Joe Fuller: Given the unprecedented speed and scope of advances in AI, it’s little wonder consulting firms are seeing an uptick in demand for guidance. At the same time, the technology is upending their business model, from staffing to billing. As they experiment internally, professional services companies are leveraging the results in an increasing array of specialized AI offerings. But as business adoption lags and early implementations of generative AI underwhelm, consultants face mounting pressure to deliver.

Welcome to the Managing the Future of Work podcast from Harvard Business School. I’m your host, Harvard Business School professor and nonresident senior fellow at the American Enterprise Institute, Joe Fuller. I’m pleased to welcome to the podcast Dan Diasio, AI lead at the consulting firm EY. Dan oversees EY’s AI service teams and is increasingly involved in redesigning the business around AI. We’ll discuss the state of play in corporate adoption, workforce strategies, and training. We’ll consider the importance of framing—whether AI is viewed in terms of strict cost-cutting or as a tool to drive growth through improved customer experience and innovation. We’ll also look at the evolving skills picture as knowledge workers increasingly interact with agentic AI and a new AI-native, generation of professionals enters the workplace. And we’ll talk about what it takes to make business sense of a rapidly unfolding technological revolution. Well, Dan, welcome to the Managing the Future of Work podcast.

Dan Diasio: Joe, thank you. I’m excited to be here.

Fuller: Dan, you’re leading EY’s AI efforts in the U.S. How do you find yourself in that role?

Diasio: A couple of years ago before the advent of generative AI, I was really focused on leading teams that were focused on doing AI and helping our clients adopt AI inside their organizations. But with the advent of generative AI, now I’ve gotten another part of that job, which is to really look at how we reinvent the way we deliver professional services to our clients using AI as well. So it’s now a plus-one role, but a very fun and exciting job.

Fuller: Well, then let’s unpack that a little bit. First of all, what sort of work were you doing historically? And then how has that changed in terms of the services you’re providing clients now that you’re working with generative AI?

Diasio: Yeah, so historically we’ve been supporting many large multinational organizations in helping to set a strategy around where and how they can use artificial intelligence, designing what the workflows would look like when they do use AI, developing some of those tools and then in some cases refreshing and maintaining them so they continue to stay up-to-date and current and keep learning. And when I say “artificial intelligence” in that context, it’s the broad term. It looks at machine learning, where we help our clients forecast what they might need from a product demand perspective or from a financial perspective. It involves helping them with classification-type problems where they might want to identify where potential bad debts may lie, versus those that are good or customers that are really high-potential, versus those that they maybe would spend more time and more money trying to influence and generally looking at a whole host of other types of systems. As of late, that works still exists. But you mentioned how things have shifted with generative AI and agentic AI. We had the opportunity to help our clients navigate through some of that change, but in a lot of ways, if you make a list of all the industries that are going to be impacted by AI, generative AI, professional services was near the top of that list. So it was really on us to do something we refer to as “client zero,” to adopt that technology ourselves to try a lot of different options, a lot of different techniques, and then to take a lot of those lessons to our clients, because we were putting a lot of money, energy, and executive focus behind it.

Fuller: So then that’s kind of an interesting brief you’ve got, because it sounds like you’ve got two client groups, really classic consulting clients: EY being one of the genuinely large and impressive professional services firms in the world, and also an internal client with this client-zero-type internal world. Tell me a little bit about how that’s changed the way you think about workflows in the firm itself to serve clients. How has it affected staffing? What have you done in terms of bringing your staff up to the speed that they can competently deliver agentic or generative AI-based solutions to clients that need help?

Diasio: Let’s take this back to February 2023. We went on a journey first to teach our people: what is generative AI and how to use it appropriately. Second, we taught them and we enabled them with a whole bunch of tools so they could use this and use these tools safely and protect our clients’ data, our internal data, our intellectual property. Then we went through a journey to teach them how to use this for their specific roles. So if you’re a supply chain consultant, how should you use it in your job? If you’re a risk consultant, how should you use it in your job? Now, more and more, what we’re getting into the space of is really around how they bring the whole of their expertise into using these systems so we can get more reliable and better, more consistent outputs as we do our work. And we call that really building out a discipline of context engineering.

Fuller: How would you care to describe the state of play now in terms of your client’s understanding of the technology? In the last six plus months, we’ve suddenly had quite a number of prominent business executives making some pretty bold statements about what level of job displacement there’s going to be and how work requirements are going to change and how employees are really going to have to be vigilant about making sure their skills are state-of-the-art so that they can continue to grow with the company. When you actually see what’s going on in large companies, where are they? What’s their mindset? Are they experimenting and embracing this? Are they being cautious? How would you characterize them?

Diasio: Largely and generally, what we’ve seen is that there’s been 30 months of experimentation that’s happening, where companies have been setting up centers of excellence and working through a variety of different use cases. Typically, those use cases are in spaces like in the support desk space or in contact centers or in areas where there’s a high degree of documentation and deriving of insights from that documentation. So in the legal space, contracting space, and then also in the programming space, in the development space, as there’s been much talked about how systems are really augmenting programmers in the future. But I would largely characterize that as a lot of experimentation. And what I’m starting to see is that many investors and CFOs are starting to ask, “When is that turning into results that show up on my income statement and balance sheet?” And I think what many companies are finding is that what got us to this point will not get us to delivering what that P&L value is. They’re starting to look at different motions. Instead of setting up centers of excellence, maybe I set up AI-native business units. Instead of working on use cases, maybe I need to start reimagining new business processes and really streamlining the way work happens, because adding AI as another step to somebody’s job doesn’t actually create a lot of value. It’s much like the old RPA [robotic process automation] days. And so there’s been a big shift in the way organizations are thinking about how they capture value. And there’s also one thing they are in the process of navigating is that, because there’s so much experimentation, we’ve recently done some studies by asking executives how they’re doing with their AI programs. And what we heard is that 50 percent of those organizations feel that enterprise-wide enthusiasm is down over the course of the last 12 months. So there’s a bit of AI fatigue that has started to occur inside many of these organizations, where they’re starting to close down the experimentation phase and try to find some new motions to be able to create value.

Fuller: That’s really interesting, and it does fit with what we’re observing in our Managing the Future of Work project here at Harvard—that companies have started off trying to use AI and even generative AI as if it were a rather standard new SaaS application to improve their current efficiency in performing a process. But that was a bit disappointing. A lot of them had challenges with the quality of their data and their ability to have “clean data,” we call it, to let the algorithms do productive things. And that’s led to some disappointment, and, as you’ve mentioned, some fatigue. Are there examples now of people who are really getting beyond that in your estimation and beginning to think more in terms of how do I build processes around AI as opposed to appended to existing processes?

Diasio: As we sit here at the end of the summer in September, I am getting more RFPs from clients focused on, not helping to discover ideas of where to apply AI across their value chain, but to figure out how we can start to really reinvent the way work happens. And to that end, there’s a methodology that we’ve started to build out that involves two steps. The first step is that we figure out how we control the inputs and the outputs and change the way work happens. And we have a lot of examples where we’ve applied this to our business, where we’ve seen we can take something that takes 45 hours and now do it in 15 or 16 minutes. That creates a really, really big, unprecedented scale and a really big unlock. The second step is to say—now that I can do that in 15 minutes—“How do I challenge my inputs and outputs to now do something that was never before possible?” And that might be in a scenario, instead of pulling samples, now we are looking at everything in a real-time basis. Instead of segmenting a population of our customers that we want to go after, now I can get really personalized with all of those customers. So there is this really big opportunity as you reimagine business processes and as you start to reinvent ways of working that we shouldn’t necessarily just be focusing on what we do today and we can really challenge some of those goals to do things that were never before possible.

Fuller: Do you anticipate that’s going to affect workforce talent strategies and companies? How do you see that, as this matures and gets—that philosophy you just espoused—starts getting embedded in larger clients? Where does the firm come out in terms of thinking, “How is this going to affect the size of workforces? How is this going to affect productivity?” As you know, there’s a big debate going on now. Is this going to be the ultimate abundance technology that really unlocks productivity and allows companies to do much more customization, be much more efficient, have better sustainability outcomes, all sorts of things, and there’s another that this is going to be a bit of a culling of the herd than a lot of companies are just going to need fewer employees. Any sense of how that’s going to shake out?

Diasio: I think the knowledge worker is being underappreciated at this moment in time. A lot of talk is that AI can streamline or automate what people do today, but they’re starting to emerge as we use more and more agentic in our work. They’re starting to emerge a new skill set to be able to get these agents to work in ways that are meaningful and reliable for us. When generative AI first came out, there was a new job created called a “prompt engineer.” And if you were a prompt engineer, you could get a salary of $300,000 or $400,000 working for a tech company. And very quickly, as soon as that job was created by generative AI, it was then eliminated by generative AI, because the narrative was everybody needs to be a prompt engineer. That’s no longer a job. Today in the Valley, you hear the term “context engineer” quite frequently. And a context engineer is somebody that’s responsible for designing a system and manipulating everything you can give to the AI system so that you get controlled reasoning and outputs in a meaningful way. The knowledge worker needs to bring their expertise to be able to use these systems in the most effective way. As you’re probably familiar, there’s an AI film festival every year, where directors are making entire movies with AI. We’re talking with directors to understand what skills they are bringing when they create AI. And as it turns out, there’s just a whole vast control system and a set of dials that they use that I wouldn’t even know existed because I’m not in the film industry. So when you have somebody working on tax, you probably actually want a tax expert working with the AI system as opposed to a generalist. And in essence, perhaps those knowledge workers become more valuable over time as opposed to being entirely commoditized.

Fuller: Where does that leave us, Dan, in terms of entry-level workers? Because I can well understand how a senior software engineer will look at software generated by Copilot or Cursor and other systems that support software engineering and see, both, how it’s working, but also spot how it’s kind of inefficient. But maybe that entry-level individual contributor, a recent graduate from a computer science program, isn’t going to have that type of insight. Are we going to see that truncation of the entry-level jobs for knowledge workers?

Diasio: Yeah, it’s something that, again, is the question really facing our industry of what happens to the apprenticeship model in the face of AI when we want to get more and more experts over time. And what we’re finding is it takes three things to be really successful in working with AI, both at an organizational level and also at an individual level. It’s, you have to have the right mindset, you have to have the right skill set, and then, third, you have to have the right tool set. Many of those entry-level professionals have this mindset and skill set already baked in, because they are using these tools at a much more formative stage in their life, and they’re able to apply this discipline a lot more broadly. You teach a pilot how to fly, even though most of the flying they do is controlled by the technology. Or, if you teach a surgeon who uses robotics to be able to perform an operation, they have to have a lot of hands-on expertise. We have now got to change what we teach these individuals as part of their job, because they might be using different tools to be able to go execute their work. And that question of “Is it going to be more, or is it going to be less?” is still something that we’re trying to work out. But the fact of the matter is that, if we take that abundance approach, there will be a lot more work to do as opposed to just doing the same work that we do today.

Fuller: It’s interesting to contrast that characterization and that nice triplet you provided with this sense of some exhaustion and AI fatigue in organizations. How can you reconcile those two things? And how is the firm helping C-suites and companies encourage their workforces to embrace this as something that’s going to be competitively important, but also something that they shouldn’t view as threatening?

Diasio: Yeah, Joe, we’ve done a study with a bunch of workers in the field, and 65 percent of workers when we released that “AI Anxiety for Business Survey” said that they were anxious about AI replacing their jobs. So just think about that: two-thirds of your workforce is worried about AI replacing their jobs. I mean, that’s where a lot of the exhaustion starts—is that they’re being told, “Adopt these tools. We’re going to track you and put you on a scorecard for adoption. We want to see that we’re using these tools.” Then we ask, “How much time have you saved by using these tools? What has been your increase in productivity? Can you do more with the same or more with less in a lot of ways?” I’m really impressed by the level of engagement that we’re seeing between CEOs individually or CXOs starting to lean in more to shape the agenda and the narrative for the company. We run a master class with many of our clients, and that is when we get the C-suite going through an exercise where they are using AI to be able to go solve a business problem or to be able to create a new business opportunity for them. They’re putting their hands on the keyboard and starting to see what they were able to do that was previously impossible, as opposed to just thinking about how we cut costs across our organization—which is one way that we see as a really strong technique to be able to cut through that fatigue and that underlying dread that many employees feel that maybe the AI will ultimately come for their jobs as well.

Fuller: Say more about how you think this is going to evolve and what kind of burden it’s going to place on educators, training programs, what we call “skills providers” in our research. You mentioned earlier that young people are showing up with a mindset and a skill set that’s bred of having been engaged with the technology earlier. Is that just a function of their own curiosity and lack of intimidation by technology, or do you think that’s beginning to reflect adjustments by those skills providers to make AI more integral to teaching and learning and training.

Diasio: Yeah, well, there’s definitely a bifurcation that’s happening in the educational space: those that believe that we need to teach people the fundamental skills without using any AI first to make sure they have those underpinnings and then they’ll figure out how to use AI later; and those that have readapted their curriculum. And I think we can certainly find, when we hire those employees into our firm, where they’ve come from based on what proficiency levels they have. But I think part of what we see culturally is just the digital nativism that many of the entry-level folks are coming into our organization. We all need to do a better job of teaching all of our employees how to effectively use these tools to be able to bring our expertise to them. And I think there’s a general commoditization of quality output that will happen if we don’t teach people how to bring their expertise to the table when they use AI. I need to bring the expertise and the context to the system to be able to get something that’s really nuanced, something that’s differentiated, something that’s personalized for me or for my organization. A lot of learning and a lot of training is still going to be required to be able to help bridge that gap.

Fuller: Now you’ve got tens of thousands of colleagues in the consulting arm of EY. How have you managed the reskilling of people in this? Because I can understand, I can hire digital natives. But if I’ve got a 45-year-old partner—and I say this as a former CEO of a consulting firm, so I have actual people in mind who have been highly productive in a pre-AI, pre-generative AI world—how do you bring them on that journey, particularly if 65 percent of them, not necessarily partners, are saying, “My anxiety here is if I work with this a lot, and it gets really good at what I do, it’s going to end up replacing me.”

Diasio: So a couple of things that we’re doing. One is we made the training for AI mandatory. And that was a big step for us as an organization, because it wasn’t just a single training. We had people first take a quiz, so then they could show us how much they knew. It was 25 questions, and if they scored below a certain level, we recommended that they take some more preparatory one-on-one–level courses to be able to really enrich their skills. But if they advance through that, then we put them through training that was really tailored to their level and their responsibility. And we’re making that to be more personalized by what their discipline is. So if they’re in supply chain, if they’re in risk, if they’re in our customer team, if they’re in our marketing team, then we give it much more contextual meaning. And we give people rewards, we give people badges for starting to accomplish that, but we’re also starting to recognize those people who are demonstrating it and demonstrating the use of these tools on their projects. So we set up a champions network in the U.S., where we are really investing in people to be the ambassadors of how AI can shape what they’re doing on the ground level with their clients.

Fuller: Do you view this as particularly people being reskilled, they get a familiarization, and then you’re basically done with this as a training regime, and they should go forth and start learning by doing? Or is this something we’re going to have to have an ever-growing array of interventions to train people to keep up with changes and to broaden their understanding of how something could be put to use?

Diasio: Yeah, Joe, what a great question. I mean, I think this is consistent and continuous learning. So I think there is a continuous stream that, while the application of that knowledge on the job is where the real magic happens, we are not one and done with respect to our investment in our people to make sure they’re getting the right programmatic and consistent expertise. But the culture of what it takes for people to use these systems is something that’s really important. And some of those newer professionals that were much more proficient with the skill set and the mindset than some of our traditional folks shared with me, that if you’re really good at working with these tools, the best thing is to be quiet about it, because some of your managers may think you’re being lazy for finding the answers in a more expedient way, and you don’t want to come off as being lazy, or you don’t want to come off as not putting in the hard effort, because our entire business model was built by demonstrating effort and working for that billable hour. And that was a real moment for us as an organization, where we realized it’s not the adoption challenge at the lower levels, at the entry levels; it’s the middle management—those that have been indoctrinated into this culture of work hard, stay late—that were the ones that we really had to invest in—in breaking through and getting them kind of onto a different mindset, investing in them to be able to see how this could be and make their job more fun, more valuable, more enriching.

Fuller: That’s really interesting. Dan, we’ve found very consistently in our research that, while very often senior decision makers are aware of needed changes as it relates to the future of work, that often when they set a new strategic direction, they miss a step, which is: How do we make sure that the in-place processes in the company are augmented, changed, modified? How do we make sure that those intermediate decision makers know how they should do their job differently in light of this decision? A great illustration of that would be so-called “skills-based hiring.” Of course, the problem was, if you didn’t teach what are called “hiring managers” how to interview differently and how to evaluate skills as opposed to degrees or what they should do differently from when they were recruited into that role or the people they recruited last month, you don’t get any change in the output. Dan, if I were a little bit of a skeptical CEO and said to you, “Gee, Dan, I know my board’s interested in this, I’m interested in it, but I’ve got a very strong competitive position. I don’t want to upset the apple cart. I can’t start making sweeping changes here, especially when I’m feeling maybe my revenue stream is a little soft or profitability might be an issue here.” What cautions would you give me? What would you say to try to get me to stop sitting on the sidelines and get serious about taking forward-looking actions now, as opposed to deferring?

Diasio: I’m sure there’s a level of skepticism that many CEOs are feeling with this technology. They’re just not telling that anymore in the surveys, because we see 97 percent of them are excited and are investing more and more in AI. But what I would challenge, I think a lot of that skepticism comes from job replacement and the opportunity to really just take cost out. And you referred to it earlier, Joe, this difference between scarcity and abundance, which I think is really the most important point that we as a society figure out for the use of AI. So, in that instance with the CEO, what I might challenge them was, if your revenue is feeling a little bit soft, what if you could go hire 1,000 more employees inside your organization? What is it that you would have them do? If you could go ramp up in a certain part of your business, what is it that you might try to do that you had not yet or you would not be able to do today? And when you find what that opportunity is, maybe they would say, “Well, really I need to grow international, so I need to set up a salesforce to be able to support my U.K. business or my European business.” Then we could figure out, “All right, how can we use AI to be able to solve that particular opportunity?” Because there isn’t a separate AI strategy for organizations, it needs to be an extension of the business strategy, and every CEO has a business strategy. So just figuring out where and how AI might be additive to that is usually the first way that we would start to try to unpack that.

Fuller: Dan, just as a final question: How do you stay abreast of developments in AI? What are you most interested in tracking? And, if for someone who’s not in the business of supporting companies on a day-to-day basis but wants to feel they have a greater command of the topic, what do you recommend they do? Are there sites or blogs you read? Are there periodicals that you think are regularly reliable in portraying what’s going on in the market?

Diasio: I subscribe to a variety of newsletters, some newsletters which are more skeptical of AI, where they are referring to the fact that we are near the top of the S-curve of AI and things are plateauing quite fast, and then some that are more coming from the technology space that dive deeper into how the technology is working. But I would say the most important thing that a person should be thinking about is how can they actually schedule time to learn about AI. I mean, this is a very important topic for everybody and just about every job. And I think just delegating this to nights and weekends or just when you have a minute is not going to be sufficient. For me, I allocate four hours a week to staying up to date on the topic, and that is down from eight hours a week in the early days when things were really coming at us quickly in 2023. I think it’s really important to start to carve that time out and then to get really purposeful the way that we do with the calendars that we manage inside our week to week.

Fuller: Well, I’m sure people are impressed that you’re still dedicating that amount of time to pure learning. And it does show that even somebody who’s right in the midst of using this technology and state-of-the-art ways and advising leaders on it still needs to make that type of commitment to feel comfortable that they’re keeping up. Given the velocity events in AI, I think it’s very, very good advice. Well, Dan Diasio, global lead for AI consulting at professional services giant, EY, it’s really been a pleasure to have you with us on the Managing the Future of Work podcast.

Diasio: Thank you for the invitation. It was a great conversation.

Fuller: 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.