Is AI More Biological or Technological- It’s Complicated!

In Season 2, Episode 3 of The HumanUp Imperative, Rex Wallace is joined by Dr. John Sviokla, co-founder of GAI Insights and Executive Fellow at Harvard Business School, for a grounded, forward-looking conversation on AI and its implications for healthcare leaders. Dr. Sviokla argues that by 2030, every competitive organization will operate as a hybrid of human and machine intelligence, and that healthcare is no exception. He walks through his RISE framework for AI adoption, explains why AI is a capability that must be grown rather than a technology you simply install, and makes the case that senior leaders, not just IT teams, need to be hands-on with these tools. The episode also explores the human stakes of AI deployment: organizational values, decision authority at the edge, and what it means to optimize not just for human audiences but for the AI models increasingly sitting between organizations and their customers.

Transcript

SPEAKER_01 0:00

In a world increasingly shaped by technology, the Human Up Imperative explores the significance of human connection with each other, with the communities we serve, and perhaps most importantly with ourselves. Join Rex Wallace and his guests as they discuss the ever-important role of authentic, meaningful connection in healthcare. It's time to human up.

SPEAKER_03 0:22

Hey everyone, welcome back to the Human Up Imperative. AI and even generative AI have been around for years in some fashion, but um it all took new life in November 2022 when ChatGPT debuted. By January, it had become the fastest-growing consumer app in history, eclipsing 100 million users. And if you think the adoption is evolving quickly, it pels in comparison to how quickly the machine itself is learning and evolving. Just a few months ago, the former CEO of Google said this artificial superintelligence, the theory that there will be computers smarter than the sum of all humans, will no longer be theory in the next six years. Um and that this is happening faster than our society, our democracy, and our laws will address. And then lastly, in May 2023, Sam Altman from OpenAI and Bill Gates and many other prominent AI researchers and tech leaders signed a statement on AI risk stating that mitigating the risk of extinction from AI should be a global priority alongside things like pandemics and nuclear war. Um so today we're gonna talk about AI and humanity, and we're gonna talk about it with the co-founder of GAI Insights, who's also an executive fellow from Harvard Business School. Really excited to uh um invite you to join me as we hum it up with Dr. John Sfiokla. John, good to see you.

SPEAKER_02 1:42

Good to see you, Rex. It's been a while. It's great to be with you.

SPEAKER_03 1:45

Yeah, you too. Thanks so much for taking taking some time with us. Um, you know, can you I gave a really quick intro there. Can you share a little bit more about your background so listeners will understand who you are?

SPEAKER_02 1:55

Yeah, absolutely. Yeah, so um the we founded this firm, GI Insights, basically as a way to accelerate uh organizations to go from human-only to hybrid human and machine. And we're a combination of research firm, strategy, executive development, uh, community, and uh conferences. And you know, my background, I got interested in AI and back when I was a doctoral student at Harvard Business School back in 1983, and uh did my finished my doctorate in 86, went on the faculty, did the first AI course there, because I was always uh impressed by what's the difference between how people and machines think, what's it going to do to economics and society? And um, it's one of those wonderful topics that has a combination of everything from philosophy and identity all the way through uh, you know, give me a cheaper widget, right? So uh anyway, uh it's been a part of my career for a long time, and I uh really uh excited about where we are today.

SPEAKER_03 2:49

Yeah, I uh follow you obviously on on LinkedIn and and uh and some very interesting posts about AI. And um, yeah, I'll I'll invite you later to kind of share your your LinkedIn information and but um but you're definitely the the person to have on today. So yeah, AI is everywhere. Um, right, it's it's making life so much easier, and it sounds like it's you know it's also gonna be the death of all of us. So how how how should you know our our listeners are primarily you know in the healthcare space, so think of healthcare leaders um across health plans and providers and vendors and uh you know regulatory agencies and things like that. What what what should healthcare leaders, how should they be thinking about AI right now?

SPEAKER_02 3:35

Yeah, well, I think that there's there's three things to keep in mind. One is that um uh by 2030, every organization that's competitive, um, not just in business, but competitive, it's gonna be a combination of machine and human intelligence. I think that's pretty easy to see. There's no there may be a bubble in certain kinds of infrastructure building out, and you know, is Tesla worth the XAI worth all who knows? Who cares? But if you look at adoption, there's no there's no bubble in AI adoption. So people who are in uh adopting are adopting more and getting value. Second thing is that uh we should live at the intersection of the of the possible and the practical. And what I mean by that is that you know these these tools are unbelievable in terms of capability. I think of them as power tools for knowledge work. So, you know, if you came back from work and you'd hired an electrician and you know, put a couple outlets in and she was there with a hand drill, you'd be like, hey, what the heck are you doing? If you have anybody in marketing or customer service or reviewing legal documents or trying to do a regulatory filing, and they're not starting off by using these power tools, that's just like you know, imagine a cube farm and everybody's sitting there with hand drills, right? So it's this is a now thing. And the third thing is probably most importantly, is it's really not a technology, it's a capability. And the distinction I have there is you can buy a technology and stick it in. A capability you have to grow. So we we uh hang a lot of our work off what we call the Rise adoption framework, which is research and education, islands of innovation, then scaling and really doing regular projects, and then emergent intelligence, which is where the AI starts to build the AI. And so how to think about this is very practically, but also you have to think about how capable is my organization to do anything with it.

SPEAKER_03 5:28

Yeah, so so 2030 is not very far away, right? Um what so if if if I'm a healthcare executive, what what should I be how how do I start laying the groundwork for that? What should I be doing as an organization? Yes, to become a high to become a hybrid organization, right?

SPEAKER_02 5:43

Yeah, first is to realize, you know, look, you can still be in business, you're just not gonna be competitive. Like yellow cabs are still around in New York City, but you know, their value went from a a million uh million four per medallion down to about 30 grand and then has has recovered about 200,000 bucks. So they're still in business, they're just not competitive with Uber. And so that's the that's the idea that you have to kind of get in your head. Then the second thing is to think about two sets of things. You can do really simple stuff, you know, um, so all the stuff you hate doing in work, you know, it's kind of like I think it's like kind of picking up the trash, you know. Okay, we have to we have to straighten out this uh you know, these regulatory filings, we have to look through all kinds of customer data, you know, we have to do you know basic reports, that kind of stuff. Much easier with AI, right? Just save a lot of stuff. Then on the high end, it's like can you can really use it as a thinking partner because you talk with these things. We've never had a technology that we can talk to it, and you can even walk up and say, hey, I'm John Sfiokla. You know, this is what I'm trying to get done. How can you help me? And the machine can tell you, right? And so I think that um I think that leaders need to understand that. And I think that leaders need to start with the simple stuff that's low risk, that's internal. Then you can go to a little bit higher risk stuff that's internal, then you can go to the low risk stuff that's external, you know, FAQs about your business, you know, how to apply for a job here, all that stuff. And then you can go to high-risk stuff that's external, right? Kind of that progression. And people have to realize that AI is the new UI, AI is the new user interface. So if you send me to a traditional, you know, uh interactive voice response system, an IVR, which from an AI's perspective is like the dumbest AI in the world, right? It doesn't remember anything about you, it doesn't know anything about the context, it makes you repeat everything. It drives you nuts in terms of, you know, well, none of that. Anybody who interacts with one of those, after you've, you know, you can hold up, you can literally hold up your phone to your car and say, what kind of oil does it take? Where do I put it? How do you know? And Gemini will tell you, like, oh, you need this, and you know, I mean, uh, you know, so somebody using a a traditional interactive voice response, they're gonna say, this company's broken, right? They're from the dark ages. They're they're they're not even you know qualified to be called a dinosaur. So I think that executives need to understand that. The other thing is if you try to outlaw this stuff, the surveys are showing that a full third of your people are using it. So you say no AI or only this restricted piece, 30 of 33% of your employees are going out and using it at home. And then you have the worst of both worlds. You're not learning anything as an organization, and you have none of the risk controls because you want to have verified, secure chat and research and agent environments, right? And you can get them from the major vendors. But if you've outlawed it, you know, people are gonna route around it because this has been the biggest uh career accelerant, and your best people are the ones who are gonna use it first, and they're using it. University students, over 80%, are using it all the time. So you have to wake up and realize you can't control this thing, you want to guide it.

SPEAKER_03 8:50

So um, you know, uh around those lines, Anthropic has a resident philosopher, right? So they actually have someone on staff, Amanda Askel, who her whole job is to is to teach Claude to be to to give make sure Claude has morality, right? Like that that it makes um that it's that it's good. And Chat GPT's gotten some bad press for you know um what its platform is doing sometimes when people uh in the guidance that it's providing, right? So anthropic, in in my mind at least, I think that in the in the public's mind, it's sort of the anti, I feel like the anti-Chat GPT or something, but but anthropic is seems to be investing a lot more in in humanity and morality and making sure this this tool does good, right? Um if I am a if I'm an organization and and I know that my employees over the next three or four years are going to be all leveraging this and and and my um organization's success is going to be driven by it, um, I want it to be making good good decisions, right? I'm curious um what your guidance is when working with organizations and they're adopting this now. Should they have, are you are you advising them to have like a chief morality officer, chief humanity officer? I guess what's I see what anthropic is doing, but what what what should like a healthcare organization be doing? What kind of role should they have responsible for making sure that this tool is doing good for my customers?

SPEAKER_02 10:25

Sure. Well, first let's talk about anthropic. They have this notion of what they call constitutional AI, and that includes uh combination of morality and policy. Um and you saw they got the dust up with the uh with the Trump administration around the fact that they wanted to put a bright line around surveilling American citizens en masse and also um autonomous weapons with authority to kill. And and the Trump administration came down very hard on them. This uh the designation they put in of uh anthropic being a threat to the global to the military to the industrial supply chain has never been used on an American company before. And um it means that you know, let's say you're a cursor and you've built something on top of, which is a a tool to help with software productivity, um, and you've built on anthropic, you can't use cursor, right? So this is a pretty harsh thing, and they're litigating it and so forth. Um, I do think that uh the idea of uh putting in values and so forth is important. I don't think you need to have a you know an internal um morality person and philosopher. I think it's but you do need to bring in the values and the tone of your organization as well as your risk policies and so forth, right? And so uh and there are ways to do that. You know, for example, Blue Cross Blue Shield of Michigan um you know has implemented a tremendous amount of um traditional and generative AI. And if I if I say to you, hey, Rex, you know, you're sitting there in an emergency room, uh, you know, am I covered for this? Right? And so they're using a combination of traditional AI. Hey, now I can use your voice to identify you so I don't have to do all this uh background when you know you're pressured for time and so forth, because by law, as you know, regulation, they have to verify that it's you. And so you can you can squish that down with traditional stuff, and then the ability to actually access in many cases there there's over 450 APIs inside um uh Blue Cross Blue Shield of Michigan, which are then calling on agent, which are the agents are calling on those APIs. And so for you, for example, in that setting for that provider, they may have to actually go down into the contract in a way that previously that granularity would have been lost in the contract and you would have been at a higher level, and so you might have got rejected when in fact you're allowed. And so what generative AI has allowed them to do is to link down into that detail, then bring it up. They have a truth uh verification system, and there's different modes they do that. For example, one of them is they'll have three different models look at the answer and make sure they get the same answer. There's other times they'll reach out to a database and verify a critical variable. So they have very strict truth and security access, but they are reflecting the values of Blue Cross Blue Show, Michigan in every interaction, and they're reflecting the policy and risk. So I don't think you need to add uh that philosopher within your company, but I do think you need to know that you have to teach the AI your policies, your risks, what to do if it doesn't know the answer. Don't just guess at it, say I don't know the answer, right? These are all things that you can train the AI to do.

SPEAKER_03 13:32

Can you uh you mentioned your RISE, the the RISE um not platform, but sort of the the process, right? Of how I'm curious if you can unpack that a bit for the listeners.

SPEAKER_02 13:45

Yeah, so when something's a capability, uh what what's going on here, and and this is different from any other technology, right? When you think about the implementation of uh traditional machine learning, or let's go back to automation, right? You would learn how to work with the machine, so you learn how to operate it. Uh, or in the case of machine learning, you'd learn what the what the algorithms were intended to do, what data was coming in, what the answers are, and what you know from a certainty or uncertainty perspective that the answer is, right? This, so that's very much engineered toward an answer or a task. This is a dialogue, right? This is hey, uh, how confident the first of all, what can you help me with? How confident is it? Um, you know, uh how do I get this to work? Right. And literally what I do now, I don't use software manuals or anything like that. I hold my phone up to the uh to the computer if I'm stuck on a, you know, because there's so much software we all interact with. Right. I won't even think about going to the help function. I'll say, hey, uh Chat GPT, or hey Claude, or hey Gemini. Um you know, this is what I'm trying to get done. This is where I am, this is the software I'm using. I take a picture of it, I said, I don't understand why I can't do this, and it will come back and tell me how to use it.

SPEAKER_00 14:59

Right. Yeah.

SPEAKER_02 15:00

Okay. So I mean that the power of that, um, of actually having a dialogue uh and not having to look up, and you know, 99 times out of 100 for the software stuff, it's right. I mean, you know. Um, and so you know, you can just imagine like how many places inside a payer um are these poor people dealing with uh I forget the number of contract variations. It was in the millions, you know, um, in terms of a lot of these payers. So no human being can remember all that stuff. And you spend all this money creating rules and regulations trying to, you know, imagine being able to take all that complexity and have it distilled down. Uh and you don't have to do it beforehand, right? You don't have to say, okay, we're gonna think of every possible question and spend all that money and compute all that stuff and create an answer lake, like a data lake or whatever. You don't have to do that. If you architect it correctly, it'll go out and find it, make a make an assessment, then you put it against your truth function, you know, maybe it hits a database, maybe it goes against, you know, three or four models that vote or something, and then goes back if it's not right. So, right, so you get verification. And then it's just distilled, it's kind of made to the answer's kind of made to order for you in that complex environment, and that's incredibly effective and efficient.

SPEAKER_03 16:10

Yeah, I'm I'm taking an AI course actually, and we're using uh they're teaching us Python, which I'm not a coder, right? So that's uh but it's but it's a great exercise to learn to learn kind of the foundations of it, and and we're using Gemini, right? So what you just said, I mean, whenever if if I'm trying to write something in Python and and I something is wrong and I can't figure out what's wrong, right, there's just a little ask Gemini button, right? And it it will it will define, it will explain the error and and then it will offer to fix it for you, and that's all you have to do, right? And it it does it all.

SPEAKER_02 16:44

Exactly. And and if I you know were to get stuck in in something like that, which I haven't yet, is I take a picture of it and I go to Chat GPT and say, Hey, Gemini's screwing up here, tell me what's going on.

SPEAKER_03 16:53

Yeah. Yeah, I I I heard, and I think it was the ex-google CEO said this too, that um, you know, in like one to two years, like all programming will be done by by AI, right? Like humans won't be doing I mean it was a very short timeline for that piece of it.

SPEAKER_02 17:08

Um Yeah, I I I'm not quite there. I think the role changes like the bulk of early computer science when you're taught it and when you come on is the execution of stuff in a complex environment. That part's coming way down. But design, uh it's it's hard to configure uh uh uh to properly specify a task over time. Let's say the task takes about two weeks to do or something like that for a human to sit down and articulate all the possible questions that that agent's gonna have over an extended period of time is actually a management art form, right? And very few people know how to be that precise on the on the instructions of how to do a complex task over time with understanding of what the agent might want and need as it goes along or if it hits a dead end, what to do. That is actually that is a skill, it's a human skill still, right? And and it's not a human skill fully in the sense that the the machine can plan it, but it it can't plan all the exceptions and things like that. You have to teach it the context. And so that skill, and you talk to anybody at these sophisticated software shops, you have uh like one of the uh hyperscalers who who remain unnamed. The way they're doing it is they have a bunch of senior uh uh software, not only architects but developers. So people who really built big complex code and driven teams. So it's not they're not just individual contributors, they know how to structure the work for other people, right? So they have driven teams before. They have that, and then they have very junior engineers, and they've gotten out of the they've they've removed the middle folks who are wedded to the way of doing the work as opposed to the outcome of the work. Okay, so a lot of people are wedded to, you know, I'm a software programmer, I do this process. Those people are gonna have a very hard time in this environment. If people say, look, I'm a software programmer, software engineer, and I get these kind of outcomes, those people are flourishing in this environment because they're open to different ways of doing the work, not just to get to the outcome. And so that's where so I I think it's a little more nuanced than simply everybody's gonna go away in that category. I think the the execution part's getting squished, the design and the and the uh and the the the oversight of the agents is gonna grow. And just to give you an idea of of how powerful this is, the cursor using uh Claude code built um set the agents uh autonomously, and the record up until um sonnet and up until Sonnet 4.6 was uh about two days where agents could go off and productively before it went off the rails, right, without instruction. This went for 10 days and Rex, they built a fully functional, fully functional C compiler. If you know anything about building a compiler, that would take a couple months of a good software team. This is ten days agents by themselves created a fully functional C compiler. So that is amazing. But you have to hand it to the cursor engineers who knew how to instruct it to be able to go that long in such a big complex area with all the dead ends you hit when you do something complicated, and to teach the model how to think about thinking, right? That specification that is not an easy skill. And that and it's gonna be incredible demand for that because these agents can do a ton when you set them in the right direction.

SPEAKER_03 20:49

Yeah, yeah. I am I'm curious I'm curious you're thinking. I was having this conversation the other day. Um something so in these sort of early, I mean, if you if you call this early days still, but for you know, I think many organizations are still in the early thinking, right? And in how in how to leverage this and and how to train their staff on it. You know, and I've read quite a bit about prompt engineering and how like it feels like that's still a place where there's a lot of opportunity to train people, but at the same time, like we were talking about with you know, like Gemini will just you just kind of tell it in natural language what you need or send it a picture. And it it doesn't need a lot of engineering in some aspects, but I feel like I still feel like um if I'm gonna be a heavy user of AI, I need to be trained on prompt engineering. But but I'm curious your thoughts. Is that is that something organizations need to be doing with their staff right now, or or is it not necessary any as much as it was earlier?

SPEAKER_02 21:43

I think prompt engineering is very, very important. And and back to the RISE model. So, you know, this notion of the that's the research and education part. So I I think of prompt engineering is basically learning how to work with a co intelligence. So prompt engineering is one thing. I think instructing agents over time is another. Thing. I think understanding how to constantly work at a new velocity where I can take stuff I used to do myself 100%, and now I only do 20%, machine does 80%. Like that process, we haven't seen that process in in technology for many, many years. Like in the in the 19, certainly the 60s and 70s, and to a little bit degree 80s, it executives actually had to think through do I want to have a relational database or not? Do I want to buy stuff on a timeshare basis or not? Do I want to invest in custom software? Like back in the 70s and 80s, like let's say you're a mid-size, and I'm sure you know a bunch of the payers can look back in their code base and they were building stuff themselves, right? Because the applications weren't stable, the the providers were varied, right? And then what happened is it got really simple, right? It became a bunch of dominant players. You're in this category, you buy workday, you buy this category, you buy sales, right? And it's all in the cloud, and like you don't have to worry about the stack, and you know, you pay the bill, right? And so that's kind of been the core. Well, that's going away. We're back into the 60s and 70s, where a senior executive, even if they're not a technology executive, have to know enough to understand which path we're on, right? And and the IT department has a much more um complex job because they have to learn to buy from small vendors, they have to learn to configure their management in such a way that they can swap out vendors, right? People haven't had to do that for years, right? Uh two generations of technology executives have not really had to open up under the cover, really. Um and I know there are exceptions, but in the main, right?

SPEAKER_00 23:41

Yeah, yeah.

SPEAKER_02 23:42

Um, and and a lot of your hard work with selection among stable alternatives. Well, the alternatives are not stable here, and you wouldn't care, except in software, for example, we're seeing 30% or more productivity increase. And now, you know, Google just released this thing on you know, reading COBOL, uh, IBM's, you know, stock price went down 15% that day. You know what I mean? So the the legacy code base, refactoring it and all that other stuff, you know, so it's the good news is more productivity innovation. The bad news is you can have to work harder and take some risks in terms of alternative supply. So, anyway, that's the so in the RISE model, that's the research and education and islands of automation, because you can't just know a capability, in my definition, has three things. You can know, you can do, and you can teach. Okay, if you have lots of people can know, do, teach, you have a capability, right? And if you have people who just do or don't really know like the core things or can't teach to other people, then you don't have a robust capability.

SPEAKER_00 24:42

Okay.

SPEAKER_02 24:43

So that notion of no-do-teach altogether is vital. And then um, so how do you build that? You get the no part and the research and education, and this is hands-on. And I've done this with CEOs of multi-billion dollar companies. There's no substitute for hands-on because this is an experience until you experience these models with good prompt engineering and you know the ability to do stuff, you have no idea what you're looking at, right? And and your intuition is not right because we've never had technology like this. We've never had technology to talk back, can do stuff, can you know. So, absolutely essential, hands-on.

SPEAKER_03 25:19

And you're you're saying this is not this is not something you delegate to your IT team and the senior and the rest of the senior leadership team doesn't really need to be hands-on. You're saying the senior leadership team needs to be hands-literally, hands-on.

SPEAKER_02 25:32

Like, go build an agent. Go, cause you can't, it's like trying to describe swimming. I can't describe swimming for you, right? You have to hop in the water.

SPEAKER_00 25:40

Right.

SPEAKER_02 25:41

And and I'm not saying they have to become expert, but they have to know enough so that it can actually recognize opportunity, right? And so that's vital. Second, the and so so top down, you need to do that. Bottom up, you need to start training people so they have some knowledge. Then you have to apply that to a bunch of areas because everybody knows the difference between knowing and doing, right? So you want knowledge, but you want skill because as you're building skill, you look at the knowledge differently, right? And so, so a number of people have said, Oh, we don't want to do a bunch of small stuff, we just want to do the big stuff first. We haven't seen anybody successfully do that because they don't know enough to specify correctly, even if they're using a third party, as to what the heck it is, right?

SPEAKER_00 26:26

Right.

SPEAKER_02 26:27

So that's the I, the islands of innovation that's absolutely vital in building up your capability base. Then after those two things, you have to kind of up your game, you have to deal with IT, data, security, a bunch of stuff. And then you create scalable thing. We're gonna redo, you know, customer service front to back, you know, we're gonna redo you know, member enrollment, okay, or something like that. Okay, that's and those you manage. The the first one's an investment, the research and education. The second one, I like to think of it as break-even. You know, you want these individual projects to have net present values and business cases, but you know, on average, if they if they come out break-even, you're great, right? Because you've built capability, some win, some lose. Okay. So it's not a big investment, but uh, but don't think of it as gonna add to your EPS, right? Or something like that. Once you go over the gap and you do scale stuff, that you manage a traditional return on investment, just like any project. And then the somewhere between S and E, between scale and emerging. In emerging intelligence, AI is starting to build the AI. Now, the place we've seen that the strongest is in Silicon Valley. And, you know, um Anthropic said that 90% of code, of the code in anthropic co-work was written by the robots. 90 90%.

SPEAKER_00 27:43

Yeah.

SPEAKER_02 27:44

And the cycle time of of building and shipping code now is happening very fast. I mean, you we've I mean you've been around technology for a long time. I've been we have never seen series of technologies of this complexity and this quality launched on a global basis every three to four weeks. I mean, these aren't minor, these aren't minor builds. These are like bang. I mean, and you can only do that with when the AI is building the AI. What happens between S and E? If you do S well, you start to have a return on investment and your earnings are go up, your costs go down if you're mutual or whatever. And then between S and E, you start to see a change in the operating statistics of the business. So the and here's a the most prominent one is revenue per employee. You look at AI native companies, their revenue per employee is one, two, four, five million dollars. As Scott Galloway so put it so succinctly, he said AI is ozempic for executives, right? Most executives think I grow revenue, I have to grow my popular, I have to grow my base, right? Right.

SPEAKER_00 28:50

Yeah.

SPEAKER_02 28:50

Uh this delinks that just like you know, when I'm on my GL1, I can sit in front of a pizza, which, you know, if I learn Death Row, it's pizza for me. I, you know, when I'm taking my you know, zip bound, you know, I can look at the pizza and say, oh yeah, that's food. I don't have to eat it. Same thing with executives.

SPEAKER_03 29:07

Okay, so I love the rise platform. And uh one thing you said, I I want to make sure we double down on for for listeners too, because you see you you mentioned top down and and bottom up. So this this is a big transformation, right? This is something so you're you're saying this is this is not top-down or bottom up. It's not one or the other, it's it's it's both, right? It's both, yeah.

SPEAKER_02 29:30

Okay. Yeah, I mean, if you if you and you know, having been a business academic and you know, a fellow now and and so forth, I've I've I've read more, you know, org change research hard research than than you would want to. And um yeah, and if you kind of step back and squint your eyes, this is what you see. Yeah successful transformations have three things top-down commitment, bottom-up enablement, and clear program management. Okay, we're headed here. Let's not let's not go all over the place. If you do a big transformation, though that it's been the same like forever. But either of those pieces missing, the top-down or the bottom-up, you're not gonna get there. And the clear program management, you're gonna wander around. So that's what you need to do here because look, this is this changes everything. I mean, uh uh, how do you do employee reviews? How do you do presentations? How do you do underwriting? How do you do uh how do you get your name out in the market? AI is the new UI. So many people now, if you haven't looked as an organization at as to where your brand shows up in the answer engine, so you have search engine optimization, we've been doing it for 30 years, right? Answer engine optimization. People are the people don't want alternatives, they want an answer. And so you go into these models and it will give you an answer, and you better understand where you are. I mean, everybody you in the old days, you could buy your way to the top, right? You want to you want better presence on Google, spend more money. Now, not so much. Because the answer, first of all, you have to organize your data differently because all the data inside and outside organizations has been optimized to humans. Well, now you have two audiences. You have to optimize it for a robot, and you have to optimize for a human, and a robot reads differently than a human does. And if you don't do that, and you have to remember that in more and more transactions, service, purchase, you know, whatever, there's a robot between you and your end customer. And so if you look, let's say you're a payer organization, and let's say all the good reviews, like let's say there's you know, been uh an independent review of payers in this geography, and you're doing fantastic, you're way up there, right? That's great. Okay, now let's say that's in a PDF file, you know, in your marketing channel or something like that. Robot doesn't see it, completely irrelevant. Doesn't matter, right? And the and the models like to look for third-party verification, right? So third-party verification is somebody else, they don't just want to hear your opinion, they want to hear the internet's opinion. And uh and very this goes from very simple stuff, you know, to have have you organize your data like with JSON LD and things like that to make it robot readable, uh, all the way through uh what's your content strategy, because I can't buy my way to the top anymore.

SPEAKER_03 32:22

Yeah, um so much to think about as we if if we want to be competitive, like you said, by 2030, be to be competitive, you need to be a hybrid and and um and be all over this. Two two two things that I've heard you say before. So so one, and I just want to make sure that um that you have the opportunity to get this out in front of the listeners because it's so interesting. But one, you said you you you look at AI, I think through a biological lens. Can you can you unpack that?

SPEAKER_02 32:51

Yeah, yeah, that essentially um look uh this is a co-intelligence, right? And um and the way I think about it is that um uh birds fly and planes fly, right? They fly very differently. And and no one would ever kind of ask, you know, does a plane think like a bird? Okay. Well, with machines, I think they already think. We can't execute Wall Street if if you're uh you know, if you're a farmer organization, you know, you can't execute SNLP, you know, your sales and operations planning without computers, right? And without machines thinking for you. You think about Wall Street, what is Wall Street? It's got a bunch of agents out there with rules, oh, program trading, you know, in in literally nanoseconds or at least you know microseconds, you're making decisions on the context and you're executing real trades, right? That's so those things are thinking. And um, so that's important. And then traditionally we've thought about things mechanistically, like, okay, we get this policy, we have a bureaucracy, here's our standard operating procedure, we teach to everybody, we put it in all the policy, we put it in all the software, stuff like that. Well, this is much more of a dialogue, and it grows because what the machine knows about you increases over time, what you know about the machine, and that's much more like an organic analogy. Um, and you know, I actually think of them um more as alive. And um, and you think about like look at look at our bodies here. Okay, so what is what's going on in my finger here? One of the things that's going on is that the my entire um code base is in available in here in DNA, right? Then it's getting expressed, and there's all kinds of localized stuff happening. If I hit it with a, you know, it starts to get infected and so forth, right? It's incredibly localized. Now it connects to the mainframe, right? But there's incredible redundancy and localization of intelligence and and information in biology, not so much in mechanical systems, which tend to be more centralized. This because the cost of the cost of intelligence is going down and the availability of localized energy is going up, over time, this is going to look much more organic than it is mechanistic. Okay. Now, what what does that kind of high-fluten thing mean? It means that some of the stuff you ought to be really thinking about is what decision rights do you give out to the edge? And we know a lot about this from service management. There's some fantastic work done by guys like Earl Sasser, Jim Heskett, and uh Gary Hart who looked at service organizations and showed that if you're willing to give more to train on values and and um values and uh procedure out at the edge of a service organization, and then you give them local authority, you can improve the quality of service. So, for example, I don't know if it's still true, it used to be true that the Southwest Airlines person at the edge could spend up to 1,500 bucks to make you happy. So that person at the gate. Okay. Southwest is the only person, the only company that would allow that. And it was on local authority and they wouldn't get grief for it. Okay. Well, uh, you know, I think Southwest is a little different now, but you know, yeah, 15 years ago, I mean Southwest or 20 years ago, Southwest was like over the top, I mean, they were like in a whole nother category of satisfaction.

SPEAKER_00 36:15

Right.

SPEAKER_02 36:15

And a lot of it had to do with that ability to push authority out to the edge. Well, AI was gonna allow lots of authority to be pushed out to the edge, but you have to think like the special forces, not like regular army. If you look at special forces, what do they do? They cross-train their people, they select them for the morals and and values that they have. Like I saw, for example, I saw some some guys, I I love watching, you know, Top Gun, and you know, I love you know all the aerial photography stuff. You know, it's and I saw some real Top Gun people talking about that. And they said, you know, is is the stuff in the movie real? They said the biggest thing that's not real in that movie is the attitude of the pilots. They would never have that kind of baloney burado. He said that would get you deselected, so they didn't have the right values. They said the flying and all that stuff, absolutely. Um, so that's kind of cool. So value.

SPEAKER_03 37:08

So less Iceman and Maverick and more goose, right?

SPEAKER_02 37:11

Right, more goose, less maverick. And so um, or more second generation maverick.

SPEAKER_00 37:16

Yeah, right, not first generation maverick.

SPEAKER_02 37:18

The um, so there's that. They have a shared information environment, right? They have uh dynamic decision rights. If you and I are working together and you're better at than I am, you I may be your boss, but you lead, right? And and so, and then clear objectives, like what are they aimed at? Well, those principles, those special forces principles, I think are uh AI is gonna make people companies that implement AI with that are gonna be much more powerful because instead of having just like in the Vietnam War, if we continue this analogy, a lot of the problem was that a bombing run uh in some place might have to go all the way up to the Pentagon and back. Well, guess what? On the ground changed way faster, and that bureaucracy costs lives. So that was where they came up with some of this thinking. If you look at the first Iraq War and you go back and look, you might have seen occasionally the camera would pan up in the sky and you'd see F-16s or other aircraft circling. Well, in that war, a corporal could call in a bombing strike on his or her local authority and call it down from that inventory of of munitions in the in the sky. Well, that that notion of putting it at the edge, now governing is another whole thing, but that war, if you look at it, was amazingly prosecuted, right? Uh, I mean, just the and it wasn't because they were a pushover army, right? Anyway, um, so that notion of being able to put power to the edge is gonna be critical. Um, because you can have the best of both worlds. So you have the knowledge and the policy at the edge, and then if you give them the training and the authority, they can take action to really make a difference.

SPEAKER_03 39:02

Okay. Well in in um and one one more really quick when kind of related to this before we get into the wrap-up. Um because to me it ties in with this. But you mentioned um, I've heard you mention before the bear the bear analogy. So from the mid-sized company. So can you can you talk about that for a second?

SPEAKER_02 39:21

Yeah, uh the the the the most important thing for executives to understand in this domain is uh forget averages, right? Because when you're competing in a market where the market's transforming, you just care about on average, most companies aren't going to adopt. So if you sit there and watch the averages, yeah, hey, I'm not behind, right? Most people aren't. That's irrelevant. If you have one competitor, one competitor against you that's really adopting this stuff, they're gonna steal market share or growth that you should have had from you. And it's just like the old analogy you don't have to outrun the bear, you just have to outrun the other person, right? And so that's what you have to ask is you know, is there somebody who's faster in your market that you care about? Uh, and that could be a new entrant, you know. So you have folks like Harvey in the law field, you know, if if you're competing against a law firm that's using Harvey, you're gonna be crushed if you're not, in my opinion. And then, because you know, there's a tremendous amount of productivity of the law and quality and so forth. Or um you have somebody like a JP Morgan Chase who has gone at this hammer and tongs, and they they did the research and education, they did the islands of automation, they've committed the platforms, they have three major platforms now. And some of the uh some of the things that the investment bankers are saying is let's say um you're you know, I'm gonna go see Rex, he's a high net worth client, he's got you know a house in Singapore and here and whatever, and you know, bringing all that stuff together, and you're gonna go visit the London office because you're you know some, you know, you just get divorced or whatever, you know, and you want to like straighten yourself out. Um preparing for that, they're saying that they're seeing 50% or more decrease in prep time to get ready for a high net worth client. So I believe in 26 we're gonna start to see the operating statistics of JP Morgan Chase pull away from its competitors because it's so fully committed to this, and this is about the time it would start to show up.

SPEAKER_03 41:14

Sure. How how will we know when AI's gone too far?

SPEAKER_02 41:19

When you're looking up at it, argue with it. No, the uh um no, I think uh I would watch for uh two big areas. One, and I think these are gonna be the the first ones, one is uh when we uh when it's got really detrimental effects on um political um activity because uh uh these things are incredibly good at getting you to relate to them. And so now the you have robots that are are you know you're relating to and they're empathetic and they never forget and they don't get pissed and everything. So it's that's a that is a very, very powerful and potentially dangerous area, and politics you know allows you to create it's very easy to create a cult or a subgroup with these things because it's very good at both the the thinking and the emotional part. It's good at face looks good at dealing with my amygdala, okay. This thing's good at dealing with my prefrontal cortex, right? So that's a powerful combo, right? Yeah, and so I think there's that, and the other is an autonomous weapons. And depending upon where um where this current war goes, um, I mean, if you look at their hypersonic missiles the Iranians have, right? They they cluster, they've got like 80 warheads coming down and so forth. What we haven't seen, and they also promote the uh the small boats versus the carrier groups and stuff like that, are gonna be much more effective because of the asymmetry of cost and and flexibility. And I I don't I have yet to hear the kind of stuff that the Chinese have been demonstrating. The Chinese have demonstrated massive swarms of drones and stuff like that. So those are those are hive minds, you know, their swarm intelligence. I think that as swarm intelligence gets local authority to kill, that's when we're gonna say, oh my goodness, this has like gone too far. And I don't think the Iranians have the Chinese capability, but if the Chinese want to ship them a couple of shipments of uh shipments of those swarm drones, I mean I think you want to make life miserable for the Fifth Fleet.

SPEAKER_03 43:30

Okay, yeah, yeah. So I think that yeah. No, I mean those are yeah, th that's um it's scary, right? Those those those two um I mean that's that's the right answer to the question. And um geez. Um okay, last couple of questions. One what's what's what's one book everyone should read?

SPEAKER_02 43:54

Uh I really like Ethan Wallock's book, uh, Cointelligence. Uh I think Ethan's done a heck of a job there, and it's the right Title, which I love. It's not, you know, winning with AI or something. It's a cointelligence. Okay. It's like. And this is deeply human. I mean, if you go back in the history of education, you know, I've been an educator on and off of my life. And look, the long history of education for literally thousands of years was tutors. Alexander the Great had Aristotle. All the founding fathers had tutors. If you look at, you know, the Oxford and Cambridge, right? The the reason they're in colleges is that you have the tutors living there and you're having a conversation with your tutors. You're in King's College, Bellio, whatever, right? Christ College. And in the old days, the way you get a degree is you read with a tutor, and then you went through a reading period where you went and read on your own. And you came back to the tutor, and you had to make the argument to the tutor that you had a new insight in that field in order to get granted your degree.

SPEAKER_03 44:55

Oh wow.

unknown 44:55

Yeah.

SPEAKER_02 44:56

Very different. You're not sitting there with a you know multiple choice test, right? Right. Okay, then the the Prussians get their butts kicked by Napoleon, you know, Horace Mann goes over, we have industrialization happening in the US, we inherit the one to thirty classroom. And that's a terrible way to teach people to have them all at the same level, moving in lockstep, one to thirty, but it's efficient. And if you look at the American educational system, the crown jewel of the American educational system is the doctoral programs and the research. It's not the undergraduate degree. And what happens to doctoral programs and research? You're returned to tutoring. Like if you've been in a doctoral program, you're getting tutored, right? It's not big one to five hundred classes like econ 101, right? Those are moneymakers, they're not teaching vehicles. And so um, so that's gonna that's why I like about Ethan's thing, because it goes back, it invites that notion that humans want to learn in dialogue, and that's what these things can do.

SPEAKER_03 45:53

That's great. Okay, I'm definitely gonna read that. Um last question. How how can how can people follow you? How can they keep up with you?

SPEAKER_02 46:01

Sure. Um, if you go to GAI insights.com, uh, we've got a daily news show. The stuff's moving so quickly. We do a five-day-a-week news show, and we have a newsletter and the whole routine, and that that all works. And you can either watch the show live at 7 to 7:30, 7:30 to 8 o'clock Eastern every day. We have the robots go out and look at hundreds of thousands of articles, we're skinny it down, and then we review six a day with AI analysts, rating it essential, important, or optional. Um, so there's that. Um, I also have uh a Substack, John Spielka, you know, Substack, and then uh I also write a regular Forbes column. Um but at GI Insights, we have a whole resource center. We have this thing of the Rise framework, we have other frameworks, the WINS framework to look at how vulnerable you are. We have a thing called Prism, which is another framework that helps you choose among projects. And then we have our conferences. Uh we have a big conference in September, but we also are going to be participating in the MIT conference in April, and we go to Davos and so forth. So GAIInsights.com, just John at GI Insights, and also just if you put my last name in in AI, you'll you'll find me.

SPEAKER_03 47:08

I remember seeing you uh at at Davos uh last year. And you mentioned Scott Galloway earlier. I saw both of you there. I thought I followed both of you. Yeah, that was interesting. John, uh man, you are a great teacher. I uh this was a great conversation, and your answers were um were so good. So um thanks for all the insights, and I feel like we could we could have a whole series of this. Uh so great to have you on. Thanks for taking some time. Um, I really appreciate it.

SPEAKER_02 47:33

Well, Rex, it's always a pleasure to be with you. It's great to see you again. Great to see you looking so well, and uh thanks for uh the opportunity.

SPEAKER_03 47:41

You too. Thanks, John. All right, thanks everyone. We'll talk next time.