Ep 144: Redesigning Work and Accelerating Responsible Innovation with Michael Davies

[00:00:09] [SPEAKER_01] Hey everyone, it's Bob. Welcome back to Elevate Your AIQ, your go-to source for insightful conversations on human-centric AI readiness, talent transformation, responsible innovation, and the future of work. This week, I'm joined by Michael Davies. He's the founder and chairman of Endeavor Partners and a senior lecturer at MIT, an engineer and lifelong learner who has been thinking about AI's economic impact for longer than most of us.

[00:00:33] [SPEAKER_01] We talk about why so many organizations aren't getting the payoff from AI they expected, and why the real value comes from rethinking the work itself rather than shaving minutes off your tasks. Michael also explains why responsible AI has to be built into how systems are designed and run, what decades of making air travel safer can teach us, and why the key is to free up our thinking rather than hand it over. Michael has a ton of great advice for your AI journey, so I know you'll appreciate this conversation.

[00:01:00] [SPEAKER_01] If you enjoy this show, please remember to rate it on your platform of choice and share it with your personal or professional networks. Thank you very much, and thank you for being part of the Elevate Your AIQ community. Let's go talk to Michael. Hey everyone, welcome or welcome back to Elevate Your AIQ. This is Bob Pulver, and I am excited for my conversation with Mr. Michael Davies. How are you today, Michael?

[00:01:23] [SPEAKER_00] I'm good. The sun's shining.

[00:01:25] [SPEAKER_01] Excellent. Excellent. Yeah, here too, here in New York. So, Michael, you and I have been trying to do this for a while now, and things are moving so fast. I mean, I hope everything that we plan on talking about is still relevant, which I'm sure it is, because there's so much going on in the responsible AI space and responsible innovation, which I know we're going to dig into. But why don't we just start with, you know, giving my listeners a little bit about your background, your career arc.

[00:01:55] [SPEAKER_01] You've done a lot of things in a lot of different disciplines that sort of crisscross, you know, where you are now. So, I'd love to hear about it.

[00:02:02] [SPEAKER_00] Well, there's a lot of different things, but there is a sort of, I think what the Germans call in design, a red thread that runs through them. So, originally, I was going to go to Cambridge and study math, but I did something which I encouraged other people to do, which was why. So, I took a year off in between, came back and said, you know what? I think I'll be an engineer. And as the beginning of the journey, pretty early on, I got really interested in what we could do with artificial intelligence and robotics.

[00:02:31] [SPEAKER_00] So, that led to me spending the first decade or so of my career basically using what was then the state of the art. An opportunity came up. I'm a New Zealander. The weird accent is a combination of English-educated New Zealander who's lived in America. An opportunity came up to move back to New Zealand. I thought I'd be working on cows and trees, which is what a lot of the New Zealand economy was about. But it was just as the regulations happening in telecoms.

[00:02:59] [SPEAKER_00] And that led to taking a lot of my sort of engineering knowledge and moving into strategy telecoms. A decade doing that. And then an opportunity came up in the U.S. to look at this from a more academic point of view. I got very interested in this idea of business ecosystems because we've been building a lot of partnerships and things like that. That was one of my responsibilities. That led me to moving back to this part of the world.

[00:03:24] [SPEAKER_00] In fact, the initial office that I walked to, I walked past most days of the week because it's in Harvard Square. And I live just by Harvard Square. And every time I look past this, it says, well, yeah, 50 churches. And I stayed current in the world of AI. And about 10 years ago, a lot of 10 years ago, one of my grad students at MIT said, hey, I've got someone who's really interested in figuring out whether or not this AI thing is a big deal.

[00:03:53] [SPEAKER_00] Now I'd seen what was being written about and published. And I went, it is. It's going to be a really big deal. And they went, well, how do you know? And I said, we've started talking to AI. Could you help us work through what the implications were? And that led to some of the earliest work. I think we may have made one of the very first presentations at border level. And I pulled the slides up to open it with the client's permission. It was Thomson Reuters, which says anything that can be automated will be. And we were expanding the scope.

[00:04:22] [SPEAKER_00] So then I hadn't quite foreseen how far it would go. We were more radical about this. But it had a huge impact, that work, because we said, don't outsource to India. Because they had a big outsourcing initiative underway. We said, by the time you do that, you'll be able to automate what you're going to go do or do it using tools that help people to do it, which is actually how it played out. So started to see the economic impact playing out.

[00:04:52] [SPEAKER_00] And I've been working on that since, teaching about it, practicing it. A few years ago, I got really concerned about the lip service that was being paid to it. People aren't, you know, creating an oversight board never works. Sorry. That got me really annoyed. I'm an engineer. I'm an engineer, and I did this within a decade working on manufacturing engineer. If you want something to be designed to be manufactured, it's got to be embedded in the processes.

[00:05:21] [SPEAKER_00] And that sort of brought us to where I am today, trying to accelerate the process of innovation, and particularly in digital technologies and digital and real world technologies. And trying at the same time to make sure that what we're doing is responsible and ethical and humane and sustainable. So that's the art that got me here.

[00:05:46] [SPEAKER_01] I think a lot of what you described is, well, not only has a lot of that come to fruition, but I feel like it's a lot of the areas where people are really struggling because they don't necessarily have the right mindset or the sort of systems thinking to say, how do all these pieces fit together? Whether just even thinking about, you know, things not just on the software side, but on the hardware side, the, you know, industry ecosystems.

[00:06:13] [SPEAKER_01] And as you look across, you know, the lifecycle of some of these, you know, workflows and different domains, whether that's telecom or HR and the talent lifecycle, which I know we're going to talk about a little bit. But you're also hitting on a lot of the areas that I'm passionate about, both on the being responsible by design kind of core, you know, principle, right?

[00:06:35] [SPEAKER_01] If we're all able to build things now, let's make sure we build them properly and have the right, you know, guardrails and ethics in place and things like that. And certainly accountability. And then the innovation side, I think people are really struggling with, you know, where am I on this, on my own sort of adoption curve and maturity curve as they try to bring AI into the enterprise.

[00:07:01] [SPEAKER_01] I was just on a call earlier today with an analytics firm based in the UK. And they were saying most of their clients are just finding individual use cases, experimentation kind of phase. And it wasn't until like a Horizon 2 kind of outlook are they looking at actually transforming, you know, roles in the work itself. I'm like, guys, I don't know. I don't know what you're advising, but like this, they're crawling. They're crawling while everyone else is at least walking.

[00:07:30] [SPEAKER_01] And like, that's not going to cut it.

[00:07:33] [SPEAKER_00] I even offered a discussion with that earlier. And people were saying, I was being asked, why are we seeing this sharpening statistics? And why are not people getting it out? I said, well, because what people are doing is taking the lowest hanging fruit. And that lowest hanging fruit is shaded by the opportunities, if I'm going to talk to the analogy. Piecemeal, save some time, save some money stuff at the task level is not where the real value lies.

[00:08:01] [SPEAKER_00] It comes about when you rethink what you're doing and do reinvent it. Now, we know that from prior history. But boy, am I seeing that pay off. And if you do automate something at the task level, unless you actually make good use of people's time that you freed up, there's not necessarily a huge gain from this. I remember back when we were doing our original work a decade or so ago, we started looking at the impact.

[00:08:27] [SPEAKER_00] We saw that we were very concerned about what the impact would be on employment. It was not catastrophic. And the reason was there were a whole variety of useful things you could get people to go do instead. So in fact, the quality of that often went up. Maybe the quantity. And that, to some extent, is happening. So I think there's two parts to it. One is we're not creating new value enough yet. That's a central theme of it.

[00:08:53] [SPEAKER_00] And the second one is when you get what economists call hedonic gains, the quality of something goes up. The interactions are better. They don't necessarily sharpen the statistics right away. Not until you rebase the measures. So the fact that, for instance, it's a much better customer service experience and people are happier and people are better doesn't necessarily shock very clearly in the GDP statistics.

[00:09:21] [SPEAKER_01] I know you've got a couple ventures. And so just talk to me a little bit more about Endeavor Partners and your sort of philosophy as you look at investments and partnerships and things like that. Just some of your core principles. They seem to obviously align to everything that we're talking about. And I'm just curious about your due diligence.

[00:09:48] [SPEAKER_00] Well, motivation came about was you can build big consulting firms which scale and do routine stuff. We're interested in innovation. And the big impact we have had has been where we've helped people create significant new value to build products that would not have otherwise been built. And that meant that we're a little bit different. We did actually some competitive assessment in the last several weeks.

[00:10:14] [SPEAKER_00] And a combination of we really can get deep on the technology, but we think about it through a connection to business. We're doing it in a way in which we're really very selective about who we work with. We've chosen to basically always try and work for the guys in white hats. The people actually build stuff. So candidly, we don't work for patterned folks. And there are companies whose business has been solicitous to work for them and we will turn them dead off.

[00:10:43] [SPEAKER_00] We're interested in actually trying to make good things happen. We have a couple of pro bono projects underway at the moment. One of them trying to transform single family style homes to be able to build better homes that are more robust for a fraction of the cost. Like get past the current approaches. We're also working, some of you may know about this. There's a big shakeout happening in public media.

[00:11:09] [SPEAKER_00] We're working with success, if you like, to NPR, which is, hey, where our podcast is actually podcasting is going to demolish broadcast radio. And what the next generation public media version of that looks like. So we try and do the, not just innovate rapidly, but do it in ways that are responsible and ethical. Sometimes it means that we have to be a little bit low profile. So people go, we haven't seen much of you.

[00:11:39] [SPEAKER_00] I go, well, I was an expert in the Google antitrust trial. And as part of that, everything I wanted to say probably had to be cleared for the course of four or five years. And we took that on and said, wow, it's going to be four or five years out of my life. Big antitrust cases take a while.

[00:11:57] [SPEAKER_01] Yeah. I mean, I think I take a similar lens to, you know, my business relationships as well. I mean, whether that's on the transformation side, you know, I'm not going to drag you kicking and screaming. I want you to have already come to the realization that you need to do things differently.

[00:12:19] [SPEAKER_01] And you're just not sure how to do that, how to approach that, as opposed to people that where I have to just convince you that this is not going away, that this is like the very nature of work, what work means and the meaning of work to the, you know, putting that human centric lens on it. What this means to people and their lives and livelihoods.

[00:12:46] [SPEAKER_01] You know, if you don't, if you still don't get it, if you're still, you know, bogged down in the fear of it, then I just, I'm not sure quite how to help you. I don't want, that's not the, that's not the role that I want to, I want to play. Right. And I think.

[00:13:02] [SPEAKER_00] I'm going back to a guy you may have heard of. You know, we're all familiar with the innovators, early adopters. That's, that's from a guy called Everett Rogers. He talks about the diffusion of innovations. And he, it's like, he actually did his work, which remains robust to stay in the thirties in farming. There are some people who will embrace something new without any references. They're the innovators. The early adopters, they'll follow right after them. And he talks about a couple of ways in which things get adopted.

[00:13:31] [SPEAKER_00] One of them is what's called epidemic adoption. Epidemic adoption is, you don't know you want it until you've got it, but it spreads like a disease. And that's what I think to some extent is happening here. Unless you've tried it, you've thought carefully about how you use it. You've taken a step back and forth. Let's all that. You, uh, you don't understand how powerful a useful it can be. So there's a lot of resistance.

[00:13:58] [SPEAKER_00] That fortunately is coming down, but I've been in some other places recently. I won't embarrass them by naming the countries, but we're looking at levels of usage and adoption at a 50 percentage points behind the United States. And then again, there are some places that are ahead of the United States.

[00:14:15] [SPEAKER_01] When you think back over the last couple of years, you know, as I guess the, those early adopters, you know, started making waves, right? I should, sorry, the innovators, right? The people who are at the forefront. And are they, was it just a sort of a risk tolerance difference?

[00:14:39] [SPEAKER_01] Do they have, is it about their, their foresight or were they managing innovation differently or some combination? I'm just trying to get my bearings when you saw things start to change perhaps.

[00:14:52] [SPEAKER_00] Yeah. It was typically not encumbrance. It was suddenly confronting a new problem and saying, oh, I need to solve this. I mean, one of the most significant examples for me was a friend and collaborator in a couple of ventures, a CTO called Matt Hodgson. And his problem was building a company that helped people make sense of this aspect. And it was suddenly, we got a process to make sense of a hundred thousand contracts a year.

[00:15:22] [SPEAKER_00] And these contracts have deliberately been written to try and obscure how much you're going to pay. Now they'd started out as a manual business and were looking at scaling that. It was like, this is to be terrible. And he was one of the very, he's the, I think one of the best CTOs on the planet. But then I would say that, but he thought creatively about, hey, if they can do this, couldn't I adapt that?

[00:15:51] [SPEAKER_00] And did, change the economics of the company. And that was one of a couple of things that made me start, you know, oh. The other one was a presentation at an MIT, a Jamil Clinic event where somebody was showing that you could do one shot screening from one X-ray, chest X-ray for some cell lung cancer.

[00:16:14] [SPEAKER_01] I mean, as you look at the market, I mean, I don't want to get too deep into like the economics of it, but, but as you see what's happening with the pace of change, especially obviously at the frontier level with some of the LLMs. And that can obviously have an impact on, you know, how quickly people can embrace that,

[00:16:42] [SPEAKER_01] how quickly they can put, you know, a governance layer around it, which, you know, is sort of trailing innovation as it always does. And just as you, as you look out at this space, I mean, do you see, what do you see as the sort of next wave of, you know, responsible innovation? I mean, are we going to hold each other accountable? Are we going to wait for politics to sort of calm down?

[00:17:11] [SPEAKER_00] I'm on record as saying this, that in general, we can't rely on the politicians. I wish we could. There are some exceptions to that, but in general, there's a couple of issues there. One is politicians' incentives are not necessarily aligned with ours. Most of the time, they're concerned first with staying in power one way or another, whether it's a democracy or any other sort of polity. So that's the first problem we have to coerce. There are some exceptions to that.

[00:17:39] [SPEAKER_00] That's small, relatively culturally homogeneous countries, which are highly democratic and not very corrupt, like Finland, New Zealand. The second issue is the capabilities. Even for those who work in the field, it's really hard to make robust predictions about what will happen. That means it's even harder to figure out what the adverse consequences might be, which means it's even harder to figure out how you might stop those adverse consequences, which means

[00:18:08] [SPEAKER_00] it's even harder to figure out what the second order effects of that would be. So there's a capabilities question. And in particular, it's not just whether or not the capabilities exist, but can you connect it with the people who are making the decisions? The UK, for instance, happens to be doing a pretty good job at that because there's a sort of concentration in those connections. And then the third problem is just the pace at which most polities work.

[00:18:35] [SPEAKER_00] It takes an awful long while to pass legislation and get it adopted. My classic example for this is the AI Act in the EU. The debate was taking place around it in 2019 or 2020, and it was mostly about surveillance and data privacy and face recognition technology. By the time it actually became law, it was mostly relevant.

[00:19:01] [SPEAKER_00] Now, they're embarking on another effort to do this. It is laudable in terms of its goals, but boy, is it moving an order of magnitude too slowly. And the connection between the policymakers who are distributed across Brussels and Rome and Bonn and the people actually know how the technology works is tenuous at best. Don't get me wrong. I have good friends involved. They're fighting the good fight. But boy, is that an uphill struggle.

[00:19:30] [SPEAKER_01] Yeah. I mean, I think that there's a lot of legislation here in the U.S., I mean, at a state level, sometimes even at a municipality level where, you know, they're sort of at least trying to get at some of the use cases that we haven't really, you know, tackled. Because I think there's a lot more than just what's happening with the frontier models and these existential, you know, threats.

[00:19:54] [SPEAKER_00] It's about the uses to which you put. Put it. It's not about the models inherently themselves. It's about the systems. So this is brainless systems perspective. The thing that inspired me as a basis from the MIT course was, hey, you know what? We made commercial jets flying a lot safer. It was really scary. Well, we didn't know better. But by today's standards, it was really scary.

[00:20:22] [SPEAKER_00] Depending on how exactly you measure it, it's like 40,000 or 50,000 times safer nowadays than it was in those days. Notwithstanding the complex systems failure of the tragic recent crash in D.C. There's a very instructive lesson there. But we managed to make this so safe that we had gone more than a decade without a significant incident in the United States.

[00:20:50] [SPEAKER_00] It doesn't take us 50 years in this case. We have experience in knowing how to do this. We can and should acquire it.

[00:20:56] [SPEAKER_01] I mean, I think some of this ties to the thresholds or the line that we hold AI to, the standard that we hold AI to in terms of accuracy, you know, bias mitigation. I mean, there's almost no tolerance for any bias exhibited in AI systems, which, of course, the bias is not in the AI itself.

[00:21:23] [SPEAKER_01] It's in what, you know, where the data came from and those were, you know, that data is based on human decisions. So, you know, but whether it's self-driving cars or some, you know, talent, you know, decisions and things like that. I mean, people just have, they just assume that the AI is going to be more biased than a human, which is...

[00:21:44] [SPEAKER_00] You've got to benchmark it against what currently exists. And the second is, there's a phenomenon, it's called blame discount. Like, I love the AI when it's working right. But when it fails, I'm going to go after the human being. The reality is, there's always got to be a human being in the world. You can automate up to a certain point, but ultimately, we are accountable. We have to be responsible.

[00:22:11] [SPEAKER_00] But if I pick up the self-driving car one, for example, if our benchmark is human drivers, and the systems are much safer than they are, then there's a pretty strong case we made to allow that very widespread adoption. The numbers here are something like, depending on how you're in fact here, you count it way more in a like-for-like situation. Not against some ancient clunker of a car, but against a modern car with lane departure

[00:22:40] [SPEAKER_00] and automated braking and blind spot warning and all of these things. So, state-of-the-art car versus Waymo, 10 to 15 times faster. 10 to 15 times safer. Well, slip the tongue there. And at that point, you give me the sort of choice between something that makes headlines because, for instance, a very unfortunate incident involving a cat in San Francisco or a bee thing versus lives being saved.

[00:23:09] [SPEAKER_00] I'm going to vote for that one. For me, forgive my cynicism, but at the point at which the people who are aggressively trying to stop Waymo being rolled out are the trial lawyers of America because they're not going to have as many personal injury suits. That says to me, we should probably be having it at it.

[00:23:30] [SPEAKER_01] Yeah. I mean, in the talent space, if we pivot over there for a moment, you know, certainly, you know, that's one of the, these are high-risk decisions according to the EU AI Act based on their sort of risk, you know, pyramid. And I think a lot of the U.S. laws are taking a similar approach, not as comprehensive perhaps. And I don't think we have, you know, a heat map for risk across these cases, which is part of the problem.

[00:23:59] [SPEAKER_01] And, but I think just like in the self-driving car situation, I mean, someone's got to be accountable. And I think on the driving situation, I think people have a knee-jerk reaction when, reaction when something happens because they're, they're just looking around like, oh my God, whose fault is this? Right. It should be the case if there was no, you know, autonomous settings at all, right? Like, oh wait, you know, who was, who was speeding?

[00:24:28] [SPEAKER_01] Who went through that light too fast? Who, you know, but in the talent decision, you can have, you have all the explainability there. That's the whole point of, of being responsible by design when you build this thing. And having the literacy and the readiness to use these technologies properly, someone is accountable. And some human being is accountable. And whether that's a leader at the organization or it's an individual contributor who is now

[00:24:56] [SPEAKER_01] a manager of agents, someone is accountable. And so it goes back to systems and structure and some of the, you know, the accountability sort of matrix within that decision. And you can correct it because you have the traceability and the observability of how the evidence contributed to its conclusions. And we just can't do that with human beings, not at scale.

[00:25:22] [SPEAKER_00] And I want to pick one together. These are complex socio-technical systems. It's that it's not about the moral. It's not even about the harnesses that you put around it, the safeguards. It's about the people are part of the system, including the designers. So pick up one of such in Adela's things about, we need to make sure that it's not just about the intelligence, but it's about the sight of the intelligence. I'm with that a hundred percent.

[00:25:48] [SPEAKER_00] The other thing, the flip side of that, which I would always try to reconcile is when we design systems, complex systems, sometimes just patterns emerge. The patrimonial crash, it's very hard to point to a single person who is ultimately responsible. There are so many people who contributed to that. That doesn't mean that we should remove people's sense of responsibility, but when we're trying

[00:26:16] [SPEAKER_00] to figure out why things failed, we don't necessarily want to prosecute people. We want to inquire and improve. So I'm an advocate. This is a failure. Don't fire the people. Learn from the lessons. Improve the system rather than mating about prosecuting people necessarily. And we know this from the in jurisdictions in which the response to a crash is to lock people up. You have more crashes. Why?

[00:26:45] [SPEAKER_00] Because nobody wants to own up.

[00:26:47] [SPEAKER_01] Yeah. Interesting. So when you want to ask you about your classes you're teaching at MIT, is this a breakthrough innovation?

[00:26:56] [SPEAKER_00] Two classes at MIT. One is about how you harness AI. It's a breakthrough innovation. Breakthrough innovation is new products that create a great deal of new value using existing technologies. Much of what MIT does has come up with really cool new technology, but that takes a long while to come to fruition. My classic example of breakthrough innovation is the iPhone. There was no new technology in the iPhone. Put pieces together.

[00:27:23] [SPEAKER_00] We're actually, this weekend, I'm spending some time with Blake Telly, with whom I co-teach this course, because the next iteration of it, we're going to include some of the very latest developments, which we're seeing in terms of how people are using state-of-the-art LLMs to accelerate the process. To accelerate the process. So it's not just about the tools. It's not just about putting AI into the product.

[00:27:49] [SPEAKER_00] It's about changing how fast you iterate, changing how widely you search, changing how well you evaluate a complex thing by augmenting humans. So this is an automator. Yes, it does involve multi-agent swarms. Does it? It also involves recursive self-improver, but it's human-directed, highly humane, and augmenting human capability.

[00:28:16] [SPEAKER_00] So the other one, again, systems thinking, applying the systems engineering disciplines that came out of MIT AeroAstro. When you think about complex sociotechnical systems that came out of the engineering systems division at the joint SLOM engineering and saying, the only way we're going to make these things really robust, safe, and responsible is bringing that sort of broad perspective to

[00:28:44] [SPEAKER_00] bear using those sorts of tools. We know how to do it. We just need to go apply.

[00:28:50] [SPEAKER_01] Yeah, I mean, there's probably no better place to have that type of interdisciplinary exploration than in Cambridge. So that totally makes sense.

[00:29:00] [SPEAKER_00] So it's one of the joys of MIT is people go, I remember once scaring a Marine General. I was teaching an innovation workshop to the US Armed Forces in Europe, and I scared a Marine General by my answers a question about who's in charge at MIT. And I was, good question. It's not really a word we use. Because MIT, unlike our neighbor down the street, is deeply decentralized, which means

[00:29:25] [SPEAKER_00] that when new stuff happens, you can't stop people building these multidisciplinary entities. So my other fun thing is our office is right by MITT store. And when people come out and go, where is the computing science department? I go, that's a good question. Which computing science department? And they go, well, there must be a computing science department. Oh, there's several. What? Very, there is interdisciplinary players.

[00:29:56] [SPEAKER_00] C-Sale, E-CS.

[00:29:59] [SPEAKER_01] They must not be local Bostonians.

[00:30:02] [SPEAKER_00] No, no. It's usually they're looking around disoriented and speaking in a foreign act.

[00:30:07] [SPEAKER_01] So, you know, one of the things that you mentioned and sort of ties to this interdisciplinary kind of concept is around like cognitive diversity and collective intelligence and how that can be, you know, either a spark or can actually be the genesis of some of that, you know, breakthrough innovation.

[00:30:29] [SPEAKER_01] And innovation, because if you don't have people with either nonlinear careers like you and I and exposure to some of those other disciplines, you don't necessarily have a full, you can't necessarily connect the dots that others might connect, right? And so maybe that was, I don't know the full history of the iPhone, but maybe that was part of it.

[00:30:53] [SPEAKER_01] But as we think about these decisions, certainly AI, a well-trained, you know, AI can be part of a, you know, sort of collaborative, you know, body. But certainly, you know, the foundational concept of collective intelligence is collective human intelligence, which I learned from some of your colleagues like Professor Tom Malone and Andrew McAfee and others.

[00:31:19] [SPEAKER_01] And so, you know, those are some of the things that I sort of, you know, think about that get my, you know, creative juices, you know, flowing and all those things that I've sort of looked at as sort of things I've learned inside projects and things like that over the course of my career. I feel like they're all coming, you know, just sort of resurfacing in the age of AI for all these reasons.

[00:31:43] [SPEAKER_00] Yeah, I'd agree 100%. And I think there's sort of a challenge here. So I think it says this on my LinkedIn profile. I describe myself as a polymath, like variety disciplines, and an opsimath, a word I only came across a couple of years ago, which means somebody who's a lifelong learner and, you know, takes on new challenges.

[00:32:07] [SPEAKER_00] I'm late in my career, but I just started doing work as a doctoral researcher on SDA Bocconi in Italy because of my real interest in how this is going to play out, you know, trying to get mastery of it.

[00:32:21] [SPEAKER_01] And because you're not busy enough, right, Michael? You're just not busy enough and you need another hat to wear.

[00:32:25] [SPEAKER_00] Yeah, but I have to go to Milan. It's, you know, it's a hardship posting. I think that one of the really interesting phenomena that I'm trying to attract very closely is we are seeing in some cases, unfortunately, what we call cognitive surrender. It's like, oh, I'm going to stop thinking about this. I'm going to avoid this. It came up in, you know, in a class I was in as a student where people going, why do I have to know how to do statistics?

[00:32:53] [SPEAKER_00] Why don't I just get to ask Claude or OpenAI about it? And the point was made codently was if you don't have enough basic understanding, then you're not going to be able to figure out whether or not it's answering the right question for you. And that's really important in a whole variety of disciplines. There was a conversation earlier today with somebody who works in the software engineering discipline. If you want something complex and difficult done in software engineering,

[00:33:23] [SPEAKER_00] do these? Somebody actually understands how software works really well. Yes, something simple, you might be able to do it. So I was trying to think, what's the opposite of cognitive surrender? It's not surrendering. It's sort of like the freeing your brain up from doing boring stuff to actually have a chance to spend more time thinking.

[00:33:42] [SPEAKER_01] I think within an organization, it seems like people are really struggling with this. Part of it is if you did take those baby steps we were alluding to before in terms of automation versus augmentation, any capacity that you free up, is any of that getting reinvested? Then how are you reinvesting it such that people are learning, continuing to learn the in-demand skills,

[00:34:11] [SPEAKER_01] maybe make a career pivot or whatever they need to do, or maybe it's just the evolution of their own role. But do they know, do they have the capacity and do they know exactly how to continue to push themselves forward and quell any sort of fears about job displacement and things like that? So I'm curious how you think about that and how your students may think about that.

[00:34:37] [SPEAKER_00] I think we talk about the capacity for introspection and for reflection, being able to take a step back and re-examine what you do and why. And that requires that you have some really strong purpose, some underlying meaning. If you have that, this is the, oh, hey, I can go do this. And so to pick up your point about having, you know, let me back up a step.

[00:35:07] [SPEAKER_00] The guy who taught me OB back when I did my MBA was the brilliant Charles Handy, who wrote a book called The Feet of Work. At the time, he tried to convince me that many of us would end up with portfolio careers. And I'm like, no, no, no, you're wrong. I did. I went and met with him and talked to him again and apologized and said, you were right. Yeah, I've dated my Bayesian priors. I think of this as much more of an opportunity than a threat.

[00:35:35] [SPEAKER_00] So it's having that, you know, that optimistic, opportunistic, embrace the possibility type mindset, I think is going to be one of the huge differentiators. If your motivation comes out of fear, you're going to miss out. Yeah, I agree. I agree.

[00:35:56] [SPEAKER_01] Michael, you, one of your other ventures is in the health tech space. So it was more if you wanted to spend a few minutes talking about, you know, AI in healthcare, which is very, you know, complicated and obviously regulated industry, a lot of sensitive data. I just thought I wanted to bring that up because I'm just intrigued when I, in areas that I don't,

[00:36:22] [SPEAKER_01] I don't have a lot of familiarity, but I certainly know the sensitivity of the data. And there's a lot of innovation that's needed in the healthcare space. So I'm just curious to probe that a little bit.

[00:36:35] [SPEAKER_00] Well, you know, there's an instructive lesson for the venture I'm talking about, which is it's one thing to have great technology. That doesn't necessarily mean you can get the adoption of the fusion you need. And, you know, the reality is that the pharmaceutical sectors at home is extraordinarily brisk at this. So having solved the technical problems and figured out how to go do it, we ran into sort of what you might call the front end of the clinical trials problem.

[00:37:05] [SPEAKER_00] The clinical trials problem is, for good reasons, a very arcane and complex system has observed to make sure that first we do no-call. Now, that being the case, the integration of AI into healthcare is going to take a while. It's one of those sectors where there are well-founded regulatory hurdles that must be overcome.

[00:37:34] [SPEAKER_00] So when I hear people talking about, we're going to solve all human diseases within a year or two, I almost want to laugh out loud because it takes a year or two to register a trial and get the first changes of it going. Might we be able to improve that? Yes. But it's going to take, from my view, several years before we go do that. Will we see, in the meantime, some really interesting early-stage applications,

[00:38:02] [SPEAKER_00] particularly in places where you're not putting it in the diagnostic, where you're putting it in the screening loop, for instance? This MIT Jamil single-shot screening for small cell lung cancer was, we can't be doing any harm. We grab one more x-ray, the exposure is not epidemiologically insignificant, and we can identify stuff when a trained radiologist can't even see it on the charts. That's the sort of innovation where there's very little downside.

[00:38:32] [SPEAKER_00] So my takeaway is the only real barrier, the technologies in many cases are already mature enough to be able to have significant positive impact. The widespread transformation of big complex systems, it's going to take a while because you've got to, this is where reliability really matters. I come out of AeroAstro, most other places, and I really don't want to futz too much with the process by which you get jet planes maintained.

[00:39:01] [SPEAKER_00] They're very, very sensitive, complex beasts, and rather that stays deterministically reliable for a while. You know, somebody who flies a lot, I want the plane to do what it's told rather than have an article with me about it.

[00:39:15] [SPEAKER_01] But I think you're hinting at something that I think may have some organizations putting a pause or maybe preventing them from being an early adopter or an innovator, and that's because of the controls and the regulatory oversight. And so tying that to some of the things around systems thinking, I mean, you have to have the right people, even on that side of things,

[00:39:45] [SPEAKER_01] understanding could this be improved? Yes, we know we've got this oversight and we've got these controls or whatever, but it doesn't mean it's perfect and doesn't have room for improvement. So I think sometimes people forget there's different sort of levels of innovation and different types of innovation. And some of that could be incremental innovation that does, you know, keep the guardrails up,

[00:40:10] [SPEAKER_01] but still find ways to streamline some of the processes around what we can do with the data, the nature of the clinical trials themselves, and improving the quality of some of what they're doing that is essentially still a bottleneck in some ways for allowing everything else to flow through

[00:40:36] [SPEAKER_01] without jeopardizing, you know, human oversight and safety, etc.

[00:40:40] [SPEAKER_00] Let me pick up a concrete example of this. There's two areas in which we're seeing significant payoff already. They're prophetic, but they're a really important part of the physician experience and of the patient's experience. One of them is well-implemented systems that actually just deal with things like appointment scheduling. So figuring out ways that you can make sure that you do use the available capacity effectively

[00:41:09] [SPEAKER_00] and done well gets people in to see physicians earlier, makes better use of physicians' time, you have less dead body. And that can actually, it's boring, it's present, but it's a really worthwhile application. On the physician side, in the spirit of full disclosure, my sister's a doctor, her husband's a doctor, my nephew's a doctor, my girlfriend's a doctor. So the use of the current generation,

[00:41:39] [SPEAKER_00] the state-of-the-art AI note-takers is transformational to people who work in complex disciplines. Because they actually want to be working with patients and listening to patients and being able to look them in the eye. This was a vision that Eric Topol of Scripps had a few years ago when he was laying that out. And it's finally coming to fruition. And it's giving people who are scarce, highly trained resources back their time.

[00:42:06] [SPEAKER_00] So there's somebody talking to somebody who was moving amongst from one system to another. Her biggest concern was, one of her really big ones was, how good is the note-taking ability? Because I've become sort of codependent with it. And it's trained and personalized. Done right. We ended up doing a really interesting piece of work on automating the transcription of electronic health records. One of the things that came out of it was

[00:42:35] [SPEAKER_00] they wanted to normalize all the data. We said, no. It's actually, there's information in scribbled notes in the margin. There were notes to self. It turned out that physicians' notes to themselves were of extraordinary diagnostic significance. You normalize the data, you lost that. In that case, the signal was in the noise. So complex, we're starting to see some real-world impact. My view is, we shouldn't, regardless of this, you have to do everything in a great big bang,

[00:43:05] [SPEAKER_00] but you can find ways of changing the way the process works. And there is near-term payoff to be had.

[00:43:12] [SPEAKER_01] The scenario with the medical note-takers, and like you said, you've got physicians' notes, there's an analogous situation to what I've seen inside organizations in the talent domain and elsewhere, which, and I'm sure you've seen this, Michael, data brokers grabbing conversational data from companies that may no longer exist.

[00:43:40] [SPEAKER_01] They may have recently shut their doors or whatever. Somebody goes and says, hey, you know what? I could, your data is valuable. That back-channel data, that tacit knowledge that got, that people thought was just like throwaway in a Slack message or something like that, that's actually valuable. It gives AI context, and it gives AI a better ability to understand how human decision-making works,

[00:44:08] [SPEAKER_01] and we want that, right? And so without getting off on a tangent about the ethics of that, because I think some of those companies are actually being sued, some of them might even be in healthcare, because I think one of the lawsuits is from a nursing organization, nurses' union is actually suing, right? So like what right did you have? I mean, I'm sure their lawyers are duking it out to say where was the fine print about, you know, who owns all of that data,

[00:44:36] [SPEAKER_01] but, you know, these companies are basically making, they file for bankruptcy, but now they're making millions of dollars selling this data with no compensation for the laid-off, you know, workers, of course, and so it's kind of a disaster, but...

[00:44:50] [SPEAKER_00] Things can be legal but not ethical. This is sort of two by two. Neither legal nor ethical. Legal but not ethical. Ethical but not legal.

[00:44:59] [SPEAKER_01] But the point I was going to make is like that context is critical, is a critical element of how you can extract, you know, insight from some of this, you know, institutional knowledge and tacit knowledge, because it's how work is actually getting done and how decisions are actually getting made, and so you can automate as many tasks as you want and you can try to agentify, you know, your workflows or whatever,

[00:45:27] [SPEAKER_01] but this company is not just a collection of workflows. That's not what makes us successful. That's not why people like, you know, coming to us, working with us. Please.

[00:45:39] [SPEAKER_00] Can I point out, at the end of the day, the tool is going to be a commodity. If you want to be successful as a company, you've got to do stuff that's different. You've actually got to have something that is creative, because always it just becomes a race to the bottom. And so the winners aren't going to be the people who just use AI. It's the people who use AI in a smarter way, to create more value, to be a better place to work.

[00:46:09] [SPEAKER_00] The same applies to all technologies, but I think it particularly implies here that the finding ways to create joyous new experiences that other people can't, that tacit knowledge, like we've figured out how to adapt this sort of organization that we are. I mean, in my firm, we're small, we're relatively agile, and we're actually really quite technically savage, so we have some advantages. But we have a robust set of guidelines,

[00:46:39] [SPEAKER_00] and as we learn, we keep looking at them about once a month just to figure out whether or not there's some new stuff we've learned that we should be incorporated. We have some hard guardrails that you just cannot cross. Certain sources of data is just never going to be accessible to the AI. That's it, right? Checking stuff, writing stuff. We actually have a rule that says human beings have to write everything, right? It's going to go out to a client. Why? We own it, right? So you're not allowed

[00:47:09] [SPEAKER_00] for stuff that's going to go to a client to actually have a human, a non-human being write it. Help you parse a document? Yes. Finally attest to a citation being correct? Absolutely not, right? We need to have robust tools that are deterministic and reliable. We'll do that. So thinking about that system, where's the human? Where's the deterministic, utterly repeatable automation?

[00:47:38] [SPEAKER_00] Where's the augmentation? Where's the AI automation? The LLM automation?

[00:47:45] [SPEAKER_01] Yeah. I mean, I think this is all part of what companies need to figure out and why redesigning the work itself is so important. And, you know, you've got to, I use this phrase, strategic work planning. If you're in HR, you use strategic workforce planning. But it's really the work itself that needs to evolve, whether that's done by a person in your immediate or extended workforce or by some, you know, digital labor, if you will, in the form of an agent or an AI assistant, et cetera.

[00:48:15] [SPEAKER_01] And so you've got to really understand the economics of, you know, the trade-offs because AI is certainly not free. You can't just start cutting heads. But also what is the capacity of the organization that you can unleash when people are freed up from some of those, you know, efficiency gains, which are nice. But I think to your point, you're going to get diminishing returns from the efficiencies. And so where is the real growth and value coming from?

[00:48:43] [SPEAKER_00] Yeah. It's not going to save you in the long run. It's interesting you talk about that, about change. I mean, I do think this is a profound leadership challenge because you're simultaneously trying to be very left-brained and get people to be rational and say we have to adopt these technologies because it will make us more effective and more efficient, more effective like producing better output. But at the same time, if you're sitting on the other side of that coin, there is an awareness that it may reduce

[00:49:12] [SPEAKER_00] the total amount of work that's being done. And that is a basis for people being legitimately anxious. I simultaneously need to encourage people to do something which may not be in their best interests. And that's a challenge for leadership. I think it's a unique challenge and that it requires this combination of being sort of both left-brained and right-brained. I will confess that sometimes that may not all be in one person. You may need a combination

[00:49:42] [SPEAKER_00] with a leadership team which is some of us are more left-brained and some of us are more right-brained. So my COO's much better with people than I am.

[00:49:51] [SPEAKER_01] Michael, I've really enjoyed this discussion. I want to be respectful of your time. Any sort of final thoughts as far as what you're anticipating for 2027 and beyond or where could people find some of your writings and resources?

[00:50:07] [SPEAKER_00] After a hiatus over the summer you'll be able to find me on LinkedIn. There'll be a post today for the next few weeks. Some announcements about some stuff we're up to. I'm waiting for hoping confirmation is going to come in. I'll be looking at my email about some stuff. Other than that you can find me through MIT or through London Business School and soon through SDA Bocconi. We have a substack about to go meet with one of our writers.

[00:50:35] [SPEAKER_01] Okay. Excellent. Well I'll make sure we have the right resources in the show notes of this episode and I'm on Subtack as well so we can do some cross pollination perhaps on our audiences. It's been a real blast. Likewise. Thanks so much Michael and thanks everyone for listening. We will see you next time.