Managing the Future of Work

Guest Episode: Joseph Fuller on the TechWolf podcast

Managing the Future of Work Podcast Avatar

The Managing the Future of Work co-chair and podcast co-host examines AI's implications for HR, spells out the competitive stakes in early adoption, and unpacks the technology's disruptive potential for jobs, organizations, and markets. Also, the rise of the CHRO, skills-based hiring, and more.

Transcript

Bill Kerr: Welcome to the Managing the Future of Work podcast from Harvard Business School. I’m Bill Kerr. The following is a guest appearance. We occasionally bring you interviews produced elsewhere that highlight research and expertise from the Managing the Future of Work project. My project co-chair, Joe Fuller, recently appeared on the TechWolf podcast. The conversation examines the sweeping multidimensional changes AI is bringing to the nature of work, organizational structures, and business models. It also delves into the benefits of generative AI, as well as the potential for disruption and widening inequality. This extended episode covers a wide range of topics—from what a skills-focused workforce strategy looks like to the increasing relevance of the CHRO, reskilling, upskilling, employee retention, and more. We hope you enjoy the conversation.

Julius Schelstraete: Hello everyone and welcome to another episode of the TechWolf podcast, where we interview HR leaders on everything related to the scale-based organization—what it is, what it isn’t, perceived challenges, real-life use cases, et cetera. And today I’m lucky enough to be interviewing Joseph Fuller, a professor of management practice in general management and entrepreneurship at Harvard Business School. He founded and co-leads the school’s Project on Managing the Future of Work, a project that also has a podcast, and we’ll make sure to cross-reference that in the show notes here. Joe has 30-plus years of experience in business consulting, having co-founded and been first employee and longtime CEO of Monitor Group, now Monitor Deloitte. He has research topics very close to our hearts here at TechWolf, such as the skills gap, degree inflation, the impact of caregiving on workforce participation, the growth of the gig economy, and the impact of AI on the hiring process. So, welcome, Joe, first of all, to Ghent and to the show, too. Glad to have you.

Joe Fuller: Thank you. I’m delighted to be here and to visit TechWolf in person.

Schelstraete: Amazing. And we’re as delighted as you. I want to start this episode the way I start all other episodes, by asking you the question that our previous guest asked. And it’s a question from Ben Wein, who leads the skills initiative at Bristol-Myers Squibb. And his question was: the future of HR and people organizations, given the AI search, how do they need to evolve over the next five to 10 years to be adaptable to that change?

Fuller: AI is as fundamental a technology for this decade as the arrival of alternating current was in the late-19th century. So the first challenge is to acknowledge that the change that AI will unleash will be very significant. It will not be incremental. And therefore, management both in the HR function and as it relates to the management of people will have to be prepared for a level of change, a rate of change, that will make them uncomfortable. I think, though, they are going to find that AI is the source of a solution to many lingering challenges. So while the transition will be trying at times and they may launch experiments that fail, there’s an urgency about adopting it in the human assets domain because the intractable challenges of mobilizing talent, of reducing turnover, of maximizing productivity and job satisfaction—all of those fundamental goals can be significantly advanced by using AI-based tools. And therefore, I urge people to set about it with optimism and courage and acknowledge that there will be a change management challenge, but it will be very, very much worth the effort.

Schelstraete: Okay. So it’s for you a mind-set shift that needs to happen in HR and people organizations with that context in mind being, it’s going to not be incremental, but continuous.

Fuller: I think there’s a very good way to think about it, because historically in HR functions, the correct stance is to be cautious. These are people and their lives and their confidential data we’re dealing with. We have to be careful not to make a mistake. We have to be, we have longitudinal data that we rely on. Disrupting job descriptions, incentives and metrics, allocation of tasks is something that has to be taken gradually. We don’t want to accidentally stop the engine of our company as we’re trying to improve it. So, that caution and risk aversion has been imprinted on the function. But given the type of productivity unlock that I believe AI will create for human resources functions, in some ways the greater risk now will be going slower than your competitor. Because their ability to get and retain talent, to make that talent more productive—in ways that affect the bottom line, not just employee satisfaction—will be very much motivated and driven by how quickly and effectively they learn to use this tool. And the first law of generative AI is: they with the most data that use it fastest win. Because it’s learning, and as it learns, it gets better. And if you’re late and flow to learn, you will fall further and further behind. And it’s hard to catch up with a company that’s learning faster than you are if they have an equal amount of data. If they’re much bigger than you are in terms of their employee base or their market share position or whatever you’re applying to AI to, you can get in a position where you’re never catch up mathematically. It becomes a literal piece of algebra that you can’t get bigger than they are. And therefore, you will never be as smart as they are. And if you’re in a highly competitive talent market, that’s a very dangerous place to be.

Schelstraete: Yeah, so there’s a sense of urgency, should be. And then that feeds again into the mindset shift that needs to happen in HR and people organizations. Okay. I’ll make sure I’ll get that answer to our previous guest. But zooming out, and we already touched upon the topic, like from a distance. But in this episode, in this show, we talk about the skills-based organization. And I’m keen to get from you what is a skills-based organization from you, because there’s lots of definitions floating around, and the definitions are also being adapted as we go long. So what is it for you? And then, secondly, why has it gained so much momentum as of late?

Fuller: The first thing I’d like to say is that skills-based organizations are still more an aspiration than a reality—that the tools to mobilize the organization on a skills basis and to make sound judgments on a skills basis are new, they’re innovative, and they’re only now being understood. A skills-based organization is one that can make decisions about both obtaining and deploying human assets through the lens of what is required of people to do a job effectively. That’s a new way of defining a role, because historically what companies have relied on, for all understandable reasons, is that we have something called a “job description,” we have a history of filling those jobs, from that history, we have identified credentials—¬¬whether it’s academic credentials or experiences that are co-linear with the ability to do that job, and then we design our processes, including the use of earlier generations of AI in things like the applicant tracking system for recruiting processes that allow us to seek people with those experiences or credentials efficiently. I haven’t used the word “skills” once in the last five or six sentences. We rely on proxies for skills, and that reliance is more or less efficient. Often we have failed hires, mismatches because the depth of the experience or the aptitude of the person was off or the cultural fit was off or the ability of the company to articulate what’s actually required to be a success in the role as opposed to what are the proxies that it’s using in its automated judgment of applicants is different. In our research at Harvard, one of the things we find is that for many jobs, the job descriptions are almost never completely rewritten. People just add requirements and very often lead trailing-edge or obsolete requirements, but your natural language processing means you doesn’t know you didn’t really mean a paragraph’s four and five of the job description. It treats it as part of the job description, and all of a sudden that may elevate a candidate who has good credentials than tertiary skills, but weaker credentials in the primary skills you need. So really, the essence of a skills-based organization have to do with maximizing skills mobilization, really tapping into the deep skills in your organization, ensuring that you’re viewing the evolution of your workforce through a forward-looking understanding of what your strategy will require as denominated in skills—not credentials or experience—and then optimizing to the degree possible the deployment of the right skills on the most important task at the right time, because right now most skills in the form of human skills are assigned to work based on the priorities established by the supervisor of that worker. Those persons skills may be better deployed in terms of the company’s interests temporarily to serve another supervisor or another project that isn’t in the immediate control of the supervisor—probably not even known to the immediate supervisor. And we’re moving from—you can think about it in the same way as we think about software development. Skills deployment historically has been a waterfall. There’s a hierarchy. Responsibilities are allocated. People in that hierarchy, they work to optimize the performance of their element of the waterfall. A skills-based organization is agile—and not in terms of the work processes, but in terms of the metaphor that it organizes work the way that we now think about organizing software development through an agile scrum lens as opposed to the historical waterfall approach.

Schelstraete: Like more, instead of only vertical, also horizontal and dynamic.

Fuller: And dynamic.

Schelstraete: Okay. And then the second part of the question, why do you think it’s picking up like it’s doing right now? We’ve been talking about skills for a long time, but it was maybe more referred to as competencies or around other synonyms or alternatives, whatever you might call it. Now it’s picking up pace. What shifts are you seeing, trends are you seeing that really boosts that adoption or at least willingness to adopt a skills-based model?

Fuller: I think there are several. First, a number of companies, a limited number, small percentage, but of thought leaders are beginning to make some genuine progress. So it’s no longer viewed as an aspiration that is far in the future and we know we’d like to be there, but it’s like world peace. It’s a great aspiration, but we’re not counting on it. People are making progress. I think, though, there’s a second reason, which is that the competitive environment in industries has two features. The first is that the rapidity of change and the ferocity of the market—I don’t mean the capital markets; I mean the competitive markets—are such that companies are very, very anxious to mobilize their talent to ensure that they can continue to prosper, win, keep their customers, innovate. And the second consideration is that, strangely, most companies are simultaneously under economic pressure from their shareholders. Their shareholders could be private equity firms, they could be capital, public markets—even governmental entities under increasing pressure by voters to perform differently. Look at the great pressure on national health services, and not just in the U.K. So there’s a desire, an urgency about how do I get to the next level of productivity to satisfy those shareholders or those stakeholders or those investors while the world is changing so fast. I have to unlock the skills—both explicit and latent in my workforce—and get them to higher levels of productivity by deploying them against the task that they’re best suited to execute that are most important to satisfy those constituents. So I think it’s both a capabilities and an understanding. I think HR leaders now have their head around this more and more, but also it’s just an operational necessity given the competitive environment.

Schelstraete: Yeah. And in that whole innovation attempt and it’s picking up, and companies picking up the technology, there’s also a meta story to that, which is the AI also disrupting the way work is done in the first place. And I want to drop a statement to you that we hear a lot, which is the worry of AI replacing people. And I think it’s important because it’s mainly people or organizations facing that worry, either from employees or workers’ councils or various angles. The skills-based movement, part of the promise is that that is not true and that we’re looking to internally mobilize people, reskill, upskill, make sure that they catch up with that new wave? Is that realistic? Are we getting there? Will that be fast enough? And this all summarizes in the question, should we be worried about AI replacing people?

Fuller: This is, of course, the great debate in the topics I’ve studied with colleagues and collaborators and other researchers and scholars in this space. There is no doubt that AI will make a substantial percentage of the workforce more productive. That will, in a number of instances, invariably lead to redundancies, at least as defined by the number of people occupying jobs as currently defined. I’ve been working on a model of the extent of the automation and augmentation in what are called the O*NET Codes, the occupational network codes that the U.S. government tracks. And the average number percentage of productivity improvement across 900 jobs is over 40 percent. Now, the lowest-ranked job is roofers, someone who replaces the roof on your home, and it turns out that AI can’t replace your roof. But even for a roofer, the coefficient is 9 percent because they can order inventory, they can schedule their jobs better, render bills better, perhaps that develop proposals better using AI. When you get to entry-level, white-collar analytical jobs—analysts at a consulting firm, junior financial analyst, or credit analyst—the ratios are 75 percent, 80 percent. Now those people who are more optimistic—and I tend to be on the more optimistic end of the spectrum—will point out several things. When automated teller machines were first created, the prediction is there would be fewer bank tellers by a huge amount—that no one would go to the bank anymore. Most of the big banks in Europe and the United States in the last 10 years have invested in having more in branch personnel, not to count out 20 euro notes, but to help you with client service, try to sell you other services. So the prediction that ATMs would eliminate bank tellers or that automated spreadsheet programs like Excel would eliminate financial analysts, it’s the opposite. They’re now more financial analysts in companies because it’s much more easy to expand the scope of analysis areas that, when it was expensive and cumbersome, companies couldn’t reach. The No. 1 job position in the United States that includes the word “engineer” today is prompt engineer. So there are some jobs that are directly created. There are some jobs that are indirectly created. I think the workers and the enterprises that will be most at risk will be those that are like the HR professionals that were embedded in my answer to your first question, which are those that are slow, that try to artificially break the process, break in the terms of breaking a car, not in breaking a vase or a glass, because they’re worried about the consequence. Their employer is likely to suffer, and then the reason they have to shrink positions will be a function of competitive realities as opposed to the indoctrination, the imposition of AI. I also think that what AI does is remove recurring, usually cognitively oriented tasks—often, by the way, the tasks that people find least satisfying in their role: repetitive analyses you have to do. You’re the sales manager. Every month you have to write a sales report. It takes you three days every month because you have to integrate manually, you’re looking at Excel spreadsheets, the recent data. And you get self-conscious because you don’t want to send your supervisor and her supervisor and their supervisor something that sounds almost identical to the last report, so you start trying to write new opening paragraphs and use different words and engage your good writing skills. AI can upload every sales report you’ve ever done and the data, and it will generate a highly competent first draft in 90 seconds and the seconds are waiting for your display to load. So now my three-day task becomes a two-day editing task. Now imagine that across lots of routine transactions. Think about a job—you might think of an ice flow. AI is heat that begins to melt that ice flow and take those routine cognitive tasks and automate them or augment them to a very great degree. What’s left in the ice flow? What’s left in the job? It’s the social elements, the personal elements, the dealing with things that can’t be automated. It’s leadership. It’s dealing with customer problems. It’s negotiations. It’s the broad category of what we call “social skills.” And even though AI can create avatars that can deal with lots of routine customer service problems, it’s going to be a long time until we have agentic tools that can replicate anything like the nonlinear, heuristic-based logic of someone who’s dealing with a customer who has a complaint or dealing with a sensitive HR topic. How often have we learned, especially in Covid, that for lots of very critical issues, not being in the same room erodes the ability of a decision maker to act. Dealing with AI is the equivalent of dealing remotely with a person. And so there are going to be lots of elements of the jobs and often the most satisfying elements that are untouchable by AI for a long period of time.

Schelstraete: Yeah, super interesting. And you mentioning the bank teller example being disrupted by the ATM, and that’s like the expectation of that and then the reality showing is not that bad. On the contrary, how much pattern recognition do you have when you review this innovation cycle and compare it to the previous innovation cycles we’ve been through. You’ve been in the business space for over 30 years. You’ve seen a lot of innovative initiatives throughout the years. How much do you have the feeling like, okay, we’ve been here before. I’ve seen this before.

Fuller: The closest thing I would say in terms of an analogy would be the launch of something called Mosaic, which is when we created, actually, the famous venture capitalist, Marc Andreessen, was integral in the engineering of the product, a tool that was first launched as a company called Netscape, which could bring visual displays to the internet. Prior to that, it was purely text, alpha-numeric text. The first time I saw Mosaic, I said, “This changes everything.” And it did. The thing that is much more complex about generative AI is that most previous—all the previous technological discontinuities that I have lived through—have been based on essentially a radical rethinking of how to do a fairly well-defined task. But it’s been radical. What makes this fascinating and very, very complicated to understand is generative AI will have a profound impact on how work is done, but also what the customers for that work require. I’ve been working with a large professional services firm, and we built a model of their, what they were called, billable hours. And a conservative estimate was that they could speed up their delivery of what they’re currently have to do by their clients by about 45 percent—so a little a little bit higher than the average in my more general model. But that’s what they currently sell. It’s very difficult to know what their clients will require of them when their clients are using the generative AI determining what they will want to buy from the professional services firm. It’s the same thing when we look at organization structures. The current structure—how many analysts are there versus how many frontline managers versus how many middle managers, versus how many senior managers, how many individual contributors versus supervisors, all those things—they’re all predicated on not only the current definition of the tasks, but embedded in that is a characterization of the type of leverage people up the organization are required to do their job. But every job is affected by AI—from the CEO to the summer intern. And so, this is the first time the technology is so ubiquitous that we know it simultaneously affects what’s demanded and what work needs to be done. And so it’s a very complex systems dynamic model. You can’t really model it. But the feedback loops is what we don’t understand yet. We do understand something in our research at Harvard Business School, which I think is very exciting, which is for a lot of the types of jobs that are significantly affected, the degree of improvement of new workers or workers that have historically been judged as underperforming the average is much greater than we’ve seen in previous technological waves. So in a study done by some colleagues of mine looking at consultancies, top quartile performers got 17 percent more productive using generative AI, but bottom quartile performers got 47 percent more productive. So AI has another effect that a lot of workers should be kind of encouraged by, which can be a great leveler. It can help you with the parts of the job that are hard for you or that don’t come naturally or you dislike or you find boring, as I said earlier.

Schelstraete: Yeah. Does that sum up nicely too, it will change what good looks like? As in, where we might have been impressed five years ago about, hey, this is a cool blog you’ve written or case study. Now we expect it either faster or in better quality because we’re able to have it be written by GPT or other tools out there. What good looks like changes and the bar raises.

Fuller: I think a very broad definition of the word “good”—it will allow more work to get done. That’s “good.” It will allow people to focus on the parts of their job that are most fulfilling, that are most challenging, that engage the uniqueness of humanity. That will be good. It will cause companies to drive better decision making, which will lead to higher customer satisfaction, less waste, more higher levels of achievement and sustainability, more equity, better provision of government services. Those are all good. It’s not without consequence, because in previous technological innovations, there have been societal disruptions. And we know, for example, the people who are likely too be using AI regularly, 40 percent of Americans use it at least once a week now in some way. But they’re more likely to use it in their personal lives than in their work lives, by about 5 to 8 points. But they tend to be younger people who do that. They tend to be knowledge workers, they tend to be people in better-performing school districts in both private and public education. So the benefits will not equally spread and there will be instances where there’s a significant disruption industry and there’ll be some painful transitions for those people. But I believe over the arc of time—and I’m not talking about 50 years, I’m talking about five years—the net for society will be significantly positive. And that will unlock more resources, whether that’s in government efficiencies or in the spending power of companies, to offset some of the specific negative consequences that are felt by either parts of the population or segments of the workforce.

Schelstraete: Yeah. And in all of this, you have the role of a CHRO at an organization. And there in between the effects of AI on their workforce and the tools they now have to be able to manage a new worker sentiment. How does all of that change the role of a CHRO? And we deal with a lot of CHROs ourselves when talking to organization about skills projects or strategic skills projects. How is their role changing? How do they need to adapt, leading this people department of the future?

Fuller: Well, let me reframe that in a way they may surprise you. Let’s go to the customer of the CHRO. And I’m just going to call them the CXO, whether it’s the chief operating officer, someone who runs a division, or the chief marketing officer. The CXO knows in their heart, and ideally also in their mind, she knows or he knows that their function or their business unit, their country organization really is only as competitive over time of what the human asset base of their operation is. For many decades, the CHRO has been an officer that has to run administrative processes well, meet the legal obligations—in terms of everything from benefits and regulatory compliance to tax withholding to issues of equity in the workplace, been really responsible for being a steward of the company’s environment and its compliance. But many CHROs that I’ve dealt with over the last more than 30 years, when you get them in a reflective mood, say, “I just don’t feel as if I’m a true partner, that I’m at the top table in talking about strategy.” CHROs who are listening to us, this is your opportunity. This is the invitation to the top table. This is what’s going to make, you give you the tools to really help those CXO colleagues to get to a completely different level of performance. If you have command of what skills you’ve got in your organization and what skills you need for the future to achieve what those CXO colleagues of yours are going to need—and you understand what’s available in the labor market, and you understand what your competitors are doing, and you understand how to upskill and reskill what you’ve got—because in a generative AI world, retention becomes much more important. Much more important. Because let’s go back to the ice-flow image. The worker that has a deep knowledge of the way the company does things, about the way customers behave, regulators behave, go-to-market partners behave, that isn’t embedded in the data that’s easy for the generative AI to access or have to infer it through looking at data sets longitudinally or something, that becomes more valuable. Because generative AI can do the computation and the inter-data assembly, understanding apparent relationships between independent variables, but they don’t have the context. They don’t have the subtlety of the context. I need to keep that knowledge in the organization. So reskilling and upskilling becomes—and eliminating those parts of jobs that are chronically unsatisfactory or demoralizing firm—all becomes an imperative. So it’s going to move the CHR organization from saying, I’m going to send an HR partner to sit with those CXOs so I can understand how to retrospectively help them achieve their goals to put CHROs on the front foot where they’re much more engaged in anticipating what’s going to be needed because now they finally have command of unbiased and comprehensive skills data. And we’ve never had that before. Yeah. Not close.

Schelstraete: Yeah. No, 100 percent. And on this podcast, we also like to get practical. And that’s why we invite lots of practitioners to share how their organizations are moving skills-based and how that diverges from the Promised Land being depicted in various thought-leader papers. I’m keen to hear from your experience in dealing with organizations currently. Can you share a couple of use cases you’ve seen and were inspired by?

Fuller: Well, my students are often surprised by the answers and examples I use. So let me start with maybe a bit of an exotic example, which is Amazon in the United States thinking hard about how to reduce the turnover of workers in operations like their warehousing. Very high turnover rate. And Amazon also recognized that their capacity, given their business model, to create jobs that would be materially more economically attractive than the current positions they could offer for a high percentage of those warehousing, logistic workers was very limited. The ratio of frontline workers to supervisors in those operations is very big. So the supervisory span of control could be 25, 30. So even if you’re the best worker, your odds, other than you get a promotion to the supervisor role because of just expansion of the business, as opposed to “I will actually succeed my boss,” is quite limited. And then they ask themselves a question, what skills are in demand in this geographic market that are related to hard-to-fill good-paying jobs. They then set about to say: how can we partner with skills providers to have fully reimbursed or compensated training in those skills for our warehouse workers in shift-adjacent time periods so that over a—it might be two years or three years or four years, while they’re employed at Amazon—they’re gaining the qualifications for the more attractive job that’s available here where they live than we can offer.

Schelstraete: That Amazon can offer.

Fuller: Right. So an example would be in Texas, around the area of Dallas, which is a very large aerospace center. They worked with the local airlines who are finding harder to find aircraft repair and maintenance personnel. And they went to a company like American Airlines and agreed on what are their credentials and curriculum and competencies someone needs to have to get hired as an entry-level worker in your aircraft maintenance? Which, at American is a unionized job. American agreed as to what those skills were, TK what the skills provider would be. Amazon says we will deliver training for our current employees who want to sign up for it to earn that certification. And the CHRO of American Airlines stands up and says, “If you complete this program, you will be presumed qualified for hire into this job that pays twice what you earn now. And by the way, it’s a union job.” So you’ve got models of people trying to manage their costs of turnover. Another is to maximize retention. So the opposite business case. You can see that here’s another company that a lot of my students kind of assume if not forward-looking in this area and they’re completely wrong, which is Walmart. Walmart has built an integrated program of helping workers get college degrees, if they had some college credit or entry level, no college credit. But also more importantly, how they can use what they call “academies” to upskill people to qualify for promotion. I think you see talent mobilization in some of the leading pharma companies where they’re using AI to greatly speed up their drug discovery operations, but they’re also using it to make sure that their technical talent in R&D is being deployed against the highest priority projects. It’s manifesting itself. It shows the fundamental nature of the technology, because those three radically different examples. Reduce turnover through keeping people but temporarily knowing you’re preparing them for a job outside the company but understanding the relationship between their current job satisfaction and having someone get them to that next level of earnings power for retention and for maximizing productivity and talent deployment. It shows that this is a fundamental technology if you can do all those things at different places at different times.

Schelstraete: Do you know at those companies if it was a business initiative or an HR initiative?

Fuller: I would say it was HR at Amazon, but that’s just my inference. I would say it was more business at Walmart and also came from very high level at Walmart, because Walmart received a lot of criticism about 15 years ago for their employment practices. And I think much to their credit actually reflected on those and didn’t immediately start being defensive of them and said, “Yes, we do have room for improvement. Yes, we should revisit the way we think about our employee base, especially inside the stores and logistics centers.” I think at the drug companies, it’s very much a factor of understanding that AI is unlocking potential in arguably the most important functions in the company. So it’s a much more strategic decision.

Schelstraete: Yeah, interesting. Your examples are fantastic, and I think it partly answers my next question too, but I still want to ask it. It’s the question around: are we potentially in a bubble here with the skills-based organization idea—meaning, it’s a great theoretical idea, but if you look at companies getting started on it, there’s lots of skepticism around “Is there ROI? Like, where’s the value here? Show me the money.” In your examples, we clearly see a potential or even a realized impact already. But could you respond to the skepticism of companies not seeing the value yet and wondering, “Look, I believe in the skills-based thing, but it’s not really driving impact yet. I don’t see it yet.”

Fuller: I do think that’s a real barrier to adoption. It’s no mistake that the first two examples I gave are in companies that are extremely data-driven. Amazon measures everything. And so they had a sense of where costs were being driven in their operation, which was, there was a very high cost to turnover in their warehouse and logistics operation. So then they ask themselves a question, “What data do we have about why people leave?” And they get that data, and they find out they leave because they can make marginally more somewhere else. And even though that job is just more attractive to that worker because it’s paying a dollar or a dollar and a half more per hour, it isn’t actually structurally more attractive than the job they’re leaving. It’s just temporarily paying more. Then they ask themselves the question, “What if we could create an outlet that led you to a structurally more attractive job?” And hence, they come to their conclusion. I think the barriers are that companies don’t have agreed-upon metrics as to what benefit they’re trying to get. And they’ve been sold over 30 years, many, many, many tools that say, “Oh, your employees will be more satisfied. Oh, they’ll be more motivated. Oh, they’ll be happier.” But it’s hard to track. And when they go to those CXOs I was talking about—particularly the CFO and say that—they say, “Is our turnover employee,” the CFOs says, “Is our turnover of employees equal to less than or higher than our competitors? Oh, it’s equal to. Oh, it’s less than. So why do I care about more happiness? They’re not leaving.” Or they leave at normative rates. I think there’s something more dangerous here, though, and something that I hope your listeners will reflect on. Getting the benefits requires a fundamental rethink of the associated processes. This is not some glorified SaaS capability that you bolt on to the way you currently do recruiting or currently do performance evaluation or currently do skills development. This is so fundamental that it really requires you to get the benefits, to build your process for achieving those goals around it. It’s a transformation. It’s not an augmentation. And that just sounds like a lot of work. And people are in a hurry. They want to get a benefit. HR, certainly in the United States, where most of my research is done, is endlessly and understandably preoccupied by, well, right now, changes in the kind of regulatory environment and the legal environment, but the ever-expanding pressure caused by things like health benefits, the activities of various social groups and advocacy groups inside the workforce and outside the workforce. So let’s take an illustration about skills-based hiring. A colleague of mine did a paper on that, and we found that, while the level of skills-based adoption—the removal of things like degree requirements in the United States—had achieved a very high level, the number of people who appear to have been hired as a function of changing the skills requirements, was really quite modest—about 3.6 percent of all the jobs that have been filled since the removal of those restrictions. It was a complicated model based on the timing of when the restrictions were moved. We had to rely on companies’ public announcements of doing that. But about 100,000 jobs over five years, which is an infinitesimal percentage, 0.7 percent, of all the jobs that were filled during that period. We don’t really know if that’s a good number or a bad number because when, let’s say, a science company removes a corporate policy that says everyone who works here has to have a college degree, when they hire a medical director for an oncology product, you cannot get that job without a medical degree, which means you have a college degree too. So even though that job was technically made available to a noncollege-degree older, it wasn’t ever going to be filled by one. But what we saw is patterns like the companies remove the practice, the rule, but there was no indication that they’d then reengineered the process by which hiring managers evaluate a candidate. So if you’re a hiring manager, you are a college graduate. You are hired into the job you are now hiring for. You become the supervisor. The whole time in that job, not only you but your colleagues all had college degrees. You only saw people with college degrees interviewing them. Now that you’re a supervisor, until the policy was changed, you only hired a college degree holders. Now the policy is removed. No one’s taught me about how am I supposed to think about this? What’s the standard of evaluating somebody on a skill as opposed to a credential? And let’s also be honest. Hiring is intrinsically what I think of as a relative phenomena. It’s not, is someone absolutely qualified for the job. It’s which among the finalists who I believe is the most attractive relative to the others. As a hiring manager, I care about hiring someone who I think will do the best job of any of the candidates in this job. Because that affects my performance evaluation. That affects my productivity, how hard I have to work. Do I really want to go out and be the first person to hire a noncollege graduate, let’s say, for a credit analyst job that’s only been occupied by college graduates previously? And then that person makes a big mistake? And my supervisor is saying to me, what went wrong with this person? And you say, well, they may not have had the right experience. They didn’t have a college degree. You hired a noncollege degree graduate in this job, and now we have this major credit problem? What were you thinking? And no one is offsetting that by saying, here are great stories of success of skills-based hiring, publicizing those stories. No one’s going to them and say, okay, yours is a job we think noncollege degree holders, particularly those with some job experience, can qualify for. Let’s work with you as the HR function. Let’s send you to a workshop and give you some tools and give you some ideas about how are you going to identify the skills you’re looking for and test for those when you’re not relying on the level of the selectivity of the university the person went to or how they performed in university or how much they’re making in their current job or whether the job description sounds equal to or, in their current role, equal to or equivalent to what you’re looking for.

Schelstraete: Yeah. I guess the issue there might be for practitioners that often in an organization, you want to see a win before you want to proceed to changing something as a process. So the question becomes for a practitioner, “What is the minimal amount of progress I need to show to be able to justify a new process to be created?”

Fuller: Yeah, exactly right. And how many times—not podcast hosts like you— but I’ve had a meeting with a CHR senior team or even an executive committee senior team, and really what they end up telling me is, when two or three companies we respect, ideally that we respect that our general field, have figured this all out and are doing great, come back to me and tell me, and then I’ll consider it. Well, you know, the old joke about in hiring is, we saw in this in the U.S. tech sector, for many years their entry-level jobs stipulated three to five years of experience required. They’re entry-level jobs. So maybe they were really saying we can get people with three to five years of experience at lesser companies who will settle for our entry-level job because we’re more prestigious or we pay better. Maybe. But no one wants to be the pioneer. No one wants to take the risk. Everyone is very cautious about this. And this brings me back to my earlier point that generative AI is so powerful and can be so fundamental to raising productivity that that tendency to be cautious and not want to experiment and wait for someone to prove it to me, that’s a risk-seeking, not a risk-avoiding behavior. Because here the greater risk is, you’re left behind. So I admit, big companies are not good experiments. They do pilots. So big companies in HR are going to have to get comfortable designing experiments, learning from them, and making sure everyone involved understands an experiment. And in a pharma lab, 18 of 20 experiments fail. And that’s a good day. And so we have to get more comfortable, that, as we do pilots in AI, of course, if we’re not getting a failure rate of 60 percent or 70 percent, we weren’t being ambitious enough. The more you’re getting close to 100 percent, all you’re doing is using it to improve the efficiency of your current process, which I’m telling you is the wrong approach. That’s the incremental approach. And it’s the person that overhauls the process and bites the bullet on transformation that gets the nonlinear improvement that’s embedded in AI that will lead to a genuine competitive advantage.

Schelstraete: So is it fair to say that the promise is big enough, but that we have to experiment in trying to get there, even though we see 80 percent fail, 20 percent win? And that companies cannot afford looking at others, waiting it out? And only jumping on when they see real success, because by the time that happens, you are so far left behind.

Fuller: Your listeners by now do not expect me to give a short answer. But this time, I will. Yes.

Schelstraete: Okay. Okay, cool. I want to, as we close up, my one with last question, is where do you show up to learn, digitally and physically? And this is to point our listeners to resources they might also tap into to learn about this space—and whether it’s skills, whether it’s, more broadly, the future of work. I think this is especially a question I’d like to ask you, because I’m sure you have to keep up all the time, and you’re creating research yourself. But where do you show up to learn?

Fuller: Several places. There are a limited number of academic resources, I’d like to think including the two large projects at Harvard, that are regularly doing very, very interesting research. Stanford University; our colleagues just to the east of us, to the northeast of us, at MIT; for noncollege-educated workers’ magnificent center at Georgetown University; University of Pennsylvania. There are also individual scholars on certain topics. In some areas, for example, something we haven’t talked about but you mentioned in my bio—I’ve done a lot of what I like to think is catalytic research on how caregiving responsibilities affect workforce participation in career growth, which, by the way, is an increasingly urgent thing for companies to understand for two reasons: Women, essentially now, more or less dominate post-secondary education attainment. Fifty-eight percent of all young adults in college today in the United States are women. They are an absolute majority of master’s degree candidates, they are an absolute majority of PhD candidates—not in every program. They’re 80 percent of romance language candidates and 20 percent of computer science candidates—doctoral candidates in the United States. But across all programs? The majority women. Studying to be a medical doctor, majority women. That’s true in India. That’s true in France. That’s true in Korea. So you have a, it’s a universal phenomenon. Similarly, it’s a universal phenomenon that women outperform men on so-called “social skills.” That part that is very, very secure in the shrinking ice flow I was describing. So it’s really important that companies revisit job descriptions and career paths that were really designed in the post-war logic of traditional two-adult households, where one disproportionately worked inside the home and one worked outside the home. So you will see not-for-profits or social entrepreneurs in area of like that, but they’re also research centers like the Rosalynn Carter Institute, which is a foundation but does very focused research on caregiving. There’s a group called the American Association of Retired Persons, AARP, more commonly known as. They do some excellent research. But the last part of my answer is generative AI. There’s something that Open AI launched called “Deep Research,” not to be confused with Deep Seek. It’s borderline unbelievable. It’s very much linked to scholarly research. But if I ask it a question about a topic I’m interested in, it’s looked at every PDF that’s ever been published that’s related to that topic. If I just tried to read all the papers about skills definition, skills taxonomy, if I started now, with that as my only job, I would not be 10 percent complete before I die. I die soon because I’d jump out a window because it’d be so boring, only reading those papers as my job. Actually, my office at Harvard’s on the ground floor, so I’d have to go up the stairs and then jump.

Schelstraete: Maybe that’s no coincidence. That’s on the ground floor.

Fuller: Yeah, maybe my dean knows that I have these tendencies. But its ability to be a dialogue partner about my research interest, I upload papers to it, and I ask it what would different scholars say? I give them their names. I ask them what of the most important assumptions in this paper. I upload a survey and say, “Are there topics that you feel based on other research I should have included or I’m over-emphasizing?” And it’s like having a very, very precocious young PhD holder work for you, but one that’s read all the papers. So since what I deal with… I think, if I may, that what’s unusual about a lot of my research is, I actually design it starting with what I think the practitioners and decision makers in companies or governments would really benefit from knowing and work back at the data. I don’t start with the data and say, “Well, here’s what it says. Oh, who would care?” That’s what labor economists do, and they do incredibly smart work, and you know it’s incredibly smart because a well-meaning executive or permanent secretary or minister can read it and not know what it says. I’m trying to do things that people know what it says or what it suggests. So I often learn a lot by sitting with companies that serve the space, like TechWolf, with people like the gentleman from BMS, you mentioned, my consulting clients. What are they confused by? What is the real challenge for them? And, of course, the biggest, for most of them, is, “I want to embrace skills first, but I don’t have a mechanism I trust for directing me in how to execute that,” which is where we have a lot of dynamism in the sector to create tools and insight. Still early, still entrepreneurial, but if we can break the back of that, it’ll enable the harnessing of AI, it’ll reduce turnover, it’ll improve the ability of companies to get in governments also—let’s not leave that out—to achieve unprecedented levels of productivity, which will help buffer society from the negative impacts of having this fundamental tool open to companies and learners alike.

Schelstraete: Yeah, fantastic. And then my last question is one to connect to the next episode. Oh, okay. And this is where you get free rein to ask a question to whoever will appear next on this show after you. And this can be about the main topic we’ve discussed—skills-based organizations—or it can be about something else. But know that we typically interview practitioners that are at organizations or thought leaders that are exploring this skills-based space and AI-driven HR. So what question would you like to ask our next guest?

Fuller: Well, this might be a bit unfair depending on the nationality of your next guest. But without editorializing, there are a number of at work in the world, largely—not exclusively—emanating from the United States, that are forcing HR functions to revisit the way they think about qualifications, goals they have in the workforce on issues like diversity, but also seem to be likely to throw up significant barriers to globalization and begin to distort the model of things like “best shoring” for skills. So how are HR functions supposed to respond to that? What are the challenges they have to rethink? What are the fundamental design principles of the way they’ve run their operations that these changes should cause them to be revisiting now. Because, as I said several times, in the world of AI, they that move earlier more effectively gain a big advantage. And I don’t think the types of changes that are being now catalyzed by things like the current administration in Washington actually are exclusively a function of the eccentricities of our president. I think things like nationalism and resentment of global product flows and resentment of other workers or job seekers—whether they’re immigrants or others in my home market—is highly visible in the entire developed world in the entire OECD. So we can say, oh, this is all the U.S. administration and they’ll go away and this will all revert to normal. I don’t think so.

Schelstraete: Yeah. So if we could repeat that in a question to the next guest, how would you sum it up?

Fuller: How do you think a flowing or reversal of globalization and forces that begin to constrain labor-market participation or labor-force growth in your markets should be responded to by state-of-the-art HR functions?

Schelstraete: Fantastic. Let’s see how they roll with that.

Fuller: I can’t wait to hear.

Schelstraete: Yeah, I’ll definitely get the answer to you. And the only thing left for me to say is thank you very much, Joe, for taking the time here after a busy day. And I hope we do this again.

Fuller: Well, I would look forward to that, and it’s been a great pleasure.

Kerr: We hope you enjoy the Managing the Future of Work podcast. If you haven’t already, please subscribe and rate the show wherever you get your podcasts. You can find out more about the Managing the Future of Work Project at our website hbs.edu/managingthefutureofwork. While you’re there, sign up for our newsletter.