary problem as much as it is a human problem and thinking through how that comes together.
Matt Alder 13:46
Again, as you said earlier, we’re talking about, you know, pilots and we’re very much in this age of experimentation. I know that lots of people listening are, you know, experimenting with how AI can do the things that you talked about there within their team, within their organisation. How do we translate this into actual business value that senior executives can buy into? I mean, what does it mean in terms of cost, productivity, resources? How do you put actual numbers against it?
Matt Campbell 14:14
Yeah. So before we jump into the mechanics of that, I think there’s a really important piece in the background here that’s probably not getting enough attention and AI is driving a lot of it. And that is the churn in C-suite that’s happening. So last year, across the major indices, there were 234 CEOs of major corporates left their roles, which is a 16% increase from the year before, right? And that was an increase, the prior year was an increase as well. So CEO tenure is falling, and it’s gone from eight years a couple of years ago down to seven years, and we’re expecting it to be in the six-year range this year based on the increasing number of turnover that’s happening. So you’ve got a lot of movement at the top of an organisation because boards are expecting more, and we’re looking at the implications then of how do you actually even make decisions. So before we get to the cost of productivity, how many people are putting numbers against it, what is the expectation of these new leaders coming in and helping those new leaders get into role quickly, get the team aligned around that, and then get into the mechanics of actually looking at the numbers that everyone else is starting with rather than thinking about what’s happening with the churn. And that’s something that’s kind of underreported. And for some of your listeners, there’s a 60% chance that when a CEO moves, that they’re going to bring a new HR leader in as well. So there’s lots of opportunity, but then there’s also lots of movement there as well, right? So getting into the mechanics then of actually applying the costs, the productivity, the capacity. For us, we are using some of our different partners out there. We’ve got a labour analytics partner that we pull in a lot of data in terms of what’s the cost of different skills in the market. How is that changing? What does that look like in different geographies? So we can have good detailed market intelligence. We’re bumping that up against the client’s technology roadmap and what they’re actually planning for AI deployment, putting some contingency around that because the grand plans often don’t stick to plan, and we’re then creating some scenarios that come out of that. So using all of that and mapping that against where is work done in the organisation, who’s doing that work, how do we value that work, how much work is going to be disrupted as a result of these different AI technologies, applying those five A’s to then look at what’s the work that’s actually left, and then starting to architect what are the roles that come out of the work that is left and how does that change the patterns in terms of how the organisation works? So a lot of that is around the task system, but then we’re also looking at how do you pull in that leadership role? And that takes me back to the CEO churn of the leaders really need to set the tone for how they’re going to take up their own role and the roles that they’re expecting for everyone else. So there is a perfect storm of not just the technology disrupting the work and the mechanics then of looking at what the implications are, but then also how do we understand what the leadership implication is for bringing the workforce along.
Matt Alder 17:20
And this really is the biggest change in business that we’ve seen for a very long time, if not ever, in our careers. And even with a clear plan, clear strategy, change is always difficult to pull off in practice. What are organisations getting wrong most often about this? And also, where do the humans fit in? Where does human judgement fit into everything?
Matt Campbell 17:45
Yeah, maybe I’d start with a position of, like, our firm does a lot of cost takeout programmes with organisations. It’s the heritage of the firm that it was based on. What we’ve seen happen there is spreadsheet-type answers aren’t enough, and leaders in those organisations who’ve grown up in the experience of it being largely spreadsheet exercises have now got a higher expectation and standard in terms of not just give me the answer, but how are we actually going to move the organisation through that? So that’s my opening frame here. And so thinking about the question you’ve laid there, the pieces in terms of where I see things happening relatively consistently, three of them are about misreading people. The first one is assuming that employees want AI and it’s going to be magic. The second is assuming that employees don’t and that they’re scared of it and that they’re going to resist it and they’re going to see it as a replacement and not see it as a productivity tool. And so that’s the first two in terms of misreading people. The third piece within that is treating it like any other technology implementation. There’s research out there that 80% of AI projects fail, which is kind of like twice the rate of IT projects failing because we’re treating it as technology. We’re not treating it as an organisation transformation. And so we get hung up on implementing technology rather than looking at what’s actually going to change as a result of this and what are the implications around it. The fourth point outside those three that I’d make is that AI is already in most organisations, whether leadership’s authorised it or not, right? I saw a stat the other day that 80% of employees in white-collar roles are using AI that their IT team has never signed off on or approved and that they’re just using it, right? And what happens there is people are, as individuals, getting the benefit out of it. So going back to the five A’s, that’s the assist level. It’s taking friction out of their day, makes them feel a little happier around it. But what isn’t actually happening for the organisation is getting the commercial benefit out of it because that kicks in when we’re operating at that augmentation level. So what the organisation then ends up with is all the risk of all of that data happening in those unsanctioned systems. And at the same time, not actually getting commercial benefit out of it. So the idea isn’t to police it but start to think about what’s a systemic path to how we engage those employees who are bringing both tools and technologies in and then looking at how you actually scale what they’re doing so you can move it to augmentation. So that’s probably the four big things. Again, if I can sort of carry the theme through here, is that the bigger underlying pattern is people thinking about redesigning work solely as a task system without identifying what are the patterns, and what we call the management operating system. So what are the rhythms, decision rights, escalation paths, feedback loops — how does that change as a result of AI? Because the other thing that AI is doing is making all of this happen very fast and so speed becomes a factor. And so if outsourcing was your mess for less and robotic process automation was let’s make the mess run faster, this now sets us up for a construct that we can change here, right, because we’re at the start of this revolution, which is not just doing the same things but actually thinking about why we’re doing them and how do we move from a recipe of confusion to actually thinking about new roles above and beyond just how tasks are being impacted.
Matt Alder 21:30
Totally. I couldn’t agree with that more and it’s certainly something that I’m seeing the more progressive employers really looking at when it comes to hiring and talent acquisition and HR in general. As a final question, where are we heading with all of this? How do you think things are going to develop over the next two to three years?
Matt Campbell 21:48
Well, two to three years is a very long time in the world of AI.
Matt Alder 21:52
Especially as I was talking about five weeks earlier. So yes.
Matt Campbell 21:55
Yeah. If we think about the revolution that Anthropic has been part of over the last 12 months, that was unimaginable for most of us two years ago, right? So the technology is moving in very fast paces and very fast moments. I think the piece that I would lean in over the next two to three years and how to make sense of it is working out a cost case for a reduction in headcount is pretty straightforward in this world. And the mechanics of it, if you’ve got the right tools and the right partners, you can think through those things very efficiently, very quickly. What you have to be mindful of within that though is the lessons that we can learn from RPA and ERP implementations that many of us have survived is the term technological debt that came out of that. We should be considering now is what’s the talent debt that we’re creating here? So if you cut too far, too fast, what does that mean? Because there are some of those big tech companies that you talked about before, they’re now going through a very slow process to rebuild capability back into their organisations. And from an engineering perspective, it might make sense to cut early and cut fast and take it all out, but humans don’t always respond as well as engineers would like them to. So sort of thinking about what that is in terms of for your own organisation, what your talent risk profile comes out of and trying to be pragmatic about it and critically testing what’s going to go wrong with your plans, working on the basis of scenarios and not just a single big bet. And then looking at how do you create enough flexibility and strong engagement within the organisation so that it’s a journey that everybody’s on. We haven’t really talked about adoption of AI in roles. And one of the pieces there is we really see like this bimodal type thing of people leaning in or people completely resisting and almost nobody in the middle, which is unusual for a technology adoption. And so working out what that means in your organisation and is that alignment carrying through in terms of your skills profile that you need in the future? What do those scenarios then look like in terms of how the work is going to change? Who are the people that need to be in those roles? And then what are the skills that you need to have access to to actually drive that productivity? So, again, going to your bigger question here of what happens over the next two to three years, I think the safe thing to say is the question is going to move from what should we do with AI to are we designing work around AI? And the organisations who pull ahead may not be the ones who’ve put it in first, but they’ll be the ones who’ve actually redesigned the work, redesigned the roles, redesigned that social fabric of the organisation so that the whole thing moves together.
Matt Alder 24:42
Matt, thank you very much for talking to me.
Matt Campbell 24:44
Great to be with you. Thanks, Matt.
Matt Alder 24:49
My thanks to Matt. You can follow this podcast on Apple Podcasts, on Spotify, or wherever you listen to your podcasts. You can search all the past episodes at recruitingfuture.com. On that site you can also subscribe to our weekly newsletter, Recruiting Future Feast, and get the inside track on everything that’s coming up on the show. Thanks very much for listening. I’ll be back next time and I hope you’ll join me.
