Re-imagining consulting through the prism of AI

Welcome to the future of the FUN podcast. I'm Emma Carroll, head of Contentirisource. And in this episode, we're going to look into one of the topics that firms raise with us most often, AI, and its impact on the consulting model. In fact, today we're going to re-imagined consulting through the prism of AI. And it's an exciting journey, and we've got two fantastic pilots to accompany us on our way. Dan Diazio, global consulting AI leader at EY, and Colm Sparks Austin, America's technology consulting leader at EY. Welcome to both of you. Thank you. Excited to be here. Great to have you. Likewise. So before we head into the discussion, could I ask each of you to spend a minute or two introducing yourselves? Dan, maybe I could come to you first. Absolutely. Emma, I'm very excited to be here with you. As you mentioned, I'm responsible for AI across our consulting organization. This means that I support in how we go to market with our clients, but also how we use AI across our business to be able to power our services. And last but not least, we also have quite a bit that we are doing to upskill our workforce across the globe, whether you're a technical resource or whether we are-- whether there are resources with very little technical programming skills and are much more focused on the business for people's aspects. So great to be here with you. Thank you. And Kong. Thanks, Emma. And like Dan said, we're flattered to be here to say the least. So as you mentioned, I lead technology consulting for the Americas. I've been at the firm for about 14 months. Before this job, I was had a strategy for consulting. And it's serendipitous that we're having this conversation because a lot of what we looked at when we were looking at pivoting the firm towards sector and what that meant was really taking a look at the effect that AI would have. So in terms of how I spend my day to day, I spend the majority of my time supporting our teams and serving clients. But increasingly, I'm spending a lot of time with Dan looking at how we can re-underwrite our business for the leverage advanced technologies of which AI is one of them. So again, happy to be here and looking forward to the discussion. And it would be great to kick off with you both sharing those one or two big trends around using AI and consulting. And Colin, you've got the microphone. So do you want to kick us out of there? I mean, you're supposed to ask the AI expert first, so I don't understand the intelligence. But yeah, I know. I think it's almost like a paradoxical question, really. Is the use of AI and just the general conversations we tend to have with clients? And I'm sure we'll get into it as a major point here. But I think we've approached this paradigm shift like many others in the past. I think you went through the internet, you went through mobile, you went through cloud and SaaS and RPA. And I think a lot of those discussions we're having today with our clients are reorient in them around. And even internally here, probably more here is that this is not going to be like the previous shifts. And so I think everybody has framed this as the use of AI. We're going to use it like we use automation. We're going to use it like we use CRM and ERP. And I think that's why you think a lot of-- or you see a lot of these programs are just simply not delivering what people may have imagined. So the number one conversation I'm having, which is number one, two, and three, is we need to start thinking about this a bit differently because I think we came out of the shoot processing this like another technology or another tool, which is why we're in the position we're in relative to-- I think but the MIT paper that talked about 90% of AI engagements just simply haven't delivered. So I don't think it's really anything more than that conversation right now because everybody's retrenching to think about, OK, how do we do this a bit differently if that makes sense? Thank you and done. Yeah, Colin. Just to build on that, obviously, I think that is one, two, and three. For three years in from when Chatship T kind of hit big in the market, I think many of our clients are starting to recognize and starting to get the sense that they've experimented. They've figured out what it's good at, what it's not good at. And it's now really time to re-up the level of ambition of how AI will start to challenge and transform the business. And in doing so, it's a lot more of a focus on not using AI, but how to reinvent what we do with AI. So a lot more of a question of just using the technology is bolt on, like Colin had described. Particularly, I've seen many of my clients link this towards the RPA, boom, about a decade ago. And recognizing that in order to really create long-term and tangible value, there's a clear need to reinvent and reorient the way the business flows with this technology. OK, thank you so much. We'd definitely be diving into that straight away. Perhaps first from firms perspective. Colin, you talked about how AI works best if you rethink processes from first principles, rather than trying to retrofit it into human-designed processes. It would be great if you could share an example of that, that you've used yourself, that you've used at EY. I've got to think really hard here, Emma. And stuff we've announced to the market, what we have in the market. So I would tell you right now, just simply, when you look at a lot of the business transformations that we're doing back to the point of, in the past, we would have just applied technology to the way we do business today. And typically, this is where I think that the business cases fall over in terms of efficiencies and cost take out. And that is a very familiar movement to the market. Starting from, again, I mentioned when we came into the internet boom, and there was a massive buildup of infrastructure. And then we had to seek savings on the back of that. That's where outsourcing came from. And so what did we really try and do? We just tried to get the unit cost of the widget that we were delivering to the business down. And through the various waves, I mentioned before, that that has been the pursuit until still to this day, when you look at what people are trying to do with global capability centers, when you really peel back the onion, it tends to get into, of course, we're going to increase the capability of the business. But it is a lower cost of delivery. When we look at rewriting what I'll call critical process with first-prince AI, first-principles, that's where we're having fundamentally different conversations where it's not about taking out headcount. It is really about making sure that humans are doing the highest value at task. And we're delegating some of the toil to the machines, if you will. And so I think for us internally working with our clients, as we look towards a lot of these transformations, where we've had a lot of success is just really walking them through. There's this layer cake that we like to talk about, which is at the very top, you've got your domain critical process skills. In the middle, you've got more of your traditional functional skills, think, mid to back office. And then, of course, you've got your technical skills, which is knitting and all together. So for us, that recipe with clients has really resonated. As Dan mentioned, we're still early innings on this. And there's still an entire market that wants to consume this as a tool and wants to pursue a lower cost. But I think most effectively, that's where we've rolled that out in terms of conversations with clients. And done anything else to you, Dad, there? So Emma, when you look across the body of our consulting business, in every way that we provide services to our clients, we're looking at using AI to be able to extend the value that we can provide. But let me just use one example in the way that we provide risk management service to our clients. So what column was referring to? And in terms of breaking down the first principles, I think that we found that that's incredibly valuable. There's a part of our business on how we manage third party risk for our clients that used to be a 76 step process. And during the first iteration, we asked the team that it was expert in conducting that process to identify where and how AI could really drive value. And they went through the normal progression of they identified the eight spots that are quite painful. I had a lot of toil associated with them and used AI to discreetly transform those eight steps in the process. But if you look at it, everything else really stayed and followed the same flow. So it's at 45 hours to execute a particular transaction. We found that we were able to take that down by 12 hours. So that's like that is the typical approach that you go looking for use cases. When we start to think about reinventing, we also challenged ourselves, something to just feel right there. We asked a separate team that did not have as much connectivity into this business domain, but was really expert with the technology and took a very product-centric mindset. They brought a couple of people that were well versed in the process. And they said, actually, most of these steps no longer matter. Instead of looking at the steps, they broke it down from a first principle perspective in what the jobs are to be done and reconstructed an entirely new flow that instead of executing over 45 hours or 32 hours, now took about 16 minutes. And we added in human reveal. So that kind of reinvention is really important. with this particular set of technology, and it requires kind of decoupling oneself from the old way of working and really focusing on trying to reinvent the new flow. - Okay, no, that's really fascinating. I mean, that's such an astronomical difference. So with that in mind, what are the limits of using AI and consulting? Are there even any? - There are still some limits with respect to acceptability with some of our clients and some technical constraints. The closer that you work with this technology, the more you start to see some of the things that it is not good at. And that means that there are more controls that are built in. I wouldn't say it's a limit. It's just a consideration in terms of how you build it because a proof of concept is quite easy to attain, but to put something in an enterprise grade scale and production across the business requires quite a bit of controls, a lot of guardrails, and a lot of risk management. Also the big area that is still unsettled business is around copyright laws. And for that reason, this is still quite a sticky topic with many legal firms, which is like if you use AI to generate something that infringes on copyright, where does the undemnity sit and is that necessarily covered? So there are still some challenges in working through how to get AI to really take a front row seat in terms of executing many of these processes, but I think most of those are things that we are working through with our clients to make sure they are not just limited barriers. We still believe the flow and still see the flow operating as human AI and human. And what that means is that I really haven't started to see the reliability of the technology at this moment in time, be such that entire functions are entirely run in an autonomous fashion by AI. We still see that there's a lot of creativity that's required on the front end in asking the right question, and then on the back end in making sure that we review things. OK, so no entire functions run by AI, but are there any sort of ring fence things that are entirely run by AI and consulting now for you? Yeah, absolutely. I mean, in the example that I was describing before in those 16 minutes, that is all machine time. And people are responsible up front, and people are responsible for reviewing the output that comes at the end. But that is essentially 45 hours worth of work happening in about a quarter of an hour that is run in a completely autonomous fashion. OK, and Dan, you said it's really important to be able to spot the things AI isn't very good at. In a consulting space, in the professional service space, what are those kind of things that it's not very good at? The real challenge that I think many of us see is that it would be really great if we could point the things that it was not good at, and where it would make mistakes. But with the stochastic nature of this technology, sometimes things that are basic and obvious to us are missed. Like, for instance, I have an eight-year-old and I was asking one of the AI systems to check her homework, and I had to convince the system over three attempts that three times seven is not 24. OK, so like, so, and these are the sorts of things that start to erode on some confidence, where it's really essential to do some risk management and some quality assurance. There are a lot of ways to get AI to start to evaluate other AI's work, but there's no single bullet of what AI is not good at. It's incredibly powerful at a huge breadth and variety of things and sometimes make some simple errors that requires some additional work. Can I just build on this point? Because I think it's really important that back to my previous point on more than just reduction of task and cost, right? Because I think the one thing that we've learned through this journey and this tool was built by, this wasn't a tool, it was an entire workflow and process and eventually an application was built by a team that we acquired through our New Vail and Sacquisition, which was a product engineering team that we, I guess the firm was about 15 months ago, but 150 people focused on product engineering and their original contract with EY was to come in and look at some of our processes and what we were doing, actually. We used them to really come and interrogate some of our, what I'll call our downwind businesses like risk, like evaluations and diligence. And so through the process, really what we've learned is that these models and these agents, when you first build them are a lot like Dan's eight-year-old, actually, right? They don't have great memory, right? And when you think about from a human brain perspective because at the end of the day, that's what these things resemble, is their neural pathways aren't really that programmed yet. And so you can go through a process with this technology. And the beauty of it, in some cases, it's like a child will forget the process, right? When you come back to it and you're prompting it again, it could make the similar error like Dan talked about. But through building risk.ai, which is what we're officially branding it now, is this idea of a workspace that is, you know, a model plus agent plus practitioner, right? And it's this process of codifying human knowledge. And so the initial use case was to take third-party risk management and again, remove the toil and level up the consultant or the practitioner. But over time, it's going to be a place where, you know, Emma, the risk consultant, shows up every day and codifies her knowledge and gets the benefit of all of her, you know, thousands of risk practitioners through the organization codifying their knowledge in their exchanges with clients. And so we start to build this corpus of knowledge that, you know, has again, leveled up Emma's daily activity because, you know, quote unquote, the machines are taking care of a lot of the tasks that were previously quite arduous. She has more time to work with the client and work with the models to ask questions and prompt and understand like, hey, what am I missing here from other engagements? What am I missing from regulatory filings? What am I missing from, you know, various data sources? So increasingly, our business has to be this motion where we build tools and technologies that are consultants enrich every single day with what their experience with their clients. And like, that is not about, you know, we're taking our clients data. It is simply just codifying the experiences we have and we both benefit from it. They get, you know, possibly a better product or lower cost of delivery. But for us, it really enriches the human experience with AI, which comes back to this idea of like, you know, job displacement and everything else. Like, I really think this is going to be a market of abundance driven through proper application of AI as opposed to, hey, let's try and do this with less consultants and deliver it at a lower cost. Yeah, Colin, I want to build on that point of abundance. We actually see this as a real two-step process. So the first step is to do what I described earlier of keeping the inputs and the outputs the same and really challenge how the work happens. And that is usually by getting to first principles or jobs to be done and starting to reconstruct an entire new flow. The second step then is to challenge why were we doing the work with the constrained inputs and outputs? So now that I have this new capability that is no longer susceptible to the limitations of how long it takes to execute the work. Can I start to do something that was previously impossible and create new stream value? And in that particular instance, you know, we asked ourselves, well, why do we do this? You know, we end up doing a couple hundred of these for many of our clients. Why do we do a couple hundred? Well, we do a couple hundred because it takes 45 hours to execute. If we can now do it in, you know, 16 minutes of machine time in a couple of hours of review, well, should we start to look at, be able to, you know, turning that on on a much more regular basis to help our clients better manage their risk with some of their, some of their more, for lack of a better word, for lack of a better word for some of their more risky activity. And it was never humanly possible to be able to do that. But now we're finding new levers for scale. So that that second part to the process is really when we move away from the reductive and we don't just look at what's happening today. But instead we try to unlock new things that were previously impossible. And then maybe building on that I'm last year. Our data here at source was telling us the most AI work that clients were buying was really about experimentation and really understanding the options that AI could offer. Is that still what they're buying or is it shifted now? Yeah, I would say, you know, to the trend that we laid out the outset. I think most of our clients are moving away. They all have some center of excellence and you are many of them have a center of excellence inside the organization and are working with some form of AI and learning on how that either works at scale or what some of the limitations are. What we are seeing more of our clients requesting from us now are reconstructing the flow of work and thinking about how AI changes. It's not just installing AI, but now it's figuring out how you rebuild around AI. And that means what you build from a risk blueprint perspective, like how you manage guard rails, how you start to think about manage and think about managing agent identities. You know, what skills are required in the workforce to be able to work proficiently with AI. So very much focus on the human workforce, the new jobs that get created inside the organization. Looking at. you know, something that we call quite a bit, we call these blueprints. And these blueprints are how you can reconstruct a process to be able to execute something from an end to end perspective using AI. So that seems to be a lot more of what our clients are asking for around how to get these things to scale and to start to do different things as opposed to just replicate and automate what it's been already done in the past. - And Coleman, is there anything you'd like to add in terms of what clients are using AI for or what their buying is changing? - I would say the more aggressive clients are starting to get into very interesting dialogues around what was previously considered impossible, right? Like what, especially from a technology perspective, you know, there was this age old adage, which is, you know, before you do anything, you've got to re-figure, you reconfigure all your systems, get the data out of systems of record, get it into like, you know, central place, data like et cetera, then you've got to build an integration layer and then finally you might be able to do something productive with the data. I think what we're seeing most importantly now is removing ourselves from like kind of the technology and the data and going to the edge where you have a clear business problem that has been limited by technology and the assumptions around how these systems work together and revisiting that being like, how do, like what has to be true for us to achieve an objective that was previously undiscoverable or intractable with the technology at the time? That is that and Dan and I talk about this all day with our teams. That is the most fruitful dialogue around AI. If it was possible before AI, and it was simply you want to speed up the process, that's probably the wrong way to think about this. I mean, you can get some short-term games, but I would tell you that from our firm perspective, we want to bet on the long-term value capture and release in these organizations. And we've learned that from our transactions business. When you look at our UI Parthenon business, this idea of discovering value, orchestrating that value and realizing that value, that's really where we want to be sitting with clients. Because again, the other side of the ledger, which is very crowded, commoditized, and it's very short term is this idea of reducing cost. What it leaves behind in its wake is not a very pleasant environment for anybody to work in. And therefore, this growth agenda, which lets understand where the value sits, lets enable humans to deliver more value to the business. I think if I was asking you to join one of two teams, you probably want to join the latter versus the former. And so for us, it's this idea of just generally going into clients and pushing for the really deep conversation. What is that problem set you want to unlock? There's just been previously out of bounds, and there's now in bounds with AI. That's super exciting for us. And I think most importantly for our people, from a purpose perspective, going to work every day and solving something that's going to move the top line of a business, that's something you can talk about at dinner with your family. So that huge amount of consulting work that was around getting your data fit to use for AI, listening to that, is that kind of over and done with? Or there are just other more strategic things that you're prioritizing? No, I think there's always going to be that type of work because over time, if you're going to have really powerful systems, you do have to clean up some of this in the past. But I will tell you that just in terms of some of the work and Dan can elaborate on this from a technical perspective, there are new techniques and new technologies that allow us to skip ahead in that journey and deliver value immediately. And some of our most advanced clients are thinking about on their own side, codifying their knowledge, codifying their environment such that they can, again, it causes a lot more investment up front. But over time, it allows you to have more of a dynamic decision-making environment that is just more than just about the technology and information. It really allows you to synthesize and really think about what changes you want to make and understand what effect that's going to have on your business. And that really starts out on the edge of the client. So Dan, do you have anything to add there? Because I know from a technical perspective, you can probably go a bit deeper. But that to us is the real opening in the market. Yeah, and as Colin said, I'd say a lot of that work, that preparation work of making sure we have a platform in place, making sure we have the right data ecosystem around it, et cetera, et cetera. I think a lot of that work still happens. It's just less general purpose work of general preparation and foundation. And it's much more targeted in a specific outcome. So I think the big shift has been, let's stop doing AI for AI saking. Let's start talking about how it's going to deliver some sort of material impact on my business. And what Colin and I's teams were, and what we're trying to do with UI is to shape that into a growth business, a growth opportunity as opposed to just a cost out discussion. But once you start to lay your eyes on a particular part of the business where you think there's an opportunity to reinvent the organizational context, the knowledge, the data, and all that. It's still very important to be able to run and execute these systems. It's just that is working towards a specific goal as opposed to clean data being the goal. And I have a really interesting overview on some sort of trends we've seen throughout the year in terms of appetite for AI with clients. There did seem a bit earlier in the year, both in our data and in the conversations a bit of fatigue around using AI and exploring AI. Although this quarter, it does seem to have rebounded to the top of clients' priority lists. Any views in terms of that? Have you seen that too? Yeah, we have seen-- we did also get a blip. We run something called the AI poll survey where every six months we're out asking a number of workers, external workers and external leaders on how and where their AI initiatives are progressing. And there was a blip a couple of months ago. Where it did seem like enthusiasm was down and fatigue was up, I'd say. Now with Agente AI being the topic that is really persisting its way across the market, we are seeing a lot of excitement of getting rid of the toil and off people's desks. I would say I have some statistics here, but 84% of those that we asked are really excited to embrace AI and the way they deliver their work in Agente AI and the way they do their work. Many of them are overwhelmed though, because this is still happening quite fast and quite quickly. But I think what starts to get at the point that you refer to, Emma, is that of those that we surveyed 83% of those said they were self-taught. And I think what is fatiguing is that you believe it's going to have some sort of an impact on the way that we deliver our work. But yet, many people are reporting that they're not getting the right skills, they're not getting the right training from their organization to be able to keep pace with the progression and they're being asked to use it. So I'd say that that is probably how I would capture what it means from a fatigue perspective. You must use this tool in the way that you deliver your work and figure it out. But we are starting to see a shift in companies prioritizing big, broad training programs across the organizations, identifying how people can develop different career tracks and working better with AI. We've seen some of our advanced clients start to train some of their employees on how to be really proficient context engineers in the way they do their work. So I think these are, at technology that's moving this quickly, of course, it's very difficult to be able to stay on top of it. But I think now as things are starting to hit the enterprise grade scale, many organizations are starting to really get the coordinated training programs in place. OK. And one of the really tricky points for firmness and movement pricing, and pricing around AI enabled services, how are you thinking about that, Dan, at EY? Yeah. So so much of the consulting business model is based on effort. We were pricing for input. So it was like rate's times hour. We see AI as an opportunity to share skin in the game with our clients and get much more into outcome base pricing or output base pricing. And an AI is a great driver of that. We've seen that our clients are very interested in moving down that path to it because they do feel the skin in the game. And they know that they're signing up for value as opposed to signing up for a plan. However, I'd say most procurement models that we that we've seen be built up over the years have really been geared towards the P times queue type of business model. So we are excited. And with many of our clients pushing into ways that we can jointly create value and get paid for the value that we're creating. And we are finding that a lot of the drag is just trying to work through how we really engage in the contracting process in a way that meets our clients where they are. Because we often-- they close roughly and saying to us that they're really attracted to outcome base pricing. But they like it a lot to protect themselves from the downside, as opposed to reward firms for the upside. Do you think AI will shift anything there? I sure hope so. because this is an opportunity to create very differentiated value. But, you know, similar, we also have run some joint research with procurement functions to understand where they are. 81% of them are really excited into moving into output or outcome-based pricing with AI, and they're very interested in hearing how professional services firms like ours can start to use AI to be able to drive their work, but yet 17% of them had mature processes in place to be able to contract around that. So that's kind of what I was referring to, Emma, is that many of our clients, just like we are reinventing the way that we deliver our services, our clients will start to, you know, need to reinvent the way they buy those services, and that always takes a little bit of time. Okay, thank you. And, Cole, you've talked a few times now about talent, and about the ways that working in this way in firms, using AI, can actually be quite rewarding if it's employed in the right ways. And so I'd love to hear from you what you think, you know, the workforce in professional services will look like maybe, you know, five years ahead. Yeah, I mean, I've had this discussion with the several of our teammates here, and I believe that, you know, again, if we treat this like a tool and think about like, hey, I'm going to measure adoption of a co-pilot or I'm going to measure adoption of an agentic tool. And, you know, I've seen some other firms talk about putting this into end of year evaluations that like we're going to measure how much you use AI. I just, I think that's probably a bit of a knee-jerk reaction like this isn't their technology, right? Like, you know, it's back to the days like, you know, how fast can you type, right? Or like, you know, let's make sure Dan was using, you know, Microsoft Word to do his word processing versus, you know, having, you know, your pee, he annotates it to somebody else and they do it. I mean, I think this is where we have this knee-jerk out of the gate to say, hey, this is what we're going to do. We had the head of engineering from Shopify, a great Canadian company, just plug in Canadian companies, just for the record, I'm Canadian. Farhan Tharwar from Shopify, he talked about, you know, their journey on engineering and how they've enabled their people with AI tools, right, from code agents to co-pilots, you know, what have you. And this is a company that, you know, doubled revenue, well, you know, reducing their staff by 30%. Right? And so he talked about that journey. And, you know, the AI moment for them when Toby sent out that the note saying my expectation is that this is an AI first environment, they have not measured any adoption of AI, right? Their take is, and again, I look to them as a leading engineering firm is we're going to provide you with the tools and we're going to provide you with the challenges in that end in culture to go and execute with those tools. Your productivity and performance will be very clear whether you're leveraging those tools or not. And then that's kind of step one. It's like just create the environment with the tool set. The strong, the strong engineers will will gravitate towards those two tools. His other point was not every task is for AI, right? And again, these unintended consequences of driving metrics that might cause the wrong behaviors is there's just some tasks that aren't that for AI. You know, we've been looking a lot at coding agents lately and we're going to come out at the end of the year with a software factory working with two other external firms. And when you look at the code that comes out from a lot of these coding agents, you know, they are, it's quite verbose, right? Compared to what you'd probably write. And so it takes a lot of human interaction to clean that up, right? And in some cases, it'll just come out with, you know, you hear AI slop all the time. And to use AI to fix AI code is almost like, you know, far-hand use the example of like, you know, using a power drill to try and turn a screw a quarter of a quarter of the way, right? You can just clean up the code right in front of you. So, you know, along an analogy there, am I just to talk about how like we've really got to think about how we enable our people, be trained them. But most importantly, create an environment in which like they really own the choice of how to use these tools because again, AI is not for everything, but over time, you know, reducing the manual tasks and toil associated with some of these jobs. And really getting into the deep problem sets, which is like, hey, how do I move faster to solve problems, which are previously unsolvable? That's the environment we want to create for our people. So we're having a lot of discussion right now about getting best of read tools, what we should be bringing in, iterating at pace, and creating, you know, probably a flatter organization, which, you know, Dan mentioned. I mean, Chad G.P.T. has been up for three years, right? A lot of the folks coming into our environment grew up with these tools. But, you know, traditionally in the consulting model, they'd be considered staff and, you know, further down the pyramid. We wouldn't necessarily listen to them. We'd be, you know, saying, hey, come learn consulting before, you know, anything else. And I think it's just going to be the other way around there. So for us, AI adoption is going to come naturally if we get those those building blocks right. It's interesting you talk about a flatter structure there, because you know, there's a lot of attention on whether the the junior layer will be the one to disappear. So is that how it will become flatter? No, I mean, I had this, we had a conversation in a couple weeks ago, one of our major hubs with partners and managing directors and I asked, I said, who thinks the bottom part of the pyramid is, is that risk here in this, in the shift the AI. And luckily only half the room put up their hand, right. And I started laughing and everybody had like a nervous laugh, which is like, you know, the 200 of us are probably most at risk. Because our, our AI reflects it. The way we use AI is like we have to force it, right, whereas like a lot of these earlier generations, like they just do it. Right. So I think what's going to happen is the floor of the floor of of skill will rise, but so well, you know, the top part of the pyramid. And I think it'll free our partners and our, and I'm answering directors to be out there. And leveraging what they do best and understanding our clients, meeting the where they are, and then pushing more sophisticated tasks into the middle and bottom part of our pyramids. And that's where I think we're just going to see incredible creativity and lift from some of our staff and seniors that normally may have been doing toil task because we feel it's part of the training method. Thank you. And Dan, I'm really interested to hear your perspective on some of the conversations I've had with clients where they've been saying eventually I think we'll be using consultants less because we might be using AI to do some of the things they do ourselves. How does that resonate with you? What do you think there? Yeah, I certainly think that the nature of the way that we support our clients will evolve. But, you know, what is successful AI adoption inside of an organization? It's not just using AI, you know, like that, like, you know, step one is if everybody used AI, most of the output is going to be the same. There's no real differentiation. It puts you down a statistical average. You know, so, so like you'll often hear referred to the fact that context is king. And what is context context is domain expertise. It's critical thinking. It's systems thinking and understanding how to turn all the dials with the AI systems. It's being able to bring creativity into the mix. When I start to describe those things, I think of consultants. You know, and consultants are kind of in many ways really trained to be able to ask the right question to figure out how to create value inside the organization. And that is going to be the differentiation for many of our clients. They start to use AI in different ways across their business. So while I do think there will there will need to be an evolution. A lot of the things that require very little creativity and just execution and just require some extra arms and legs. Like that will likely be done by agents. But that will be done by agents with context that we've built from a wide variety of clients that we've supported in a particular industry or sector. And the real interesting value creation activity. I think that is always going to require some additional support and people that bring like that outside in perspective that have a wide breath of of different experiences that they can bring into the way that they engage with AI. So so long, long winded way of saying Emma that I hear a lot of times people refer to AI as taking the cost of knowledge down to zero. And that means that the impact on knowledge workers is going to be really acute and we're going to, you know, we're really going to see mass unemployment in the knowledge work business. And you know, if we if that is actually the way that things turn out and what happens then then we are just converging to everything starting to look the same, you know, based on AI. And I just think, you know, most organizations are going to continue to look for differentiation and that is going to be the context that works with these systems and consultants are really consultants and knowledge workers at whole are really best place to be able to bring that those skills in working with AI to create output that is unique and differentiated. Yes, that's the real value, isn't it? Yeah, if I can build on that. The source to give you credit, you came up with a great chart last year, which was the quality to value gap that we're seeing in the market, right? I think it was 81% of clients perceive value or sorry, quality in the work that consultants do and then it was something like 34% you know, feel like the value out of that. There's two parts to that issue, right? I mean, again, as we mentioned before, there's an entire market that's conditioned to buy on costs and so they procure a certain way, right? And when we deliver a program to the procurement requirements, like, yeah, it's quality work, but when at the end of the day, the value is almost forgotten. By the time we get to the end of it, you know, through many rounds of contracting and, you know, technical specifications. So I think that's one problem. But again, there's an entire industry out there that is programmed to think like this. And so, I do think for us closing that quality to value gap, the consultants that simply configure, install, advise, I don't think that's going to have a ton of value going forward. Again, I think there's there's, there's, there's converging markets here. So that will last for a while. Our thesis and where we're investing is that we have to provide value through value discovery, value orchestration and realization. And our ability to get at that is investing in a combination of, again, that domain expertise, leveraging our functional expertise. And for my organization, building out the technical depth to then go deliver on that. And I do think once you can start to really condition the market to procure a certain way and price the value fixed fee, getting away from, hey, like what's in the labor model, what's in the effort model? You can really get towards closing that gap because that that gap is is is is an indictment of our industry. But it's also an indictment of the way things are procured. So it's going to take time and it will be threading an needle with our clients to see those that are ready to move more towards, you know, value discovery and unlock versus, you know, those that have to go through a necessary, by the way, I'm not discounting this part of the market, a necessary, you know, cost reduction and reset in their environment in order to be able to go to the next phase. So I think it's a huge amount of opportunity, but unless you over index your business on getting towards the value side of the equation, I do believe there'll be less consultants on on the previous market I discussed. We always like to end on a really practical takeaway. So if you, imagining, if you were advising a large professional service to sperm on how to get the best out of AI, what is the one pitful each of you would advise them not to fall into? Yeah, I can go. I mean, I think we talked about it earlier on. I mean, it sounds like when I was in the strategy role, a lot of what we need to do in terms of strategy is very obvious, right? And just it takes discipline and it takes focus. I see the same thing for our clients on this on this journey and professional first services firm as well as this idea of don't do what's obvious, right? That this obvious thing of hey, let's treat it like RPA, let's speed up processes, let's try and get the cost down, let's use top-down rules to force adoption. It's just it's just not the way that you're going to win with this technology. I think it has to be ground up looking at, you know, first principles, how people work every day, how do you improve that experience and, you know, how do you get at some of the tougher problem sets that were previously deemed intractable? I mean, that's how I just think you're going to, it takes a lot of effort and it takes a lot more investment upfront. But I think if you bet long term on, you know, humans plus machines or human plus compute and really leveraging, you know, as Dan's point, the human creativity with the scale and in capacity of the other side of technology, you're going to win in the long term. And Dan, do you have a pitfall? Yeah, the pitfall would be to think again, incrementally and just cost out, as Colin had mentioned, you know, I think it's a revolutionary technology. And at first, we always tend to use these things quite incrementally before we start to reinvent. And, you know, I'd say the pitfall is just thinking that everything is a use case and that everything is a way to make a step that happens faster as opposed to really challenging why we are doing that work in and of itself and how we might be able to have more value. Dan, Colin, thank you so much for your time today. Really appreciate it. Thanks for having us. This was fun. Thank you Emma. If you found today's discussion interesting, you can find more episodes on Spotify, Apple podcasts or anywhere else you get your podcasts. To find out more about how we're helping shape the firms of the future, head to sourceglobalresearch.com.