The River Doesn’t Wait: How Far to Take AI While the Technology Keeps Moving - with Blaine Mathieu의 전체 녹취록

parts of today's AI excitement feel familiar from previous technology cycles? Blaine Mathieu: Well, first I have to say the way you're describing me, I feel really old, but I'll I'll consider that wisdom instead of Karl: Ha ha Blaine Mathieu: instead of age, Karl. But enough about the 30 years, Okay. but but all but all good. No, it's it's good to have seen these things happen many times before, as you said. And and like I just mentioned, probably pace is is the fundamental is one of the most fundamental differences. But the other one is they're the category of things, I guess, is that in the previous revolutions. of technology, you could adjust to them by effectively figuring out what tool you needed to buy. Okay. What tool or eventually became what service you needed to buy through software as a service. But this is different. It's not just about a tool. It literally is about changing the way you operate, right? You mentioned earlier reimagining the way you operate because these are not tools. These are, I'll use the word you know, thinking. Not in a human sense, 'cause I want to be careful that we don't anthropomorphize this, but but these are thinking systems, fundamentally. It's not just about another tool. And so that's besides the speed, that's why this revolution is something quite different than anything we've seen before. Karl: Okay, well that that brings us neatly to the the central argument of of the river doesn't wait. You you previously explained to me that you could almost remove the words artificial intelligence from from your book and it would still be a book about running an organisation through unusually rapid change. What what does the river metaphor actually mean for a for a managing director maybe who is a a profitable, functioning business today but knows the technology landscape is is moving faster than their their normal planning cycle. Blaine Mathieu: Yeah, you make a great point. And when I f got the idea to write this book, The The River Doesn't Wait, it was really at the culmination of a long career where I've been thinking about corporate strategy and working in strategy circles as a Gartner analyst at startups and with large enterprises. And I, you know, I I really had in my mind the idea of writing a book on strategy and how to think about and adjust your strategy for a long time. But what was missing was the the inflection point that would cause it to have to think differently about about how and where you're directing your company. And so that was what began the idea of creating this book. And fundamentally you could remove the words AI from every page of this book and it would still be a valid guide to the methodology and framework of how to understand, set, and then constantly reset your core strategy and direction in the face of constant change. It's just that AI is the, as I said, the inflection point or the instigator for having to do this. You could think about doing that before. Maybe you could be an agile

or flexible organization at any point in time, but now you have to be. And so that's why that was the instigation for the book. And and this notion of a river is actually, I have to give my son, who's a very accomplished content person and writer, credit for that. I was trying to find a core. Analogy that could explain to somebody without any technical terms what was really going on here. And what my son helped me arrive at was this notion of this AI and this agentic AI transformation that we're going through is like a river that's constantly flowing. It never stops, you know, never flows backwards. sometimes it surges, sometimes it slows down a little bit. Maybe we'll talk about some of those. But Your organization is a boat in that river and you're either actively steering it and and setting it and directing it, or you're being pushed by it, one way or the other, right? And so I think that that I've used that river analogy to and the notion that if it's not waiting for you, it's happening, you're in it, whether you want whether you're ri driving your boat or not. That was sort sort of the core obviously framework and model that I've used throughout the book. And you know, I think as far as as Consultancies, it's a really interesting time because fundamentally these are very knowledge-based organizations. They're not just running a process, right? They're not stamping widgets out of a machine. This is about knowledge. And like I said a few minutes ago, this is a fundamentally a knowledge revolution that we're happening with AI. It's not a tools revolution. It's not about making your your your press a little bit more efficient. This is a knowledge revolution before anything else. Karl: Yeah, and just I'll congratulate you on the metaphor there. I think it's a it's a great metaphor for for explained Yeah, Blaine Mathieu: Well I I'll I'll share that with my son. Yeah. Karl: yeah, definitely congratulate him on that. I mean if we d if we take a three year strategy used to sound sensible. I if if the capability and economics of AI can move materially within a quarter even, let alone a year or more, what what what should replace that style of long range maybe technology plan without without leaving the the business permanently reacting. Blaine Mathieu: Yeah. Yes, and and to be clear, I'm not at all putting forward the hypothesis that we should abandon long term goals or directions for companies. Of course, we should figure out where we want to go in the in the long term and then figure out a a plan to get there, right? So that that doesn't go away. Karl: Yeah. Blaine Mathieu: But what is changing now and I think has already changed is the way we would set that kind of strategic plan is we would read do a read of the environment. Literally, they call it a reading of the environment. And then based on that reading, you put the plans and processes and systems in place. And then maybe you update that reading, your annual strategic planning cycle, you do an executive offsite, whatever the case is. But that now has to be replaced with something I call not a not a reading, but a watching, a continuous reassessment of what's going on in the environment. And think about back to the boat analogy, the captain doesn't just walk out and look at the ocean once and go back in and say, All right, we got it. Let's cross the Atlantic. No, he's got he's got somebody at the top who's doing a constant watching for what's going on with the weather, with other boats, y you name it. That's the only way you can navigate a diff you know, a difficult sea, right? And so the same thing applies to organizations. And there are a few things you need to be watching for, if if you'd like me to talk about that. Karl: Course, yep. Absolutely. And do Blaine Mathieu: Okay, well yeah, go ahead. Karl: it I just mentioned that I just love the fact that you're bringing in water analogies everywhere. So I'm looking looking out for them from now on. Blaine Mathieu: Okay, well, and to be clear, you do not have to be a sailor to understand the core framework that I'm talking about here. I've I've had Karl: Yeah. Blaine Mathieu: to be a little careful because you in writing these the this book, you could really go overboard and start talking about the mizzenmast and all kinds of crazy stuff and find the analogy. Don't worry, nobody has to be a sailor. But I think the the core analogy will be understandable by by virtually anybody. So Karl: Course, yeah. Sorry. Blaine Mathieu: Yeah. So the thing you're watching for, b back to back to the the question is probably three things. First, have the capabilities of the technology themselves changed in some fundamental way that opens up possibilities for you that weren't open before. Okay. And it can be as a quick aside, it can be hard to detect that because we're just flooded with news about what's going on in AI, but we can we can maybe talk about how we get over that one. In a bit. The second one to watch for is has a competitor or maybe a comparable company raised the bar? Are they using AI and agentic AI to do something fundamentally different in a different way today that that hadn't occurred to you yet? Or maybe you're not you're not doing it yet, and you realize, wow, to to remain competitive, we have to think about something different. And then finally, the third instigator to of this kind of change is. What's happening internally? Has there been some kind of internal change, an inflection point, something different in how you want to run or operate your business, or maybe what your long-term goals are that is, you know, causing you to want to reassess these things? So it's sort of external forces, technology, competitors, and internal dynamics. You have to be constantly watching for significant changes in any of those three things. And to be clear, Driven by many different forces, there's always gonna be things driving change in your environment, but I'm really talking specifically about

AI and agentic AI for in this context. Karl: Yeah, of course. Yeah. So I I'm seeing it myself, just feedback from different companies. th there is a lot of pressure on on the leaders to to do something in AI, and and they don't know what necessarily, they just know they haven't got time to do the analysis themselves. They just know that, you know, that there's this you know, just seemingly urgent requirement to do something. Just do something, yeah. Blaine Mathieu: Just do something, anything, right? Karl: And they're giving it to the IT Blaine Mathieu: Right, right. Karl: guys maybe to to look at thinking it's a it's just purely a technology solution. But Blaine Mathieu: Mm. Karl: but how what how do you really distinguish from this this maybe healthy urgency from what sometimes feels like executive panic where an organization just starts collecting technology tools and pilots simply so it can say it is moving forward, it's it's embracing AI. Blaine Mathieu: Right, right. Well, and panic to some degree or an a heightened sense of urgency, let's just say. I'm sure most of our listeners aren't actually in panic mode right now, and or and they Karl: Yeah, hopefully not. Blaine Mathieu: shouldn't be, but but they probably have a heightened sense of urgency about this. And and in my experience, not just through the AI revolution, but generally in technology, that often leads to pilots. Okay. We have to do a pilot. We have to test some new technology. And these are often characterized as technology pilots, right? And and what that leads to is you mentioned the IT department or the CIO or whoever's running the technology part of your business. And in small businesses, it could be your son in law or something. But in any case, it leads to AI figure AI out, okay? Because AI is thought of as a technology problem. Th and therefore you solve that problem by running a technology pilot. And the point I'm trying to make in in the book and in the frameworks and what we do with clients at the River Group is it's fundamentally not a technology problem. In fact, already the technologies of AI and agentic AI are far more advanced than the vast majority of companies are even able to take advantage of. This is a strategy problem. It's it's a chall or not a problem, but a challenge about How you want to reimagine your organization's operations, its effectiveness, where it goes in the organization. And then AI is just an enabling technology or an enabling tool. But panic leads to pilots. And pilots, as we know, it's well, you know, well accepted that the vast majority of pilots fail. And in my experience, most of them don't fail because of the technology failed. They failed because you get to the end of the pilot, you realize, okay, that that sorta worked, but what difference would it really make? To the business. Not much, right? Karl: Yeah, so i it's interesting, you you mentioned the word obviously im reimagined then we mentioned it at the start. So there's this retrofit versus the reimagine. So at one end you can use AI to to make an existing process faster or cheaper maybe, and at the other end you can redesign the process, the roles and perhaps even the services or the service itself. How how how should a leadership team really decide where a process belongs on that spectrum? Yeah, where do they get to grips with that? Blaine Mathieu: Yes, yes. Yeah, the the most fundamental framework of the book is this notion of retrofit versus reimagine and how you can use agentic AI to to do something on that scale. And I do want to make it really clear. I've I've used the word reimagine a few times now because it is, you know, it's sexy and and it it brings forth the idea of wow, we're gonna really do things differently. But I I'm aggr I make great pains to say that there's no wrong position. On that scale. It may very well be that a retrofit of an existing process or a retrofit of some operational logic in your organization to make it faster, more efficient, cost less, that might exactly be the right thing to do. And maybe you need to be doing that right now, maybe because your competitors are doing it and their process getting 50% more efficient. So if you don't, you're gonna have to charge higher prices to your clients at the end of the day. Or you're going to deliver that report or that analysis in six months and your and your competitors can do it in three months. So a retrofit of the way you currently operate very well might be necessary to do right now using Agentic AI. Having said that, what retrofits generally don't accomplish is long-term differentiation from your competitors. Because retrofits, because they're relatively easy. And simple and don't fundamentally change how you think about operating the business, your competitors can do them just as easily as you can. Okay? So it might be necessary to do, but it's it's rarely sufficient to actually pull ahead in the market. If you want to pull ahead in the market, you bet you should think about how do I think about doing something differently than it's being done today, not just faster or cheaper. And that's what we call the reimagine. And Agentic AI technologies, AI can enable everything on on that spectrum. But I would encourage leaders of organizations to to open their minds to all possibilities, not only the quick and easy retrofit. Karl: Okay, so just moving on slightly from that, so recent enterprise resear research that don't have seen i is showing a striking gap between individual productivity and and measurable company level value. So i Blaine Mathieu: Mm. Karl: is is that what you would expect if firms automate tasks but leave the surrounding workflow maybe or the handoffs and decision rights largely untouched? Is that is that what's gonna happen? Blaine Mathieu: Yeah, I think there's a few things going on there. So that is to some degree an example of a retrofit first. Let's just give a new technology to our employees and see what they can do

with it. And and maybe they can, you know, produce the the speed of their an analysis or their research or their report creation or whatever their knowledge work task is, they c they can, you know, im improve the s the efficiency of doing that. That is a sort of a retrofit by by definition. And it's also the case historically in in technology evolutions that and it's the same in this one, that technologies are being brought into companies from the bottom up. So even without the IT departments officially controlling whether it's mobile or the web, and now AI employees and knowledge workers especially are just bringing it in. And in many cases the the leaders in the organization don't even know it. Okay. And and that's not new. That's a time worn element of how technology infiltrates organizations. The challenge, though, as we just discussed, is generally speaking, that is going to lead to a retrofit or a bunch of mini retrofits maybe happening in each individual's workflow and what they do. It's very unlikely to lead to any kind of long lasting differentiation in terms of how you actually fundamentally operate the company. And so it's not surprising to me that that's happening today. It's being led on the one one-by-one basis. And but but I think that has to change to get the full value. And on the technology side, we're seeing that evolution happening too. The reality is most of the AI technologies up today, today, have been personal in nature. So chat bot one knowledge worker talks to and collaborates with. And it's only recently that we're starting to see the technologies be focused more on teams and on, organization-wide implementations and not just individual implementations. so that you know, so I think the the evolution, the use cases, the infiltration, the technology adoption lifecycle, you might say, are all evolving more toward organization-wide reimaginations. Karl: Yeah, it's interesting, isn't it? Because some of the drivers that you've been talking about for implementing AI or or almost the opposite of what some people are saying, that you know, people are are worried about their their jobs because of AI. We'll talk about that in a bit. But there are the it's the opposite as well, isn't it? There's the the AI champions there who are inadvertently they're spotting issues with their workflow on a personal level. And then they're implementing their own little AI solutions to help themselves. So actually take it a step back from that. They are the ones who are a spotting their area inefficiencies with the way that they are already doing things, maybe not telling anybody or or thinking they're not going to get any purchase by telling any anybody else in the business, or maybe wanting to steal an advantage themselves. Who knows? But you know, that they're also coming up with ideas themselves, they're designing ideas for themselves to find solutions as well. So when we're talking about companies being reactive to AI and seeing things, actually they've got a really good resource potentially within their own business to spot these ideas, to spot ahead about where they can find these in you know, efficiency improvements through the use of AI. And if they start using them collectively, actually that becomes much more more powerful as well. Just Just an observation really from from what you were you were saying then. Blaine Mathieu: No, it's a hundred percent right, Karl. And I and I think, we're gonna talk more about AI impact on jobs and and that kind of thing, but I think the the people earlier in their careers, you might say, are they closer to being AI native than the senior executives, right, to some degree. And so to they probably they could very well be doing mini reimaginations of their own jobs without anybody even knowing it. Karl: Yeah. Blaine Mathieu: to some degree and and so they could be a should be, I think not could be but should be an incredible resource to tap when you're trying to think through, okay, now how could that impact us at the bigger picture? Karl: Yeah. Okay, well maybe that that maybe is one of the answers to my next question really, which is so if you walked into a consultancy, maybe it's got fifty to a hundred people in there, what what would you want to understand before you picked a single AI initiative? So I don't know, would it be processes, data, people, customer proposition or I don't know, something else? Blaine Mathieu: No, what I'd what I'd want to do is dive deeply into their technology stack. Just joking, Karl. That's that is not not what I would do. Okay. Did no, no, yeah. I'm just making that point. So hopefully your audience said, what the heck is he talking Karl: I I I thought this this does not sound like the answer I was expecting there. Absolutely, yeah. Yeah. Blaine Mathieu: about right now? No, that that is the opposite of of what we would want to do, obviously, right? What we want to do is do it, you know, have an overall look at the Operations, I call the operational logic, because it's it's not the tactics of the operations, it's the logic underneath those tactics. Okay. And Karl: Yeah, yeah. Blaine Mathieu: and try to identify which of those operations carry some level of volume of repeatability, right? that would make them good ones to good candidates for agentic AI implementation enablement. Where's the money actually flowing from it? Obviously, that that's really critical. I'd want to know who owns the outcome, you know, who are the owners. And this is why IT is generally not the place to be driving your agentic AI transformation, because they are probably not the ones who actually have their name or their career sitting on hitting a number. That's what's important, right? And what is that number that they're judged on? And and then, you know, fundamentally, what would have to be true for the partners in the organization to agree that this is the direction they need to go? And I wanna those that sounds like some kind of

big picture McKinsey analysis, but what it actually ends up being is start by identifying one particular operation or system you're running that has a number on it, has a clear owner, and that has some kind of volume or repeatability to it, because if it's entirely a custom one off every single time, then that that's harder. to autumn to it's definitely more likely to be a reimagination than a retrofit. neither neither is wrong. But identify, you know, where I always start with clients is start with one identifiable operation who that has an owner, that has a a number next to it, and then figure out how we can use a gentic AI to to position it somewhere on that scale. And and again, if you're not doing that proactively, then it's being done to you, right? So that that's that's where you start. Karl: Okay, so talk about obviously having guardrails with with AI and and such but how much process discipline is is enough before you automate? Is is there a danger of automating a mess, but also a danger of spending a year documenting every process while the opportunity moves on. How how do you find Blaine Mathieu: Yeah. Yeah. Karl: this this sensible middle ground? Blaine Mathieu: Well, and and this is directly related to the the comment I just made about picking one identifiable process flow outcome. You know, with most companies, it and it's probably virtually every company, you look at the when you step back and look at the big picture, you go, my god, the data is all over the place, and including in in people's heads. I call it in in my book, I describe it as the context, right, that you need to be able to give agenda KI systems to operate efficiently and effectively. And I've never seen an organization yet, big or small, that doesn't have some level of of challenge there, let let's just say. And again, the IT approach would be, okay, let's take two years, we'll do a deployment of a global data warehouse and we'll get all the data, you know, modules lined up and and figured out. And and then by then, you know, the game game is over. Right? How many times have we heard of a five-year implementation of an ERP system? which is fundamentally one way of keeping all your organizational data organized. and again, I'm not saying IT shouldn't be creating a global data warehouse of all the organized and tagged data, but to get started driving some value with a gentic AI, start smaller. And the other thing that AI is really good at, as most of us know, it's very good at. At handling fuzzy context, fuzzy data, fuzzy situations that normally only a human could possibly parse, right? Because pre-AI, these technologies were deterministic. If A and B, then C. And if A and B aren't lined up or they aren't in the same format, forget it. Okay. Start your building your data warehouse. Right. But now these systems can look, you know, A and B aren't in the same format. They aren't quite the same, but I can glean the meaning out of these things well enough in order to create C. That's a a technology capability that we've never had before. And so that I'm and again not arguing against having good data policies and practices, but but you have more degrees of freedom with AI that than you had before. And you remind me of another of another important element this this reality that a lot of organizational knowledge is held in People's heads. It's held in the mores and practices that are never written down in an organization. It's held in the structure of the org chart and the compensation system, right? A lot of what makes a company run isn't, you know, in some database somewhere. And two things. First, like I mentioned, AI is better at dealing with those kind of fuzzy levels of context than ever before. But the manager or the leader also has to decide. How much of that am I, how much of that context am I willing to share? Maybe I was holding on to some of that to make me feel important or needed, because of course this company couldn't run without the five things I've got running in my head right now. Well, if you want AI to be effective and successful for you, you have to be willing to share more of that with your systems and in a way you've never had to before. Karl: Yeah, I I I'm thinking really that that that problem, if it is a problem of of transferring that information that's in people's heads, will probably go away as more and more people are using AI anyway, won't it? So inevitably we're just going to be constantly talking to AI, using AI. AI is going to be making a lot of the decisions there. So is that fair to say? Do we think, the individual's knowledge is no longer going to be, how they are valued? anymore. It's it's more how they work with the Blaine Mathieu: Well, Karl: system. Blaine Mathieu: well, of course, knowledge and experience and and wisdom, I guess, ever to use a term I I think I talked about earlier, is is always going to be valuable. But I think it's it's an unlock that it's no longer only sitting in the heads of leaders like like me and like you and and like your clients, right? my major mode of of interacting with this system of agents that I've built that are running the back. office of my company right now is all spoken. I don't, I rarely type anything anymore. And and it's all stream of what I what you might call stream of consciousness, sort of in the same way that I would talk to a colleague or an employee that was sitting you know on the cubic lover from me, right? The way you used to have to work with technology back to the data thing and context was very structured. Right. Now you can It's very I find it very easy to just offload some context about an issue I'm thinking about into the system, just as if I had a colleague that I was talking to at the water cooler that you could never even imagine doing before. So so, you know, think of it as as a this technology as a powerful unlock to enable context sharing in a way that you could never, you know, could never do. Karl: Yeah,

very interesting. Just going back to maybe the spark for AI adoption or or even relating to the governance of it a as well or almost so who who should actually own the AI adoption? Is it one person from the top? Is this something more for the technology people? Probably not. Or is it individual par departments, groups of people, different levels? How how should it really be Set up, do you think? Blaine Mathieu: Well, and and building on what we've been discussing so far, there there I I don't think there should be really such a thing as an AI strategy. Okay. An AI strategy is a technology strategy. And and again, your IT people do need to be thinking through all these issues, open weight models, closed models, the what harness they're gonna use, all all of that stuff is all I'm not saying it's not important. It doesn't need to be figured out. Okay. But What first needs to be figure out, figured out by a business owner who owns a number, who has a clear objective, or maybe a number of clear objectives, but but can identify one objective. What do I need to accomplish? Okay. That that's where that's where it starts. And so the book and the methodology that that we use runs sort of at the, I describe it as the organizational level. So it's not, it's not only at the company level. This is not just something that the board and the CEO have to think about and deal with, although it could be, it could be at the company level, but it could also very well be the head of sales thinking through how to retrofit or reimagine her operations. Or it could be the marketing operations or it could be the operations that are, that are doing research and and doing the initial report generation in a in a domain. So that's those are different organizations. I describe them for want of a better word that run inside a company. And each of those leaders should be thinking about retrofit or reimagine in the context of their specific goals and outcomes. And then, Karl: Okay. Yeah, that's great. Blaine Mathieu: and then get IT to help you figure out how to deploy it. Okay. So I love IT, you know, good. Karl: Yeah, absolutely. The boring stuff. And obviously with AI, the next step is autonomy. So agency AI, that is to say systems that can pursue a goal and take a sequence of actions, maybe, rather than simply giving a single answer to it to a prompt. So y your book asks how far organizations should trust autonomous Systems, what what is the practical test for deciding what an AI agent may do by itself, what requires human approval and and what should maybe remain just purely human work? Blaine Mathieu: Yeah. I've got a chapter on autonomy and governance, and those are obviously related related issues. The the more autonomy you give your agentic systems, then the the more you have to be thinking about governance. And as I think I say in the book, autonomy without governance is reckless, but governance without autonomy is just theater, right? You're putting putting tools and systems in place. Why? Right. You're just you're just wasting Karl: Yeah, yeah. Blaine Mathieu: time. So you you need I think they definitely need to be thought of together. And that's how what what I describe. The I guess the thing, the fundamental way to I think about this is built-in versus bolt-on governance. And what's happen and and leaving AI aside, just think about how you run your organization, right? Is the governance you use to ensure that the people that are actually doing the work on a day-to-day basis. Is the governance simply a final check at the end of the work when something is done? Is it good or bad? If bad, send back. Okay. That's what I call bolt-on governance. And that's sort of tried and true. We or or an even might say worse version of bolt-on governance is the report or the piece of work gets out in the market and then the market rejects it. And then you go back, or the client, and then you go back and say, What happened? what happened in our process, right? That's that's bolt-on governance, and that's what companies have been doing for eons, Built-in governance is ensuring that by definition, you cannot produce that faulty output to begin with. Okay. You you have the internal checks, the guidelines, and in many cases it gets back to the context. These are very closely related. If your people, in forgetting AI again, if your people fundamentally understand the goals and the constraints and what you're trying to achieve with that client and what the market is expecting, the better they understand that that context, the less likely it is they're going to produce that bad thing at the end to be to begin with. Because the check for the bad thing is happening continuously in the minds of the creators while it's being created. It's not just being bolted on at the end. That process has always been an effective leadership strategy. Okay. Make sure many people understand the context for what they're doing, what our goal, what are we trying to achieve with this client? That will, of course, dramatically increase the odds that what we produce at the other end will achieve the needs of the client. Okay. AI is exactly the same way. The more context you can give your agents that are assisting with the creation of your work, whatever it is, with your operations. The less likely it is they're going to produce something out the back end that is wrong or invalid or inadequate or doesn't meet the need. So built-in versus bolt-on governance. I think again, going back to my point that this is not really an AI book. This is a book on corporate strategy. Think about leadership in general and how if you can build in the governance into your team, you'll increase the odds that you're going to get good stuff out the other end. Same thing with AI. Karl: Yeah, it makes sense, but I I guess, it's it's hard to sort of visualise where this goes, you know, when when AI is producing outputs at volume, maybe more or more complex sort of outputs maybe as

well, that maybe requires a manager to to inspect, you know, in in depth really to to make sure that it it is complying with what what a a client might expect. You know, I'm thinking in terms of the sector that I work in, it's land developments, if you don't get the reports right and something gets just not allowed to to proceed, then it may end up being a taken to an appeal where it is going to get cross examined. Blaine Mathieu: Mm-hmm. Karl: You are gonna get expert witnesses in. So I mean this this balance suppose with having agentic processes and a human in the loop aspect to it is just Does it become I don't know, is it is it is it more of a comforting phrase really, rather than a real control? What what what what should effective oversight look like Blaine Mathieu: Yeah. Karl: at this sort of scale? Blaine Mathieu: That's yeah, you I think you've you've touched on the the core challenge here is everybody's talking about human in the loop in these systems, human in the loop. All that really ends up meaning is meaning is there has to be some human I can blame somewhere for this, for this if it goes badly. Okay. That's the if you're the human in the loop in in one of these new processes, be very, be very afraid because what all that really means is you're the person who's gonna get the the crap if if it go if it goes south, right? And the challenge with these systems and and for putting humans in loops is they are now these these days so intelligent and they operate at such speed that if you've got a human whose job it is to review 400 reports or or analytics every week, at some point I can tell you what that human's gonna do. Yes, yes, yes, yes, yes, yes, right. So having a human in a loop probably isn't going to achieve the fundamental goal. And so this gets back, I think, to reimagining how your process actually works and what the and I think we're going to get into this soon. The the what is the job of a more junior analyst or the kind of person that used to write, sort of author these documents, these reports, these these proposals at the end of the day, what what do they do? And how does their job shift? in this world from sim from being the actual original author, but not just shift to being a human in the loop who's just looking at the stuff coming at him and going, Wow, yep, approve, approve, approve. And then eventually that that's not that's not gonna work either, right? Karl: Absolutely. I talking about the reports that I'm talking about, the the the writer of these reports is they know that they're going to be the person who's going to have to sit there in front of an e a a barrister or something and explain their reports. So they're taking the responsibility for it and they know they're taking the responsibility for it, so they know they have to trust what what they're they're getting out of the the AI system. So it it is a tricky one because Yeah, if you really need to test what is being produced, you have to put the time in. But yeah, maybe over time you Blaine Mathieu: Yeah, yeah. Karl: start to trust it more or you start to get the automation bias and you start thinking, Yes, it it's it it must know what it's talking about 'cause it makes it so logically, you know, plausible. Which, you know, if you've got the prof yeah. Blaine Mathieu: It it it sounds so good. It sounds so good. And the latest, the latest newest I you know, I've noticed even over the last six months as I've been writing books and collaborating with AI agents and this team of agents that I mentioned earlier to to run the business, I've noticed even in the last six months, the level of intelligence and competence, the the the writing skill has improved dramatically. And And I think many people who are accomplished creators of content are are to some degree having a bit of a crisis of confidence because they're seeing what Karl: Mm. Blaine Mathieu: is what is produced now by these systems. And again, if you haven't been really involved in it, even in the last six months, it's changed quite dramatically in in in my opinion. And so that's the speed back to the speed of the river flowing. That you need to be constantly rethinking what the right approach is, or all of a sudden your competitors will be drafting, you know, incredible output in record time. And back to your governance though. Now you have to figure out okay, now how do I maybe reimagine how the people that used to be the original authors of these things do their job or what they're doing to make sure that not only the volume and the speed. improve, but the quality stays as higher, maybe higher even than you had it before, right? Karl: Yeah, and I I'm I'm thinking out loud, but I I I think if if AI is providing good reasons and citations for where it is coming up with its its outputs, then you're speeding up the process of you being able to to proof it yourself anyway. And if you've got the knowledge in your head, you know what's where the banana skins are and and where the the the nuances in wording can can have an effect. So yeah, that that goes back to what we were saying, that that knowledge in your head is always going to to to be useful. I don't see AI sitting in front of a barrister and answering questions about a report just yet. It is going to definitely be a human, I'm sure, for while and you touched on something, obviously, which is a a constant is constantly raised as as a a concern really. It's it's you know People's perceptions, staff perceptions of AI coming into the workplace. if if AI genuinely removes parts of people's work, the the the phrase is it's only there to help you and help you do a better job, i it sort of can fall on deaf ears a little bit. What what have you learned about introducing major technology change without either frightening people unnecessarily or or pr pretending there are no consequences maybe? Blaine Mathieu: Well, my first bit of advice for any leader in any in any in the face of any kind of

constant and rapid change is be honest, right? And and I would say, and I touched on this earlier, engage with all levels of employees about what's going on, not just with the executive team or the owners or the the leaders, right? Because I my would bet dollars to donuts that the more junior People in your organization are more likely to be in touch with what's going on, with what's possible on the tech side, right? Or on in again, even if they're not technologists, I'm not talking about the IT people, but just the general day-to-day users of these technologies and systems. So they're probably, like we mentioned earlier, already doing some reimagination of their own work that you probably don't even know about, okay? Using Karl: Yeah, yeah. Blaine Mathieu: their own personal accounts, so it's not being tracked by IT. I can't tell you how many people I know are doing that right now, using these systems in a way that IT has no knowledge of on their personal devices, right? Because Karl: Yeah, yeah. Blaine Mathieu: they're afraid it's gonna get shut down or or that the leadership level will say, you're not allowed to use AI for this. So they say, okay, that's fine. I'll just I'll just use AI in a way you'll never find out, right? So so I think the the a key to success is to Involve the organization in figuring out what especially what a reimagined could look like. The in on the retrofit side, these often do tend to be characterized as reductions, as efficiencies, as we no longer need the the staff that we used to have to to write these proposals, okay? Because now they're now we can write them twice as fast or 10 times faster using a AI assist, right? And If all you're thinking about is a retrofit, then and and what you want to achieve is to downsize the company and reduce cost. That's I'm not to saying that's not a valid strategic choice, just like that's always been a choice for companies to do. Okay. But if what you actually want to do, a core concept in the book here, is to pull away from your competition, the organizations that are doing this, that are reimagining, are not fundamentally thinking about layoffs. You know, JP Morgan. Just took just reduced their know your customer onboarding process from five days to a few minutes. Right? They had 300 people in their global organization that was running that process. Did they lay those did they lay all those people off and say we saved a bunch of money? No. They took these knowledgeable, experienced people who understood customers really well and moved them into more customer-facing organi elements of the organization so they could increase their lead over the other banks. They didn't just lay off the 300 people. And I think it's a great, you know, case study and example of how you can do a reimagination that does improve efficiency from one perspective, but but actually enables you to pull away in the way you do it. Karl: Yeah. it might might sound a bit of a euphemism really if you're reimagining almost people's jobs in this whole process, but there's the realities to to contend with there, aren't there? And you can look at it positively or you can, look at it negatively. So I mean you touched Blaine Mathieu: Well look and look, the average knowledge worker doesn't want to be just doing drudge work, right? They want to be adding Karl: Absolutely. Blaine Mathieu: value, they want to have agency, they want to feel like they're having an impact. And so a reimagination can figure out how to use these technologies and systems to enable that person to have that impact, right? Versus just moving paper. Right. That's that those days Karl: Absolutely. Yeah. Blaine Mathieu: are rapidly coming to an end, but it doesn't mean knowledge work is coming to an end, far far from it. Right. In Silicon Valley here right now, which everybody feels, obviously is the the bastion of leading edge use of these technologies, there are not massive layoffs happening. There's a hiring boom happening right now in in Silicon Valley, even for software developers, which everybody predicted for a couple of years now was gonna be the the First example of a job that was going to be decimated by AI because it's fairly logical, and AI is good at writing software code. And yet, if you're a software developer in Silicon Valley right now, you are the happiest person. So the jobs are changing, and all knowledge work jobs will change, I think, dramatically over the next five to ten years. and and my personal opinion is sooner on that scale than later on that scale, but I And I I don't at all believe it's gonna be an apocalypse of of knowledge worker jobs. I think it it's for people that are open to reimagining their own role and how they operate, and then for the leaders that can make it happen, it's gonna be an incredible time. Karl: Very reassuring words. So you you touched on it earlier as well, really, but you know, the junior staff are the ones who may not realize it, but but they're the ones who have traditionally learned their craft from working really from the the nuts and bolts of a a a particular job all the way through you know, drafting reports for senior people to then, you know, obviously provide feedback on and you know, mentoring around right from that that that base sort of level, that square one. So if you know, if AI is taking more on on more of that that drafting, where where does the next generation of of senior judgement come from? Blaine Mathieu: Well, a lot that's you might call that the apprenticeship level of these organizations. You you do the initial research and the initial drafting, then a more senior person will review the draft and tell you where it's wrong or what you missed, and then maybe there's a finally a partner or a principal at the very top that's gonna put her name at at the bottom and and you that that's the that's the way it's always been done, right? And Karl: Absolutely. Blaine Mathieu: I think we need to re reimagine or

rethink How that apprenticeship works. And back to what we were just talking about, give a new challenge to the people instead of using the writing of things as the way they learn, they can instead be the ones whose job it is to manage the first phase writers, the agents, because a lot of that is being taken, taken over by the agents themselves, to manage those people. The more junior people can be the ones who. Do the checking, the confirmation, the the ensure the reliability of these systems, effectively up-leveling their jobs, and that's how they can learn, right? There's many, there's many ways to learn. And it and we gotta we gotta just accept the fact that a lot of these entry-level jobs in these companies has been just a mostly a training program because Karl: Yeah, yeah. Blaine Mathieu: they got half of it wrong anyway. Some more just senior person had to go in and rewrite it and red line it, and it was a C of red, right? So let's rethink how training actually happens in these organizations. Maybe training happens by getting in front of customers more at the beginning instead of spending the first five years sitting in an office right doing research and writing reports, right? There are many companies I'm hearing about now that are that are literally doing this. They're they're taking that lower level entry work that and and converting it into what they call a training program. Because that's what the entry-level work was. It was just training with a a different name and it was inefficient and it was poorly done because that's all the junior people could do. Now they're just saying, look, we've got a training program. Our goal is to make you super efficient at in this particular domain. You need to understand the customer needs. You need to understand the market. We're going to put you through a six-month program where you're actually in the market, learning X or Y, versus learning it by sitting in the basement or the first floor or cubicle. writing something when you don't know anything. that that d that doesn't make sense, but but it's the way that a lot of knowledge work has been done, whether it's legal profession, accounting profession, all of these. Start in the start in the basement, trying to draft something you don't know much about, and then eventually you'll get to the knowledge. Why not why not reimagine that? Karl: Yeah, I I think you're absolutely right. just going back to the the report writing aspects of of these things, these these are not these are reports that are probably not going to be ever written by people again. So why do you need to Blaine Mathieu: Yeah. Karl: learn how to write a report? You don't, do you? You need to understand the implications of what are in that report and spot the the the the the wrinkles in it and and the the nuances of of what's being said there and how they they could apply. Blaine Mathieu: Well and and think of their responsibility because if if you've got that junior level job just writing and there's gonna be two or three more levels of checks and confirmation above you, then yeah, I'll just write something, you know. But if you are the one responsible for doing the checking of that something else helped author, then it changes the mindset of of these more junior people also. Karl: Well, I think, they're they're obviously going to be using AI to to check these reports as well and have conversations with AI about Blaine Mathieu: Yeah, wow. Karl: these th so the learning speed is actually probably going to be accelerated anyway. I I've said it said it before. I wish I had access to AI tools, early on when I was at university Blaine Mathieu: Mm. Karl: or in my early early career, it it would have just been a godsend. So yeah, you can look at it two ways, can't you? And i in your Blaine Mathieu: Absolutely. Karl: forthcoming second book meet them on the water and th this turns the question, outward to customers and and markets in including a world in which customers own AI agents increasingly maybe discover, evaluate and buy things on the on their behalf. What what does that mean for a maybe a professional consultancy or or other business whose traditional growth model is still relies heavily on relationships, referrals and obviously websites and and and other proposals? Blaine Mathieu: Yeah, there's a an the the revolution that the first book talks about is your internal operations, your operational logic, retrofit, or you imagine those operations to make you more efficient or really navigate to a new goal altogether, right? And that's important and I think it's a good place to start. And it's already well down the path in many organizations. The next revolution that AI is enabling is what happens externally between your company and the market. Whether on and both on the sell side when you're trying to sell your solution, your service you're offering into the market, and on the buy side, when you're trying to buy goods and services from the market. And right now, you know, historically large enterprises have had these things called procurement departments, and they might be the bane of some of your clients. They are the professional buyers. You know, when somebody in a large company wants to buy a service or a product, they go to procurements, say we need a new X. And they will go out and do research and then interview sellers and create RFPs and you know all all and negotiate prices and all of this stuff. Well, imagine if every buyer effectively had their own procurement group, but that procurement group was their agents running on their behalf to assess the market. Look at look at the offerings, negotiate offers and and price and and recommend something to you to finally sign the PO for or sign the credit card slip for, right? That's where the market is going, using agents to intermediate the buying and selling of goods and services. And as a little microcosm, for example, in the professional services and consulting that

that the River Group provides, it's not it's not visible to humans. But my website has an entire mirror website that's just built for AI agents. Because I want AI agents to be able to, that my clients are increasingly starting to use themselves, like we've just been talking about. When they go out into the world to scan the web and look for relevant services and offerings, they're going to see the agentic version of the website I have and understand it and be able to parse it very. easily and much more quickly than they could if they're trying to deconstruct this human website, right? Full Karl: Mm. Blaine Mathieu: of all the stuff they can't read very well. So that's just one is my personal example of how these things are going to change. And I think the market over it's still early days, right? We're still not saying to your agent, go out and, you know, buy me a professional level report on you know, on this particular topic and get back to me when you're done. We're we're not there yet. But we're starting to make steps in that direction. And that's fundamentally what the second book is going to talk about. I intend to release it later this year. Is first one is about internal operational logic. The second one is about how you face your market, both as a buyer and a seller using Agente KI. Karl: Okay, look forward to that. So let's let's finish this main conversation by looking ahead if again. So keeping our feet on the ground maybe rather than disappearing into predictions about machines taking over civilization. What over the next Blaine Mathieu: Yeah. Karl: two or three years, what what what change in how ordinary businesses operate do you think will maybe surprise leaders who are currently treating AI mainly as a I don't know, maybe a better chat bot or a productivity bolt on tool? Blaine Mathieu: Right, right. Well, you know, the the factor that's everything we've been talking about is the future for most of the people where that are listening, I'm I'm sure, right. And so we've we're we're projecting forward from where we are today now. I I think what we're going to see is the pace of technology change is not going the the flow of the river, the currents that back to how where we're where we started are not going to slow down. They are actually, we are just at the beginning. They are increasing in speed. And like I mentioned earlier, the fact that the capability of AI to be a competent author and writer has dramatically improved even in just the last six months. And we see that across all domains and even beyond core knowledge work, but into physical systems, robotics, all all of these, these areas are really at the at the early nascent. stage. So again, this is why setting in place the process, the ability to watch, the watching I mentioned, what's happening with your organization and its surrounding context and figure out how to respond to it is more critical than ever before. Because if you don't have that kind of system and practice set up now, imagine when things are moving five or 10 times faster than they are today. And that's and that's what's coming. So what's holding us back from all this change is not the technology. As I said, in most cases, the tech is already beyond, way beyond what most organizations are using. It's the ability of the organization to absorb and take advantage of that pace of change. And the differentiation, and that's what the two lines on the cover mean. The blue line at the top, those are the organizations that have figured out how to thrive and to accommodate that level of change. The ones, the yellow line at the bottom are the ones that haven't. And I I think probably, yeah, I guess I guess the line does actually point downward because I I think that is actually true. It's you know, the the ones that don't figure out how to deal with this are are not just gonna it's not just about not keeping up with with the the ones that do, but but it will adversely affect your organization outcomes. And so it sounds it sounds scary, but it doesn't have to be, right? The The system, the methodology, the frameworks to think about this are all there. In fact, I published them all in in the book. It's public, right? There's no there's no secrets. But of course, doing it is can always be a challenge, implementing it in your own company. That's why I started the River Group to help companies do exactly. Karl: Yeah. You just reminded me of Formula One there. I know you like your Formula One as I as I do as well. Blaine Mathieu: Yep. Karl: And it's always like if you're not moving forward, you're moving backwards. So it's looks a little bit like that on your book, I think there. Blaine Mathieu: Absolutely. And they're on the leading edge of of a great example. They're not using a lot of generative AI in Formula One yet, not for the cars themselves. But the it's a great example, as you said, of organizations that embrace whatever the latest tech is are able to pull ahead and and yeah, gr great analogy, I think. Maybe Karl: Yeah, yeah. Blaine Mathieu: I'll use that one in my next book is I'll use a Formula One analogy. I like that. I like that idea. Karl: I'd love that, yeah, great. So just to finish off the podcast then I've got a couple of quick fire questions, if that's okay with you. So what is one AI habit or practice you think most leadership teams should stop doing? Blaine Mathieu: Well, at the risk of just repeating myself, I would say stop thinking about AI as a technology challenge. Karl: Yeah. Yeah. Blaine Mathieu: Think about it as a strategic opportunity. The technology is enabling, but you can this is why the book, as I keep saying, is not a book about AI. It's a book about strategy and constant and managing constant change. And AI is just the thing that's causing that constant change. So, you know. That that's the the most important takeaway. If your listeners can take that away, they'll have a victory, I think. Karl: Yeah. And I love that that phrase

that you use there that AI is really the enabler in all of this and that's exactly what it is, I think, there. So okay, what is one question that about AI that every CEO or D should be asking their their management scene this year, do you think? Blaine Mathieu: Well, and then what I'm going to say is don't ask a question about AI. Ask a question about your business. And the question you should ask is what is just one operation that we run that meets some of the criteria we talked about before. It's it's a lot of repeatability, it's it's got some volume to it, that that we think we have a hypothesis that these technologies that we're talking about could have an impact on. And then let's let's dive into that. Let's assess that. Let's see. And if it's not that one element of your operation, then look at another element of your operation. Don't try to think through how will this impact my company? That's too big. Okay. Pick, pick one thing and start there. Karl: Yeah, so you you've managed to avoid my two trick questions there, so so well done. So and so finally I I ask all my guests this question. I I don't know you've written two books almost two books if if if we count the new one coming up. And we've talked about it a little bit, but you can talk about those again if you want to put them in in a cleaner sort of context. But otherwise what is one or two or the must read books that you highly recommend that I and our audience read? Could be anything. Blaine Mathieu: Well, I'm gonna give you something to su might feel from left field, although it's it is actually closely related. And your UK audience will find this quite interesting and relevant. Blood blood in the machine, sometimes it's called blood on the machine, but it's blood in the machine. Brian Merchant is the is the author, just though you can spell it just the way it sounds. And it's about the Luddite revolution and what happened in in England. back when textile mills and and cotton weaving was all being overtaken by machines. Really interesting book and so many so much relevance to the place we're in right now because we know there is some backlash, you know, brewing about AI and AI data centers and all of that. And the big mistake that that I I did as well before I read the book was I thought Luddites were anti technology. People. They were not anti-technology. What they were were experts in their craft and their expertise and their position in society was not accounted for as part of this revolution that was happening. They were rapidly shunted aside and pushed into poverty almost almost overnight. And this gets back to what we were talking about before, about how you involve the people, especially the newer people in your organization, in this reimagination. If these folks had been involved, I bet you, you know, England would have retained its leadership in this industry far longer than it actually did. But that's not what happened. So if you want to understand the true story of Luddites, and I guarantee you'll see many parallels to what's happening right now in AI, blood in the machine, Brian Merchant, I highly recommend it. Karl: That sounds fascinating and hopefully we'll stop history repeating itself there. So yeah Blaine Mathieu: Exactly. Exactly. Karl: Blaine, thank you for for sharing not only your journey but your a very practical way of thinking about a period of change that otherwise could be some dominated by tools, technology, terminology and and noise. So I've really enjoyed this conversation and I'm sure our audience will take plenty from it as well. So it's a bit a pleasure to have you on the podcast. So thank you for joining me, Blaine Matthew. Blaine Mathieu: Thanks, Karl. It was fun. Karl: Brilliant, and for our audience you can find Blaine and his monthly briefing on what's actually moving in agentic AI at therivergroup.ai. And if you enjoyed this episode, please follow myself and aligned mindsai on LinkedIn for practical AI tips and updates on future episodes of Aligned on AI. I'll share the details of today's podcast and guest on LinkedIn and it will be on YouTube, Spotify, and Apple Podcasts. But please do subscribe to the Aligned on AI channel. on your favourite podcast app for more content like this.