AI Restacking the Knowledge Economy with Sangeet Paul Choudary

00:00:09: Hi, this is Philip Maian, welcoming you to another episode of Talking About Platforms.

00:00:16: And we are Daniel Trebuchia

00:00:17: and Thomas Buganza, the co-host of this podcast, where we present and discuss relevant discoveries from the field of platform research.

00:00:27: In every episode, we have a guest sharing with us one of his or her latest insights on platforms, starting from a paper, a project or a book to make it accessible to everyone.

00:00:39: And with that, let's just jump right into the conversation.

00:00:49: Welcome, everyone, to a new episode of Talking About Platforms.

00:00:53: I'm Daniel Trobukki, and I'm here with my usual co-host, Tomas Buganza.

00:00:57: Hi.

00:00:59: We are very happy because this is the first time we actually have a video podcast of this podcast.

00:01:07: that is now at its fifth season.

00:01:09: It was founded with Philip Mayer and we are very happy

00:01:13: to open

00:01:14: this new season with these news of actually going in video and with a very special guest for a podcast called Talking About Platforms.

00:01:24: Yet with us today, there is going to be Sangit Paul Chaudhary.

00:01:28: And it's going to talk about AI restaking the knowledge economy presenting its latest book.

00:01:35: Probably to the vast majority of people that actually listen to this show, the name of Sangit already tells something extremely meaningful, since Sangit is one of the authors of one of the milestones in the platform world, that is platform revolution, that in mid-the- published an amazing book on AI and our AI is actually changing the systems around us titled Reshuffle.

00:02:09: Sangit has advised leadership teams at over forty-four to five hundred companies, including Nestlé, ExxonMobil, Daimler, and many other, as well as pre-IPO tech firms.

00:02:23: Sangit currently serves as a senior fellow at the University of California, Berkeley, and has spoken at global forums such as the G-Twenty Summit, World Fifty Summit, and the World Economic Forum.

00:02:35: His Harvard Business Review article pipelines, platforms, and the new rules of strategy was named one of the ten management ideas of twenty seventeen and is featured in four HBR ten mass streets anthologies.

00:02:50: His ongoing research and commentary are published regularly on his sub-stack newsletter of platforms, AI, and economics of Big Tech.

00:02:59: Let me welcome Sangit thanking him so much for being here with us.

00:03:05: We are super happy to have you here.

00:03:08: Thank you, Daniel.

00:03:09: Thanks, Thomas.

00:03:09: It's a pleasure to be here.

00:03:12: You know, people listening to our show know that we always start with the same question.

00:03:17: Then today we're going to slightly change the topic, but the main idea behind the podcast talking about platforms is having people that have to deal with platforms

00:03:27: that

00:03:28: come here, even though sometimes the meaning of platform is not unique.

00:03:34: And so it's important for us to start asking, what's your definition of platform?

00:03:39: What's the definition you refer to?

00:03:41: And also to get a bit more personal to know how you end up stabbing platforms.

00:03:47: Yeah.

00:03:48: Well, you know, there is a definition of platforms that I use in the book Platform Revolution.

00:03:54: I'm not going to use that definition here today because over the course of the almost a decade since that book was written, My view has been abstracted significantly from the mental model that I had at that point in time.

00:04:10: So the way I would explain the idea of a platform today is not so much purely in contrast to the idea of a pipeline, which is the model I use in both the HP article and the book.

00:04:25: But I'd start by saying that To understand a platform, you need to understand what an ecosystem is.

00:04:30: An ecosystem is a system of complementary activities and investments typically made by actors with conflicting incentives.

00:04:41: And the key problem in the ecosystem is to coordinate these incentives towards value creation.

00:04:47: And what a platform does is it is the most scalable and comprehensive mechanism to coordinate these activities and investments.

00:04:55: towards value creation.

00:04:58: And that to me is the real essence of what a platform does and what a platform is.

00:05:03: How it does that is by creating an open infrastructure and specifying the rules of governance.

00:05:09: But fundamentally, if you go back to why a platform exists, it exists because value today gets created through complementary activities and investments and in order to get to value creation, you need to resolve the incentives so that all stakeholders work together.

00:05:29: And so that's really, you know, at a slightly more abstracted level, that's how I would define a platform as opposed to other forms of mechanisms through which ecosystems have been managed in the past, bilateral contracts, agreements, even standards, but platforms are much more extensible for a variety of reasons.

00:05:49: What really got me interested in platforms to get to the third point of part of your question is not necessarily the idea of platforms as such.

00:06:00: The question that has guided my intellectual curiosity for the last sixteen to eighteen years ever since I started working has really been just this, what differentiates winners from losers when new technological shifts happen?

00:06:17: And very often we think that winning with a new technological shift is about adopting the technologies, about getting really good at using the technology.

00:06:29: What my research has taught me is that most companies win or lose not because they get really good at adopting the technology and get really good at executing with it.

00:06:40: In fact, the companies that lose very often get really good at executing.

00:06:45: with the new technology but within an old frame of reference and playing an old game that no longer makes sense and so they get really good at playing a game that is no longer valuable whereas the companies that win not only play the new game but more importantly they set the rules for others in terms of how they should play the new game.

00:07:04: And that is typically, you know, why the idea of platforms has been so central to my work, because one of the most compelling ways to set the rules for everybody else is to provide an open infrastructure and set the rules of governance around how you engage with that open infrastructure.

00:07:18: And what we are seeing with AI today again, and this is where a lot of my work on the shuffle also focuses on, is that in order to win with AI, stop.

00:07:29: trying to get better at executing tasks in the old frame and playing the old game.

00:07:33: And think about how you can use AI to reinvent the game and flip the rules for everybody else.

00:07:40: So that's really what has kind of driven my work all through.

00:07:44: That's the source of my intellectual curiosity.

00:07:48: And that's where the path has led me from looking at platforms to looking at what I look at over here with reshuffle.

00:07:55: Well,

00:07:57: that's that's very interesting because it's easy to know to reconnect what you're saying also with the last work.

00:08:03: so and then the books reshuffle.

00:08:05: actually what you say is that AI is a restocking the knowledge economy and I mean partially.

00:08:13: you already say that but I would like to ask you more about it.

00:08:17: so to what extent what is happening is different from the typical traditional stories of digital disruption?

00:08:24: well, I think The mechanics are not entirely different from what used to happen in the past, but there are a few factors that are increasingly becoming different.

00:08:38: The first factor is which type of economic activity is getting impacted.

00:08:45: So with a lot of digital disruption so far, we've seen the economic activity that could be clearly codified.

00:08:53: and captured in the form of structured data that got impacted.

00:08:57: So whether it was John Deere and Climate Corp changing how agriculture works or whether it is Airbnb coming in and setting up a reputation system to manage how your spare bedroom should be put on the internet, these are all mechanisms where we relied on structured information, we relied on creating some kind of standards, whether it is Stripes API or Airbnb's trust system, it's inherently creating a standard for that activity that others align around.

00:09:31: And accordingly, there were vast portions of the economy, vast sections of activity that Were immune to these technologies because they were not structured.

00:09:45: a lot of that was unstructured but more importantly a lot of that knowledge that You know gets traded and gets coordinated in other parts of the economy is tacit and what I mean by that is a lot of value that we create is not easily captured in terms of very structured data.

00:10:04: A lot of value is created by sitting in meetings, discussing with each other, capturing it on emails.

00:10:09: There has so far not been a very good mechanism to coordinate all that activity.

00:10:13: So the first thing that I think is different is just the fact that what was typically seen as more tacit, more quote-unquote human and hence away from the realm of algorithms is now increasingly also getting impacted.

00:10:30: And that's why I talk about the knowledge economy.

00:10:32: Much of the knowledge economy involves that kind of activity.

00:10:35: The second thing that I think is fundamentally different compared to previous waves is that a lot of what happened in the early two thousands all the way till the mid-twenty tens is what I would call demand side disruption.

00:10:57: shift in how you not just serve customers, not just in the parameters in terms of which you compete, but more importantly, a shift from just serving customers to owning the market mechanisms by which others serve customers.

00:11:12: And this again goes back to my point of setting the rules for others, because if Airbnb owns the trust rating system, it kind of owns the market mechanism by which everybody else serves their customers as well.

00:11:23: So the first... category of shifts happened, I would say, closer to the customer part of the value chain.

00:11:30: The second set of shifts happened in, you know, with the rise of microservices and APIs, a lot of so-called disruption happened around creating coordination across a relatively standardized activities across the value chain, which could be opened out so interfaces.

00:11:51: And whether you think about what happened with FinTech and financial services, open APIs, open banking, or whether you think about what's happening with connected cars or energy, all of that fits into that category.

00:12:02: So it's a different part of the value chain.

00:12:04: I think what's really happening today, which is super interesting, and AI is just one aspect of that, is that finally, that those levels and those cycles of disruption are coming into the inputs of the value chain because AI is really an input to business.

00:12:20: It's not an activity in the middle of the value chain.

00:12:23: It's not a market facing mechanism alone.

00:12:27: It's an input into business.

00:12:28: It can be infused into any value chain and change.

00:12:31: where power lies.

00:12:32: But then AI is just one example.

00:12:35: Improvements in biology are changing value chains as well as an input.

00:12:38: Improvements in chemistry and material sciences are changing value chains.

00:12:42: Improvements in the economics of energy are changing value chains.

00:12:45: And this is really what my next book is about.

00:12:47: It's the fact that the real capacity for innovation, the ability to exert power across the value chain and the ability to flip power structures is moving away now closer to the input side.

00:13:02: And again, those who can control all three layers are in the strongest position of what I call sandwiching.

00:13:08: But just the fact that we're moving closer to inputs that itself is a big shift from where things used to update in the past.

00:13:18: There are a few things that actually got me reading your book.

00:13:21: The first one is that is a book that talks about AI that has didn't need to be written in twenty twenty five like you read it and you don't know when it was written because you are not in what time GPT or copilot can do today.

00:13:39: So there are I would say almost known examples of of AI applications and this is.

00:13:46: Absolutely not a standard in the books around us.

00:13:50: And I think it's very cool.

00:13:53: And probably the main message that stuck into my mind is AI is not a technology that is impacting the task or the job, but is having a systemic impact.

00:14:08: You know, we are management engineers, systemic impact.

00:14:12: kind of resonate with the type of study with it, but probably is not the most direct thing to explain to someone that didn't study this type of mechanisms.

00:14:25: And the book is full of stories.

00:14:27: It's full of examples that have nothing to do with AI, but that actually explaining this systemic impact.

00:14:35: Would you mind sharing one of them to make this a bit concrete?

00:14:39: Yeah, absolutely.

00:14:40: I mean, the the motivation for the book sort of came with this idea that there's too much focus on how intelligent AI is, you know, what kind of tasks it performs, how well it does.

00:14:56: on benchmarks, you know, is it a PhD in your pocket or a five year old in your pocket?

00:15:00: It's like it's always talking about what's what's the intelligence score, right?

00:15:04: And the thing is that Intelligence is not how value is created.

00:15:10: It's another application of intelligence through which value is created in the world today.

00:15:14: And so, I don't necessarily believe that we need to even get to general intelligence in order to see massive value unlocked from AI.

00:15:22: In fact, whatever AI can already do, whether it's twenty twenty or twenty twenty five, whatever it can already do is massively underutilized because we haven't yet reorganized our systems around its capabilities.

00:15:35: And so that was my key motivation to write this book.

00:15:37: To make this point, I take a range of technologies as examples which are actually not intelligent at all but fundamentally transform the world.

00:15:47: i take the example of the barcodes and how they transform retail to radio communication and how it transformed competitive advantage for germany in the second world war.

00:15:57: but the most interesting example personally to me because i've lived very close to it and you know i spend the most time.

00:16:04: kind of wrestling with that example to get that into the book is the example of the shipping container.

00:16:09: Because nobody thinks of the shipping container as a way to explain the impact of AI.

00:16:14: Typically, when we talk about, you know, you need to think about the bigger impact of AI, we go back to electricity, we go back to, you know, something of that sort and how electricity changed, you know, factories and cities.

00:16:25: But the shipping container is interesting, because the shipping container probably had as big, if not a bigger impact than any other technology out there.

00:16:33: But It's not even something we see as technology.

00:16:36: And that's what's interesting about it.

00:16:38: It's just a steel box.

00:16:39: But the reason it's interesting is because the shipping container transformed not just how ports worked, not just how logistics worked, it transformed the world through globalization.

00:16:51: And the reason it was able to do that was because before the shipping container, cargo used to be transported in break bulk, which essentially meant that every Unit of cargo would have a different size and shape and so there was no standardization on how cargo was moving and Accordingly there was no standardized way of moving cargo.

00:17:11: So you had manual labor because again manual labor is very good when interfaces are not standardized.

00:17:18: Whether we're talking about cargo or we're talking about APIs, you know manual labor is very good when interfaces are not standardized.

00:17:24: and so dock workers used to move cargo up and down from the ship.

00:17:28: and because manual labor was involved and cargo was not standardized.

00:17:34: There was no reliability on how fast cargo would move.

00:17:38: What the shipping container first did was that it standardized the interface of the cargo and allowed cranes to move cargo on and off the ship.

00:17:48: So the first order impact of the shipping container was automation.

00:17:51: It resulted in port automation.

00:17:53: And today, when we think about AI, we think of automation again.

00:17:56: But if we just stop at port automation, we miss out on what the shipping container eventually did.

00:18:01: The real value of the shipping container happened with a second step, which is when trucks, trains, and ships agreed to use a standardized set of measurements around the shipping container so that it could be easily moved from one form of transport to another completely seamlessly.

00:18:20: Alongside that, They also agreed on a unified contract, which then meant that any cargo from source to destination could be moved smoothly.

00:18:28: If you did not have these two important developments, the world would have evolved very differently where ports would have become much more powerful and inland economies would have become much less powerful.

00:18:41: But what these two factors did was they enabled intermodal transport, which enabled every part of the economy to have relatively equal access.

00:18:50: to shipping capabilities.

00:18:53: Now, what that did then in terms of, you know, further effects is actually all the more interesting because once you could move something from source to destination reliably, that meant that you did not have to worry about managing unreliability.

00:19:09: And we don't understand this, but most of our, a lot of our business investments go into managing unreliability.

00:19:16: Before the shipping container, the The way we made choices was that factories and manufacturing in itself was vertically integrated.

00:19:27: Even if you had suppliers, they would be locally co-located.

00:19:30: So you would not have supply chains which were spread across the world.

00:19:34: The shipping container allowed manufacturing to be unbundled and supply chains to be spread out across the world.

00:19:39: This enabled the rise of economies like China in response to this.

00:19:45: But more importantly, it changed.

00:19:47: where innovation happened in manufacturing.

00:19:50: Because now that manufacturing could be unbundled and global, you could have component manufacturers compete with each other to improve individual components.

00:19:58: In the past, there was no incentive to do that.

00:20:00: And so when component manufacturers competed on the performance of each component, component performance improved, which then meant that the product creators now had access to better performing components or product innovation improved.

00:20:13: That's what gave us, for example, Intel is the classic case of, IBM vertically integrated, moving to Intel as a component manufacturer, and then opening up innovation across the PC value chain.

00:20:24: And that's what we see with Nvidia and others today as well.

00:20:27: But all of this traces back to what happened with the shipping container and related technologies like ERP coming out around that time.

00:20:34: There's another interesting aspect of this that played out, which was that there were a lot of middlemen who were just focused on managing this unreliability.

00:20:43: But with Shipping becoming the liable, you could now have just-in-time inventory.

00:20:48: So you did not need those middlemen.

00:20:50: You did not need stocked inventory at your end to manage that under liability.

00:20:56: So the key point that I'm trying to make with this is that very often we look at AI and we say, is it going to take away a job?

00:21:04: And that's what dock workers would have said when the shipping container came.

00:21:09: Workers in China would never have thought about this is going to give us jobs or new jobs or the middlemen in the retail value chain would never have thought this is going to take away our jobs.

00:21:19: So yes, new technologies create jobs and take away jobs.

00:21:22: But the effects that are most interesting do not happen through automation or augmentation.

00:21:26: They happen through a fundamentally redesign of the system.

00:21:29: And that actually creates fundamentally new value.

00:21:34: and destroys jobs.

00:21:35: It's not a zero sum way of looking at what you do today and seeing if AI takes that away.

00:21:42: That's not how things work.

00:21:43: That's why I really like the example of the shipping container.

00:21:45: It brings all of these ideas to life in a very powerful way.

00:21:50: I enjoyed the book as well.

00:21:53: The way you've got to frame the program and explain things is very interesting because, as Daniel said, you're not doing it by looking at the moment.

00:22:04: So what is happening, what is the last feature of an AI tool, but in many cases you link it back to different examples in different industries in different times.

00:22:17: And I think this is very valuable to extract the fundamental rules and the mechanisms instead of flashing in the eyes of the readers with something that is very new and also it's gonna last for.

00:22:32: two hours that you can easily read in a book of extracting the real mechanisms and the real real values.

00:22:42: and everything that i enjoyed a lot is that you know when you read the book.

00:22:47: personally i love when you find something that is somehow counterintuitive.

00:22:53: or that it is going against something that is widely accepted and everybody agrees upon.

00:22:59: And well, I think that we all, you know, heard the sentence that AI will not replace humans, but humans who use AI will replace humans who don't.

00:23:08: This is a typical sentence that everybody knows about.

00:23:11: And after a while, we probably stop our critical thinking and we just accept it because it makes sense because it's a nice story.

00:23:20: But then It's a matter of going behind the scenes and see if it is really what's going to happen.

00:23:26: So what is your take on this?

00:23:28: Yeah, I think that's something that when I first heard it, it felt inherently wrong and I did not have the language to explain why it was wrong.

00:23:41: So I went down the rabbit hole of trying to explain why this actually is not right.

00:23:46: Well, the thing is that it's what I call it true but utterly useless in the sense that eventually yes somebody using AI will take your job but knowing that is useless because it will happen in a completely new system and your job will not even look like the job that that it's there today.

00:24:06: but the problem is the reason.

00:24:08: this is you know I call it true but utterly useless is because there's an element of truth to it.

00:24:14: it creates a lot of confusion and people assume that what it means is that you actually need to start using AI and that's the way you will get to keep your job or you will be able to future proof your career and so on.

00:24:27: The story that I really like over here and I looked through many different stories to figure out what's the best match but the most interesting one that I found was this story of what happened to typists when the word processor came around.

00:24:43: And the reason this is important is that Word processors, when they came out, somebody would have told the typist the same thing.

00:24:52: You know, word processors won't take your job, but a typist using a word processor will.

00:24:55: And so the natural response then is, let's reskill.

00:24:58: And that's what everybody says today.

00:24:59: Let's reskill.

00:25:01: The question then is, you know, the skill towards what?

00:25:03: What's the future going to look like?

00:25:04: Well, if you look at the typist example, there are some very interesting things that happened over there.

00:25:09: First, the word processor did not take away the act of typing.

00:25:14: So automation did not play out.

00:25:16: In fact, augmentation played out.

00:25:17: The word processor made typing easier for humans.

00:25:20: All of that does not mean that the typist actually got to keep their job.

00:25:24: And the reason that did not happen is because very often when we use words like typist, painter, plumber, we are assuming that what defines a job is the task associated with the job because the name itself has a task.

00:25:36: Type, paint, plumb, you know, the verb itself, the task is in the name.

00:25:41: But jobs are not structured around tasks.

00:25:43: And that's the fundamental mistake we make when we think about jobs.

00:25:46: Jobs are structured around managing a constraint in the system.

00:25:50: And the constraint that typists were managing was that before the word processor, editing a typed document was very expensive.

00:26:00: And because editing was expensive, people who were really good at typing with very low error rates were highly valuable.

00:26:08: Now the moment word processors came out, the cost of editing collapsed because you could just keep editing live in the workflow, which meant that you did not have to be very good at typing in order to be a typist because the main constraint went away.

00:26:21: And so what typists did not realize was that their job was not about typing, it was about managing the cost of editing.

00:26:27: And when that cost collapsed, their job did not make sense at all.

00:26:31: And this is the thing that we need to understand when we look at how AI will impact our jobs today.

00:26:38: We hear advice like do what AI can't yet do.

00:26:42: Look for the cracks that AI can't fill.

00:26:44: I mean, that's sort of a losing game because A, it assumes that your jobs will always be stable no matter what happens and you just have to keep fighting AI.

00:26:53: And the second, it assumes that AI will the main at the level at which it is today and which, you know, if you look at the fate of the prompt engineers who came out all over LinkedIn a few years back, we know what happens to that, right?

00:27:05: So I think it's very important to think about what is the new system going to look like?

00:27:11: What were the assumptions?

00:27:13: I use the term constrained, but you know, just think more broadly, what are the assumptions around which your current job is structured?

00:27:18: do those assumptions still stand in the future.

00:27:21: And if they don't, you'll have to rethink what your job looks like in the new system and what that new system looks like.

00:27:27: And a very simple thing that I'll just point out over here is that a lot of our jobs are structured around the assumption that access to the performance of knowledge work, not access to knowledge, but access to the performance of knowledge work.

00:27:38: has been scarce.

00:27:39: In order to perform knowledge work, you need to hire people and you need to train them or you need to hire expensive people who have been trained elsewhere.

00:27:47: So whether in time or in money, it is expensive.

00:27:52: But what AI does is a lot of knowledge work, it makes it more cheaply available.

00:27:57: So that constraint goes away.

00:27:59: How does that then change the system?

00:28:00: How does that then change jobs?

00:28:02: So don't just look at tasks, look at what constraint goes away.

00:28:07: Wow.

00:28:09: You say that my favorite example in the book, you know, the ship because the ship container is at the very beginning.

00:28:16: So it stacks as when I read the typist one, I was like, okay, with this one, anyone can understand what the book is about, what's the high level message of the book.

00:28:29: You know, we were discussing just a few hours before the results of a survey we did with with a colleague of ours.

00:28:42: And some of the findings, some of the things that are emerging on how people use AI are easily explainable by previous theories in the field of innovation.

00:28:55: When an innovation enters the market, people tend to substitute technological substitution.

00:29:00: You tend to repeat what you were doing before, and you need time to adapt to that.

00:29:05: And I think your book is amazing.

00:29:09: in describing this mechanism.

00:29:11: So a huge part of what we are living already happened, but it happened in the past, so we don't remember about it, or even if we remember about it, it was just too long ago and now we live in a new norm.

00:29:27: The same time, my feeling, but I wonder what's your opinion on that, is that the AI revolution, if we can call it like that, still holds some peculiarities.

00:29:42: that is probably making this revolution coherent to a certain extent with what happened before, but don't somehow unique.

00:29:52: What are the features that you see as unique in this?

00:29:56: Mine a bit more because they are unique of what is going on.

00:30:00: Yeah, yeah, and I called some of these out in the book and There are a lot of similarities with what's happened in the past, and that's why I take shipping container and barcodes, because we haven't even yet seen those effects play out.

00:30:14: But in addition to that, there are a few other factors that make the current revolution very unique.

00:30:22: The first is just the fact that unlike many of these other technologies, it's not a static technology, it's a constantly evolving technology and a learning technology.

00:30:31: So that constantly shifts the frontier of what it can be applied to.

00:30:36: The second is that given the heavy investment and the fact that we live in an era which is attention poor and capital rich.

00:30:48: Now, the reason that is important is that when you have and attention poor and the capital rich either.

00:30:53: The way to attract capital is not necessarily to show the terms in the short term, but to create hype that attracts attention in the short term.

00:31:02: Because what is scarce is really attention and capital follows that.

00:31:05: And that is why the level of hype that we see is not a bug.

00:31:10: It's a feature of the, you know, the AI we live in.

00:31:13: The same thing that causes misinformation today to spread faster than the truth is what causes hype to be more effective than actual business results in the short term.

00:31:23: And so, given those things, we see greater concentration of capital.

00:31:28: towards wherever hype can be generated.

00:31:31: And the reason that is important is that with higher concentration of capital in some parts of the economy, and in fact much lower concentration of capital and relatively poorer economics of performance in the other parts of the economy, there's a growing divide between these two parts.

00:31:47: And this becomes interesting because if you take any value chain in the knowledge economy today, The amount of capital and the rate of innovation that's happening at the level of the AI models part of that value chain versus the service providers who are using those models to deliver solutions is very different.

00:32:03: And the rate of change is very different.

00:32:05: If you're a consulting firm or an agency, you change in weeks and months.

00:32:09: Models can improve in matters of seconds and minutes.

00:32:12: And so the rate of change is very different.

00:32:17: I refer to the idea of clock speed from Charles Fine, which essentially shows that this differential rate in change creates a fundamental power pull towards the higher clock speed layer.

00:32:34: In this case, the AI models and the AI tool providers.

00:32:40: So I think that's the second thing.

00:32:41: that's quite different from previous tool providers who even if they were very widely accepted, did not ever have this kind of power dynamics over the players who were using the tools to create solutions, right?

00:32:57: And the third thing that is what I mentioned towards the beginning, unlike a lot of other inputs to business in the past, traditionally inputs to business were commodities.

00:33:11: So whether you talk about Minville, so you talk about you know, electricity, a lot of these were eventually commoditized.

00:33:20: And so competitive advantage was never with the input.

00:33:23: AI is not a commoditized input.

00:33:25: And the reason that's interesting is, and it's not just about AI, the fact that materials, biology and energy are programmable means that they are not commoditized as inputs either, which essentially means that for different use cases, the input can be morphed to serve that particular use case better.

00:33:46: And that gives the input layer control over a vast portion of the value chain much higher and much closer to the customer.

00:33:53: And so that's, again, something that's very different from what's happened in the past.

00:33:55: So to me, those are three things which are very different from previous technological devolutions that we haven't seen before.

00:34:05: Wow.

00:34:06: I know that

00:34:07: we invited you

00:34:08: because the book in the book is mainly about AI, but the reality is that it doesn't really happen that often to have the opportunity to talk to somebody who's an expert with a different view and take on AI, but also be expert on platforms.

00:34:23: And so I would like to try to reconnect these two things basically because I'm very curious.

00:34:30: Uh, you know that our research is all about, you know, platforms and legacy firms and how the existing firms can actually, you know, try to, uh, to apply some platform models and platform thinking in order to, uh, to reborn and to use the assets that they've got.

00:34:49: Uh, but what we are observing in our research is that, uh, you know, these companies are struggling a lot.

00:34:56: So it's very difficult to.

00:34:59: fully embrace and understand the platform mechanisms and the artist is there still, you know, making confusion between a digital service and platform and then they just, you know, are just surprised because it's not scaling or is not, you know, growing is not involving in ecosystem and so on.

00:35:16: So the situation that we are looking at is legacy companies are struggling with this.

00:35:22: And now the AI comes on top of it.

00:35:26: And the question is what's going to happen?

00:35:28: So what is your idea?

00:35:31: So is this going to be just another complexity layer that is making their own life even more miserable or difficult because they are already in the middle of it?

00:35:40: Or is there any way that AI might become a way to... you know, somehow to make it smoother this transition towards a platform approach for legacy companies.

00:35:55: Okay, that's a question with a lot of layers.

00:35:57: So I'll pick a few of them, which I believe are important, right?

00:36:00: So the first layer that I'll just pick is, you know, why do legacy companies struggle?

00:36:06: Very often the most common and, you know, I would call it a cop out answer that's given is that Large legacy companies are bureaucratic.

00:36:18: They are slow-moving and everything is very slow.

00:36:21: They don't have a culture of innovation and the startups are much faster.

00:36:25: and It's partly true, but again, I'll say it's true, but utterly useless.

00:36:29: There's not much you can do with that, right?

00:36:31: We need to look for more clues into when legacy companies actually win and when they don't win, right?

00:36:38: So the second thing that's important in transformation is to understand that every company has a certain architecture of value creation.

00:36:48: And what I mean by that is that there are a set of interconnected assumptions that have been built over time into how the business works, who it's serving, what it is competing on the basis of, and the entire business is structured around those assumptions.

00:37:03: When even one of those assumptions changes, you cannot reinvent the business overnight because there are multiple other assumptions closely interlinked and built around, you know, working together to create the powerhouse that that business is.

00:37:17: And, you know, while there's a lot of talk about Christensen's, what is called demand side theory of disruption, there's very little talk about another set of theories that came out of Harvard Business School around the same time, which, you know, I go to some extent in reshuffle as well.

00:37:35: I would say the key proponents of what I would call the supply side theory of disruption where Henderson, Clark, Baldwin, and the key ideas over there are about this architecture.

00:37:48: How do you set up your capabilities?

00:37:52: How do they interact with each other?

00:37:54: And the key issue that I see with technological shifts is that there's a certain level of lock-in that the previous architecture creates, which you can't just move away from or work away from.

00:38:12: I'll give a few examples over here to illustrate this very, very clearly.

00:38:16: Take the example of Adobe.

00:38:18: It does not fit into this logic of slow-moving incumbent because it's actually a very fast-moving incumbent.

00:38:24: It moved to the cloud very well.

00:38:26: It's now embraced AI, but all through, it's a few steps behind Figma.

00:38:31: It's never able to catch up with Figma.

00:38:33: And the reason for that is that architecturally, when the shift to cloud happened, the nature of the business model completely changed with that.

00:38:43: And it's not about shift from selling one-time software to selling subscription software.

00:38:48: The shift was much deeper.

00:38:50: And the shift was essentially this, that Adobe sees its business as helping designers with execution.

00:38:58: and Figma sees its business as helping all the workflows into which design is an input, governing all of those workflows.

00:39:06: So there's a difference.

00:39:08: You're not serving designers.

00:39:09: You're not helping with execution and you're seeing design as an input into larger organizational workflows and hence governance is important.

00:39:16: Now, why did Figma get to this point?

00:39:18: The reason Figma got to this point is because with the cloud, you could unbundle the design file.

00:39:25: So Adobe's design, even when it moved to the cloud was still structured around the design file and the logic of the design file.

00:39:31: And it just helped designers on a new medium.

00:39:34: What Figma did was it took the design file, it unbundled it into its various elements and allowed each element to be individually addressable.

00:39:40: And the reason that's important is because then when those elements get used across the organization, they can be managed and governed centrally in the library.

00:39:48: And that shifts.

00:39:50: competition from, we'll help you execute design better, to we'll help you govern how design fits into all of your workflows internally, right?

00:39:59: And so that's something that Adobe simply cannot compete with because its architecture is optimized for a different way of competition.

00:40:06: And we see this again and again, you know, if you look at Stripe, it's done the same thing to payments.

00:40:10: Previously, you know, payments were much more bundled around specific use cases.

00:40:15: So expense management was ramp, if you will, right?

00:40:19: management is different and you know vendor negotiations is different.

00:40:24: and I'm taking the example of ramp here which is essentially managing internal finances.

00:40:29: both stripe and ramp.

00:40:31: unbundled the payment or the financial transaction into the individual transaction.

00:40:35: And then they restructured every workflow around governing those transactions.

00:40:40: And we see the same thing, you know, with Roblox versus traditional gaming studios, which are structured around creating the game versus Roblox is structured around enabling modular game designed by anybody because the unit of innovation is not the game.

00:40:56: by a studio, but an element of the game by a user.

00:40:59: So what happens in all these cases is that the incumbent sees what's happening, but they are unable to copy or replicate that, not because they are slow, not because they are risk averse, but because they're architecturally locked in to a previous architecture.

00:41:16: Now, all of this kind of brings us back to your question.

00:41:21: Does AI offer a better advantage to the incumbents compared to what we've seen so far.

00:41:30: And I think the answer to that is that there's definitely a lot of value, or let's think of three layers of value unlocking that can happen with a new technology.

00:41:44: First, there's value to be unlocked by just adopting the technology to play your game as you play today, which is really the task mindset.

00:41:52: Everybody does that.

00:41:53: everybody did that earlier as well, right?

00:41:55: That did not really help you move to platform business models or new business models.

00:42:00: fundamentally It just helped you execute faster and eventually you might have been commoditized or not depending on you know other advantages you had.

00:42:08: the second way you can think about the impact of a new technology is that you use the technology to reimagine you know, the basis on which you compete, which is that you actively commoditize a certain part of the value chain because of the capabilities the new technology gives you.

00:42:29: A really good example over here is what happened in India when the Alliance Geo entered the telecom industry, when the technology shifted from three G to four G. And because four G allowed voice to be carried over data, it allowed reliance to subsidize voice and get four hundred million users overnight.

00:42:51: And the dominant value capture model of traditional telcos was commoditized because everybody made money on the voice before that.

00:43:00: And of course, there were other advantages, how the lines use capital, how it uses political connections, but all of that notwithstanding the key point is that you can use new technology to commoditize the existing game to to to change the rules of the existing game by commoditizing today's value and finding value somewhere else.

00:43:19: And then the third thing is really where the architectural piece that I talked about comes in, where if you really want to be sort of what is today called AI first, if you really want to maximize the value of playing with the new technology, if you really want to reimagine your business, you need to see if your architecturally locked in or not.

00:43:40: And that That is the place where the value gets really truly unlocked.

00:43:48: And so I would say there are these three levels of value unlock.

00:43:51: In my experience, most incumbents start with one, many stay at one, some get to two, very few actually get to three.

00:44:00: And so you need to see how far up the spectrum you can move.

00:44:04: I feel that everybody can move up to two if they think.

00:44:07: creatively and intelligently about the opportunities for business model redesign.

00:44:13: But everybody has an opportunity to move in that direction.

00:44:20: You kind of anticipated a few elements that I was going to ask.

00:44:25: While I was reading your book, there were many things resonating.

00:44:31: And at the same time, I was reflecting on the different discipline.

00:44:40: we focus on innovation management and we had a huge expansion towards leadership.

00:44:45: So the behavior of people.

00:44:48: in a way, while I would easily frame your book as a strategy book, whatever strategy still means in in in in twenty twenty five, but you're looking at the system in a way.

00:45:03: Going back to the survey I was mentioning before, We realized,

00:45:10: well,

00:45:11: we proved once again what everyone is talking about on the news in these days, the famous Gen AI paradox.

00:45:17: So everyone is using it, and basically we have a Ferrari to go to work, five hundred meters down the road.

00:45:25: So everyone is using AI to draft emails, to summarize documents, and when they go for an advance usage, they generate fresh ideas.

00:45:35: That's what's emerging.

00:45:37: And very few people are actually using it.

00:45:42: I don't want to say in relevant ways, but at least in different ways, to do something that the AI can do and other technologies couldn't.

00:45:52: And so I would like to ask you two questions nested into one, and then you pick the ones you prefer.

00:46:05: This is innovation theory.

00:46:06: This is diffusion of innovation.

00:46:08: It's difficult to accept a new paradigm.

00:46:10: So we're living something that that we're used to.

00:46:14: But at the same time, this technology is not related to the task is related to.

00:46:18: the system is changing how the different things interconnect to each other.

00:46:23: And so it's more difficult to be accepted in a new way while everyone is adopting it in a straightforward way.

00:46:34: So On the two questions are on the one hand, what would you suggest to people and or what would you suggest to companies incumbents, but even some startups probably to help them in this transition in trying to get something more out of this shift.

00:46:56: Yeah.

00:46:57: And, you know, it comes back to the word that you used.

00:47:01: Strategy right and to me strategy is just choosing your game and figuring out a way to win it right.

00:47:07: and when technological shifts happen As often as they do today You need to keep revaluating.

00:47:15: Are you playing the right game?

00:47:16: Should you be playing a different game?

00:47:17: And how do you then win it?

00:47:19: Is it you know, they've done your favor rigged against you.

00:47:22: So essentially Everything boils down to strategy Whether it's at the company level or at the individual level, I mean, when you're planning your career, you should be thinking about, am I playing the right game?

00:47:34: Right?

00:47:34: I mean, you know, if I'm a professor or if I'm, you know, a manager at a company or if I'm a consultant, whatever your role be, what is the basis on which you get to command a premium today and you get agency in your work?

00:47:51: and how are those things going to change because of how new technology is coming in.

00:47:57: So I think the starting point is that whether you are a company or an individual, start by thinking about how the rules of competition are changing.

00:48:07: Now, what that means is if you're a data analyst, you know, is a certain part of your core skill.

00:48:14: on the basis of which you were commanding a premium, is that getting commoditized because AI is coming in.

00:48:20: And if it gets commoditized, what happens?

00:48:22: Does that mean somebody much more junior can do that or with less skill can do that?

00:48:26: Does that mean somebody who is in a complementary function, who used to work with you can now start doing it directly in themselves and doesn't need to work with you?

00:48:34: What does that mean for you from the perspective of your ability to charge a premium for what you did?

00:48:41: And another practical way to look at it is the many conversations in which you have a say today and what I would call quote unquote voting power in the organization.

00:48:54: The more teams you are and the more important teams you are and the more important workflows you contribute to, the more I would say is roughly is your voting power in the organization.

00:49:03: What kinds of decisions do you make today that determine value for the organization?

00:49:07: And then ask yourself, do I still get to make those decisions tomorrow?

00:49:11: Because one of the simple things that happens when data or AI enter the system is that decisions that are being made at the edge of the organization with local data can be made at the center of the organization with centralized data.

00:49:23: So does that shift decision making power?

00:49:27: So in all of this, Think about how the game is changing.

00:49:30: Don't think about how can I do today's work better or faster.

00:49:34: Think about how the game is changing.

00:49:36: Think about it at the level of your job, your workflow, where you are in the organization.

00:49:41: But more importantly, think about what does this mean for your industry and for companies in your industry?

00:49:47: Are there fundamentally new forms of companies that will emerge?

00:49:49: Will they work according to a different logic?

00:49:52: You know, if I take the telecom example again, will... companies stop charging for voice and make money somewhere else, right?

00:49:59: So think about what's that fundamentally new logic that will come in and then think about what that means for your company and for your job again.

00:50:08: So you need to, you know, I use this framework in the book.

00:50:10: It's a very simple framework, the system of work, right?

00:50:13: Don't think only at the level of the task.

00:50:15: I mean, it's not a, you know, very sophisticated framework.

00:50:18: The only reason I use that framework is to shift our focus away from looking at the task and what AI does to it.

00:50:25: and moving to looking at the entire system, you know, how is competition changing and how will companies differentiate and compete in your industry?

00:50:33: What does that mean in terms of the capabilities that an organizational middle need to have?

00:50:37: And what does that then mean in terms of which jobs will hold value and which jobs will not?

00:50:43: So that's the key idea.

00:50:45: I mean, you need to think about technology in those terms.

00:50:50: And, you know, it sounds... metaphorical and abstract, but it keeps coming back to this question of, am I playing the right game and am I playing it by the right rules?

00:51:03: Keep asking that all the time.

00:51:07: Well, time is ticking, so it has been a very interesting

00:51:12: chat

00:51:13: with you.

00:51:14: And as well as we open every single episode, we were saying the same question, we were so close.

00:51:19: every single episode with same question which in this case is interesting because normally it's a question we don't know anything about but in this case we got a lot of knowledge about it.

00:51:28: because uh because as dana said before it's very interesting.

00:51:31: you know you added at the end of the book you know the list of things you are working on.

00:51:36: the question is actually this one.

00:51:39: so what is the next very interesting thing?

00:51:42: that is you know catching your time and attention and energy in this moment.

00:51:47: Yeah, so, you know, I, before I started writing the eShuffle, this question had already captured my time and attention and I started writing this and then the idea for eShuffle came in.

00:51:57: So, so I wrote, you know, the idea for eShuffle came in January this year and I wrote it between Jan and June and then it came out in July.

00:52:06: But last year, I had started working on this idea about understanding, you know, more broadly the nature of uncertainty because We keep saying that the only thing that's constant is change and so on, but the very nature of uncertainty is changing.

00:52:26: Traditionally, what we used to think of as uncertainty was what I would call operational uncertainty, which shows up as demand shocks, shifts in demand, supply shocks, and the way you manage.

00:52:37: operational uncertainty is you have buffers and stocks and so on.

00:52:42: Operational uncertainty at the end of the day, means uncertainty within a fixed structure.

00:52:48: What I believe is happening increasingly today is what I call structural uncertainty, which means that the structure itself is up for change.

00:52:54: So industry boundaries are no longer fixed.

00:52:57: Accordingly, how you compete is no longer fixed.

00:52:59: How you make where you capture value and where you don't capture value are no longer fixed.

00:53:04: All of these elements are up for grabs.

00:53:08: The way companies compete is fundamentally changing from better execution to changing the rules of the game for everybody else.

00:53:14: So I know I talked about that for a bit.

00:53:17: But that's really the core thesis of what I call unfair advantage, which means that we used to work with the idea of rule-based advantage.

00:53:27: Here are the rules.

00:53:28: If I get really good at doing that, I'll win.

00:53:32: But it's really now we're shifting to unfair advantage, which is you win.

00:53:36: during moments of structural uncertainty by defining the future structure for everybody else.

00:53:41: And so you get to define how others play.

00:53:43: And this is very interesting implications.

00:53:45: because, you know, a common question that I get asked by clients very often is, are economy the skill going away?

00:53:51: Is it all economy the scope?

00:53:52: And it's not as simple as that.

00:53:53: You know, that's a question like, is digital going, a physical going away that all digital?

00:53:58: or, you know, is it all data?

00:53:59: It's not as simple as that.

00:54:02: What's really, you know, what's really changing is that you can't win on the basis of execution, you have to win by changing the rules of the game for others.

00:54:10: Now, in certain cases, the economy of scale will still apply in that model.

00:54:14: In certain cases, the economy of scope will better apply, but it's not one or the other on that axis.

00:54:19: The axis that matters is the nature of uncertainty that we are dealing with.

00:54:23: So that's really what I'm looking at.

00:54:25: There's some very interesting themes in that where I talk about why we are moving back to vertical integration and why it's the simple idea of, you know, platforms are better does not apply across the board and whether doesn't apply.

00:54:39: And so there are many other aspects of that, which I look at, which essentially also links to the nature of change that I mentioned, you know, change in market mechanisms, change in the middle of the value chain to change in the nature of inputs.

00:54:55: The more you move towards changing the nature of inputs.

00:54:58: the traditional logic of platforms as we know them does not apply anymore.

00:55:02: And we see that, you know, when we look at companies that are optimized around the performance of the input, Tesla is a great example, which is optimized to extract the maximum performance from the battery.

00:55:14: And so it is very clearly because of that vertically integrated.

00:55:19: And we see that repeatedly, you know, and there's accordingly a different different maturity cycle, which then determines when such value chains modularize and how long they stay vertically integrated.

00:55:29: So that has been an interesting theme that I've been looking deeply at.

00:55:34: And again, this book is full of stories.

00:55:37: So I really love collecting stories, especially those that do not directly come from the same domain.

00:55:43: So yeah, that's something I'm quite excited about.

00:55:47: That's very very cool.

00:55:49: The last thing we would like to know from you is how can our listeners stay connected with you?

00:55:56: I mentioned something reading your bio at the beginning, but I leave the floor to you so that they will know exactly what to type whether.

00:56:03: Yeah, absolutely.

00:56:05: So I died a newsletter quite regularly.

00:56:07: It's at platforms dot.

00:56:11: I'll be launching a new website in the coming days very shortly by the time this podcast comes out.

00:56:17: It's called reshufflebook.com.

00:56:20: It essentially is a place where we'll be mapping how this idea of a reshuffle is playing out in different value chains and how the nature of the input, not just AI, but different inputs are creating different reshuffle effects.

00:56:32: So that's another place you can look at.

00:56:35: My previous work is available on platformthinkinglabs.com, which is where most of my work on platforms is.

00:56:44: pick a copy of reshuffle on amazon if you'd like to go deeper into it.

00:56:48: but those are the different avenues through which you can connect and i post quite regularly on linkedin as well

00:56:58: thank you very much.

00:56:59: you know having you on this podcast meant a lot

00:57:03: for

00:57:04: your history but i think your present is as brilliant as your legacy.

00:57:10: this book is a must have for.

00:57:13: I must read for anyone that wants to try to understand something on what is going on.

00:57:20: Thank you very much for being here with us and see you in the next episode.

00:57:24: Thank you again.

00:57:25: Thank you so much.

00:57:26: Bye.

00:57:34: Thank you for listening to this episode of Talking About Platforms.

00:57:39: You can rate the podcast on every major podcast app.

00:57:42: This helps us a lot in growing.

00:57:45: Visit talkingaboutplatforms.com to get the links of the short notes.

00:57:50: reach out to us and see the previous episodes.

00:57:53: Hear you in another episode to talk about platforms.