009 - Heads Talk - The Analysis - Katja Rieger's Analysis on Episode 287 - Prof. John Amaechi OBE

Follow me to see #HeadsTalk Podcast Audiograms every Monday on LinkedIn

Heads Talk - The Analysis - 𝐓𝐡𝐞 𝐝𝐢𝐚𝐥𝐨𝐠𝐮𝐞 𝐛𝐞𝐧𝐞𝐚𝐭𝐡 𝐭𝐡𝐞 𝐝𝐢𝐚𝐥𝐨𝐠𝐮𝐞

_____________________________________________

The Analysis is part of Heads Talk®: Nexus Rerum: Boardrooms & Statecraft. Here we extend the conversation beyond the principal exchange.

1 or 2 consequential questions from the main episode are placed before a second distinguished voice. An individual deeply embedded in the worlds of business, policy, or geopolitics. Their role is not merely to respond, but to interrogate: to examine the framing of the question, challenge its assumptions, and surface the deeper strategic and intellectual currents that may otherwise remain unspoken.

This is expert analysis, a companion discussion that offers a more deliberate and expansive reflection, where ideas are tested, perspectives are sharpened, and the dialogue evolves beyond its original bounds.

What you will hear is a continuation, not a repetition. A considered counterpoint. A deeper reading of the question at hand, a nuance.

We hope you find The Analysis  both illuminating and indispensable as part of the broader Heads Talk experience.

In this episode, we feature Katja Rieger's analysis of Episode 287 with Prof. John Amaechi OBE Episode Title: 🇬🇧Psychological Power & Generational Warfare🇬🇧

• Question 4: When nations act from historical grievance, how rational are their decisions really and how dangerous is that psychologically?
• Question 5: In your experience, does power corrupt people, or does it simply reveal who they already are?

ABOUT THE HOST

Transcript

SPEAKER_02 0:00

Coming up on HeadStalk.

SPEAKER_03 0:01

What AI can help do is it can see where teams are diverging, where teams are converging. It can you know um expose these aspects out and by converting a lot of unstructured information within teams to structured information or structured knowledge that can be shared across teams, it creates that implicit coordination without requiring all these meetings to get to consensus. So that's what it means at an organizational level.

SPEAKER_02 0:26

This is Head's Talk with me, Elaine Pringle-Schwitter, the podcast where we talk to C-level executives, leaders of institutions, and heads of multinationals. What are the current and pressing topics? Well, we like to start the conversation.

SPEAKER_00 0:40

Can you imagine getting into a business or a market where you actually spend a hundred billion plus on a piece of paper?

SPEAKER_01 0:48

Are you kidding me?

SPEAKER_00 0:49

That was like a frying pan overhead.

SPEAKER_01 0:50

I got nothing against CFS. It was not just the job of a lifetime, it was the job of a thousand lifetimes.

SPEAKER_02 1:05

My guest today is a leading scholar, author, and strategist in the fields of platform ecosystems and artificial intelligence. His new book, Reshuffle, challenges how we understand AI's role in reshaping power and knowledge across industries and societies. He combines that with a deep record of advising governments and global institutions. This episode will explore the intersection of research, policy, and real-world transformation, delivering hopefully fresh ideas and practical insights. But before we get into that, here's a brief message. This episode is sponsored by Axio. Axio is the leading private cloud platform in the Alessian and Matamos ecosystem, combining intelligent solutions with security and control. Axia's clients profit from digitalization and automation of critical business processes in a cloud and hybrid architecture. 150 staff provide migration, engineering, and support services to over 200 leading organizations in 32 countries. He is also the co-author of earlier works such as Platform Revolution and Platform Scale, which have become foundational in the study of platform economics. In addition to his writing, Sange holds uh roles as a senior fellow in competition policy at UC Berkeley and a non-residential scholar at Dartmouth. He has advised governments and regulatory bodies, spoken at forums like G20 and World Economic Forum, and contributed to shaping policy responses to platform regulation in Europe and Asia. His career combines rigorous scholarship with direct influence on public institutions and large corporate ecosystem. Let's have a conversation now. Sangeet, welcome to Heads Talk.

SPEAKER_03 3:12

Thank you, Elena. It's such a pleasure to be here.

SPEAKER_02 3:14

Excellent. At last, um this is an episode we've been waiting for. And many thanks to Paolo Saroni, the global research leader in banking from the IBM Institute for Business Value for his introduction, his uh his prompting and foresight in bringing you onto the show. And many thanks for the book um reshuffle and giving me the time to uh to digest the content in order to cover the topics today. Before we dive into the details of your new best-selling uh um book, uh Sankey, you've been writing about platforms and ecosystems for years. But what made you decide that now was the perfect time to write, reshuffle, and focus on AI's impact on the knowledge ecosystem?

SPEAKER_03 4:03

Yeah, Elaine, uh the guiding principle across all of my work, whether it was the work on platforms and ecosystems or the recent work uh in reshuffle, uh, the guiding principle has been understanding and explaining why certain companies win and others lose, not because they don't adopt the technology. It's not technology adoption that differentiates winners from losers. What really differentiates the two is adopting new technology, but continuing to play by the rules of the old game when the shift in technology has actually shifted the nature of the game. And so, really, what I do with all of my work, uh, what you've seen with platform revolution and what you see with Reshuffle now is helping clearly contrast what the rules of the old game were and what the rules of the new game are, and why companies that continue to operate with the wrong mindset and the wrong mental models uh end up losing, not because they haven't adopted, but because they have failed to uh uh you know, play start playing with the new rules.

SPEAKER_02 5:07

Okay, thanks for that. Um okay, so let's dissect your book. Um, you just talked about it. You said in Reshuffle, you argue that success now depends on playing the right game rather than just using AI to play um yesterday's game better. I mean, this all sounds a little bit uh like a riddle. You've sort of half-bought it to introduce and talk about it, but I'm sure you'll explain it better now. So um so so for AI strategists and executives and the leaders that you talked about, how do you determine what game should be played in this shifting landscape? How can an organization start when the rules of competition have changed versus when you know it's safe to double down and improve existing processes?

SPEAKER_03 5:53

Right. That's the central uh argument that I bring forth in Reshuffle that today a lot of our framing around AI is what I think of as very task-centric framing, which is we think about AI as a way to speed up today's tasks, to make them cheaper, better, uh faster. And accordingly, our uh view on how AI impacts jobs and organizations is built around the idea of how AI impacts tasks. So we look at you know, what can AI do today, which uh humans can't do at the same uh level of performance, or which AI can do much cheaper, and we think of that as an as automation, and we look at which tasks AI can help humans do better, and we think of that as augmentation. So it's all stuck in this frame of here are the tasks, here's what AI can do to those tasks, and accordingly, your organization and your jobs will be affected based on that. Uh, what I argue in Reshuffle is that that's fundamentally uh missing what's really happening because AI doesn't simply impact individual tasks in the knowledge economy. Uh, it changes um the entire system in which those tasks exist. So, what I mean by that is that whenever uh a new technology, in this case AI, comes in, certain older forms of differentiation and uh older basis on which companies would differentiate themselves and compete with each other, they get commoditized. And a classic example of that is how various forms of knowledge work performance are becoming commoditized. You no longer need to hire uh or invest in at the same level of talent and training to get access to that knowledge work. You can get access to the same knowledge work using AI models. And so that's just one way of saying that any firm that was differentiating on that basis uh would now be confronted with commoditization on that particular access. So, really, if you think about the impact of AI, it's changing things at three levels. It's changing the basis that with which firms compete. Organizations then are changing in response to that, because if there's a new model of competition that comes in, if competition, for example, in professional services is no longer about hiring and training a lot of people, but is shifting to a new basis, then the organizational structure changes in response to this new model of competition, not simply because certain tasks are being automated or not. And then accordingly, jobs change not because today's jobs are getting displaced because certain tasks are being taken over, but because this new logic of competition makes certain jobs more pivotal and certain previous jobs which were very pivotal, less pivotal in the new system. So really you need to look at it at this uh effect uh at the overall systemic level and really start with asking the question well, why do firms exist and why do they how do they differentiate and and position themselves in the market uh rather than thinking about what AI does just to your individual tasks?

SPEAKER_02 8:51

Oh you said something very quite interesting, but let's continue with this competition stuff that you're talking about. If AI is is if AI is forcing us to change the game entirely, to change the system entirely, which you've just mentioned, does that mean that size and scale, you know, which used to be advantages, might now actually become liabilities?

SPEAKER_03 9:14

Um that that's a great question because the answer is not binary over there. Uh I believe that scale still has a very important position. In fact, uh I believe that we've kind of uh uh you know gone through this um uh learning arc where we initially moved from asset heavy to saying that, well, it's actually asset light that's important. And now I believe people gradually realize that it's not one of the two, it's asset right. So you still need scale, but not all assets are equally valuable uh or as valuable in today's game. And the first um, you know, uh factor that I would like to call out over here is the reason scale is important is because it provides you uh a certain fixed cost base, it provides you a certain level of asset intensity that prevents new players from easily entering your space. And that is how traditionally, you know, industrial scale used to work. But what's different today is that because the the nature of uncertainty is so high, that the the way companies compete change so rapidly, uh, the new technologies completely change how uh an industry is structured. Because of that, your assets have to be the purposeable. So if your assets are highly specific to a particular uh use case and the whole industry's use case suddenly changes or the basis of competition changes, and your assets cannot be the purposed in that direction, then that's you know completely useless. A very good example of a repurposable asset today, I would say, is cloud computing capacity. No matter what the end use case is, as long as you need compute, compute is highly depurposeable across different use cases. And so any company that invests in creating scalable compute is uh you know in a very, very uh powerful position in the economy. And we've seen that uh with various other forms of assets as well. Uh so the nature of scale changes. We still believe that scale is important, but uh you need to have scale in a way that it can be repurposed as market conditions change.

SPEAKER_02 11:23

I'm glad you mentioned about the cloud computing capacity because I was thinking while you were talking, I'm thinking leaders are probably saying, well, you know, how do I detect? How do I detect this shift? You know, what what kind of what are the signals? So can you share sort of a concrete signal leaders should watch for that tells them the rules of their industry are being rewritten by AI?

SPEAKER_03 11:46

Well, uh there are a few different things that um we can watch out for. So one is look for what forms of knowledge do you compete on the basis of today, you know, and what I mean by that is uh you've got really uh precise and differentiated uh access to talent, you've got specific hiring practices through which you can curate that talent, you've got specific training practices uh through which you can train that talent, and all of those are central to your competitiveness. Uh, I would first look at you know, which of those um are at risk of commoditization because of AI. So the first axis that I would look at in terms of how to think about the rules of competition is, you know, what's the threat axis? Uh, what kind of uh or what's the commoditization axis? What's the what are the uh uh existing sources of differentiation that could get commoditized? The second thing I would look at is what's the opportunity over here? If you could um commoditize a certain part of uh uh you know, certain form of uh expertise, could you then repurpose it and deliver it across industries? Uh if you think about um you know what happened with the rise of uh digital photography and how uh the um uh the the the fates of two companies that used to compete, Canon and Fujifilm, vastly diverged. You'll actually see this because uh both Canon and Fujifilm knew that technology was changing the playing field where they were playing. So the value was no longer in film uh and in printing, but value was shifting towards sharing of photos. And that's where you know Instagram and other players benefited from that. So it's it's not so much that digital cameras ate physical cameras, it's more that photo sharing ate photo printing, right? And uh it's not that Canon failed because it did not go digital, Canon really failed because it had no paths to moving towards uh uh you know photo sharing. But if you look at Fujifilm, Fujifilm realized that at the same time, its capabilities on the basis of which it was differentiating was actually chemicals. That's what it was using for printing. And now that printing was no longer a use case, it started looking at ways to repurpose its differentiating differentiation and its capabilities in chemicals. And through that, it started entering uh cosmetics, pharmaceuticals, and so on, where the same chemical capabilities were required. So that's an example of investing in scale in assets and IP that can be the purposed when market conditions shift. Um, and uh really um what this example also shows is that it's not adoption. You know, Canon was very good with adopting digital, unlike uh what the um codec, what the typical uh uh explanation goes around it, but was really about looking at what gets commoditized and how can I cross industry boundaries using capabilities that I have. That's really what separated the two.

SPEAKER_02 14:42

Okay, that's good. That's good. Uh and and for the listeners, we will put a link to um Sangeet's book so that you can there uh go and purchase the book. The the original first question, uh and perhaps some of the side questions I asked uh against that uh was from chapter 12. You don't need an AI strategy, where to play and how to win. That was near the end of the book, so it was sort of like going from the end and then we're gonna work our way back to the front again. Okay, let's look at unbundling and the building block economy. Let's look at that. You described AI as a coordination technology that unbundles existing systems, separating tasks from jobs, expertise from experts, and capabilities from companies to create a building block economy. What does that mean for professionals who specialize is who specialize expertise is being unbundled? And and what I mean by this is how should knowledge workers and and organizations adapt their their architecture and strategies in a world where skills and knowledge become modulous, scalable uh components?

SPEAKER_03 15:56

Yeah, um, that's a that's a there's a lot of detail to that answer, Elaine. So I'll break that down into three parts. Um so uh first let me try to lay out uh what uh I mean by a building blocks economy with a few uh you know clear examples. Uh then I want to talk about what that um means in terms of how you think about your um your your jobs as such. And uh within that I want to also weave in a third component, which is you know, where's the real value? If if a lot of knowledge work is going to get commoditized, uh, where is the real value that sits in such a scenario? So let's start with the first piece. Um the the key idea of a building block economy is that traditionally they were uh you know the the ownership of an asset and the ability to use that asset were tightly bundled together, which meant that if you wanted to use a certain facility, you needed to either own it or you needed to uh you know invest in it in some other form. Um what uh has happened over the last 15 years, even before you know we talk about AI, what has happened with the rise of cloud computing is if you look at cloud computing, uh you unbundle the ownership of the infrastructure from the access of the service that the infrastructure enables. So you no longer need to invest in the servers, but you get access to the same compute capabilities. And this unbundling then creates a lot of flexibility in terms of which building blocks you can use. It also changes sort of the rules of competition. If previously companies were competing on the basis of owning those assets, and now those assets can be accessed as modular services, uh, as in the case of Amazon Services, then uh that basis of competition goes away. Where this becomes really interesting for the individual, and I use this example in the in the book, is if you if you look at uh um you know the example of Mr. Beast, who which is the number one um um YouTube video channel run by Jimmy Donaldson, um Mr. Beast launched a burger chain overnight using building blocks. And uh there's there are you know positive and uh cautionary um lessons that we can draw from that. What I mean by launching a burger chain with building blocks is Mr. Beast had access to uh a large audience thanks to his social media following. What he needed was restaurants that could cook burgers, and what he needed was uh a delivery system that could deliver those burgers. And both those things are today available as building blocks. You can get access to on-demand restaurant facilities, uh, virtual restaurants, um, and you can get access to on-demand delivery facilities. And and combining that, Mr. Beast essentially launched this restaurant chain overnight, which typically for uh a traditional burger franchise would have taken a long time. Um, now that's the example of you know what you can do when you have access to building blocks. Now, if you bring that back to where we are with AI today, uh, this is an opportunity that um workers in general have that they can access capabilities to perform knowledge work, which traditionally would have required a significant amount of capital. A simple example is just using Chat GPT, right? A lot of things that you can do using Chat GPT would have traditionally required uh three or four assistants doing all of that research for you and packaging those uh outputs. And even then, um, you know, depending on how diverse your use is, it could range from three to four to thirty to forty assistants. And even then, the speed at which they would bring that back to you would not be quite as fast. So there's quite a lot of knowledge work that can get performed at a much faster rate. And uh Chat GPD is just a very basic example. There are many other AI tools today which package knowledge work and uh fairly complex workflows end-to-end to provide you the output. Now, what's important in that case is first, uh, you know, you can put on the Mr. Beast hat and think about what are the new business models that I could create, given that I have all these capabilities that I can access on demand. Just like Mr. Beast did not have to set up the restaurants and hire uh, you know, uh build up the logistics fleet. Uh, you don't have to build up that whole um knowledge-based firm. You can just access these capabilities. So, how do you bring them together to create something new? And what is your real basis of differentiation then? For for somebody like Mr. Beast, it was their YouTube following. How do you bring these commoditized capabilities to access scale, but then think about what your real basis of differentiation is? So that uh, you know, that's the key um opportunity that I believe knowledge workers have today to think about things more entrepreneurially, but then that also brings uh you know to the third part of my answer, which is uh if everybody has access to the same uh commoditized building blocks, if you will, you know, where's the differentiation? Where do you really stand out? Right. And uh the way I think about it is that if you think about how knowledge work works, uh the way the knowledge work value chain, if if you will, works is that you ask questions, you get answers, you choose which answers make sense, and then you uh, you know, uh take a bet on the right answer so that you also uh take on the risk of the outcome of taking a bet on that answer. So what I'm what I'm trying to say is that there are four uh parts to doing knowledge work asking the right questions, creating the right answers, choosing uh which answer makes sense, or you know, curating that answer, and then assuming the risk associated with this uh choice. Traditionally, the cost of creating the answer was so high that we were focused entirely on creating answers. But when the cost of creating answers goes down, we you know the focus moves more to asking the right questions and really curating the right um choices on that basis. So the key point that I'm trying to make is that if you can really think about problems that need to be addressed. for which you can leverage some of these building blocks, but it's really about how you curate the right answers and bring those together to create a new solution. That is what helps you differentiate. And I'll just wrap this up with a very s uh small example to kind of uh help land, you know, um how this works today uh in in a slightly unrelated context. Um but this whole idea of you know answers becoming a commodity, if you look at what happened to magicians, uh magic as as a trade traditionally was built around secrecy. So getting access to those answers, which is essentially you know how the trick is performed, was a secret. And the fact that it was a secret uh encouraged you to um invest a lot of effort in perfecting that sleight of hand. But today with the internet those secrets go out very easily. So it does not make sense to keep on learning new more complex tricks because you won't have the ability to benefit from its secrecy because once the trick goes out, you know the magic associated with it goes away. And so magicians have now started creating magic not with the trick but by bundling different tricks together into fundamentally new experiences using new technology, using new immersive experiences and taking fundamentally you know traditional tricks but crafting them into new narratives and new experiences so that the overall experience becomes magical. So magicians have realized that magic is not just about the secrecy of the trick it's about the overall experience and that is where we value now shifts when the the the secrecy of the trick gets exposed so rapidly.

SPEAKER_02 23:40

So those are things we need to think about you know when the value of knowledge gets commoditized what do I need to wrap it uh wrap around it uh to create new solutions that deliver the same or related value to um my stakeholders how do I need to evo I'm really glad that you gave that um magician example because I was going to ask you because I'm listening to your answer yes I understand it but you're you're you're kind of not addressing the the personal the the people the professionals angle here and um when you gave the example of the magician I just thought okay that's good you you you've kind of answered the question that I was gonna um ask as a sort of a a follow-on question to to to to what you provide as as an answer and and for the listeners that um question came about based on reading chapter four AI unbundles the job where the future of work takes a lot more than just staying a guild so thanks for that and Sangi um now this one you should you introduce a key choice between coordination with consensus like shared standards and AI enabled coordination without consensus you need to explain this one um to me and Sangeet could you illustrate what coordination without consensus look like in practice? For example how might AI help different teams or companies align on goals without a formal shared protocol and and what are the trade-offs in terms of speed capital or strategic control?

SPEAKER_03 25:15

Yeah this is a um this is a point that I wrestle with quite a bit because I wanted to clarify where AI is different from um how today's technologies help us get work done and um um we can you know look at what this means at two different levels organizations and at the industry level but I'll just focus on organizations to keep it um you know focused on that and then we can open up other parts of it uh in subsequent questions but um the the key the key idea here is that organizations today um need to coordinate between teams to get work done so um every organizational system whether a traditional organization or an external loose collective of you know teams you know it's fundamentally built of teams who are performing certain uh components or parts of workflows which need to come together to get the entire work done and that creates a fundamental trade-off in all organizational systems between the autonomy of the team and the coordination between uh and across different teams because uh the more autonomous a team is the more output it can create but it also makes it less coordinated because if it's focused only on its own output it's not coordinating with other teams and the more you invest in coordination the less the more you take autonomy away from them. And so that's the trade-off that we see today. The way we solve that trade-off is through very old technologies. You know the most common technology to solve this trade-off is essentially meetings. We call up a meeting every time we want to make some coordination happen we every time we want to share information and knowledge across teams or align our incentives and our goals and all of that meetings create what I call a coordination tax on the organization. It's all of that time, all of those resources not just into meetings but also in documentation and creating internal knowledge so that knowledge gets transferred across the organization. Now where AI is really powerful is it takes away a lot of this coordination tax because of a couple of factors. First a lot of this internal knowledge creation can be automated to a large extent using generative AI today because you have the ability with technology for the first time to understand unstructured information in the form of meeting notes, meeting call recordings, etc and and just a lot of different notes from different parts of the organization make sense of all of that and create an institutional level body of knowledge across the organization which can then potentially be served and into the right workflows at the right time. And of course you need structured knowledge bases. I mean we see companies like Notion today which started out trying to create an infrastructure for helping organizations create structured knowledge bases and then they are now launching agentic capabilities on top of that so that all of that knowledge can be used by agents. What what all of this does is that if you if AI can help us get to creating internal institutional knowledge aggregating and synthesizing it across the organization across different teams and then agentically serving it at the right place at the right time in the right workflows not necessarily AI executing the work but humans executing the work but accessing the right knowledge at the right time at the right place because there's a brain working alongside them that dramatically takes away all of the coordination tacks overheads that are associated with knowledge management and organizational coordination today. So that's really you know what I think about as coordination without consensus in the context of an organization because meetings and getting onto the same page and you know agreeing on certain things, these are all very much coordination with consensus. But what AI can help do is it it can see where teams are diverging, where teams are converging it can you know expose these aspects out and by converting a lot of unstructured information within teams to structured information or structured knowledge that can be shared across teams it creates that implicit coordination without requiring all these meetings to get to consensus. So that's what it means at an organizational level.

SPEAKER_02 29:36

Let me keep throw one here based on what you said who wins who wins if coordination without consensus becomes the default who really ends up holding the power the companies building the AI layer or the organizations plugging into it?

SPEAKER_03 29:57

No that's a central tension that I believe we'll see across the AI economy um which is um in certain cases you will have tool providers who are building the AI exert a lot of um power and in certain other cases it will be the companies that are using these AI tools to solve customer problems. So what really determines uh which of the two actually win is the amount of value that you add as a solution provider on top of these AI tools. It goes back to the building blocks uh question that we talked about earlier if all of these uh capabilities are available as building blocks uh what's your right to play on top of that and your right to play is essentially understanding the customer you're serving very deeply and then assuming the uh you know guaranteeing a reliable solution to that customer because what AI tools do really well is they perform work. So they're very good at performance but performance is not the same as reliability. So if you take the example of a professional services firm their customer might care about uh you know the the sensitivity of certain topics and um knowing uh the level of risk associated with certain choices um as an as an AI tool provider um you could provide all the you know uh you could perform all the background analysis to get to those uh choices but the actual uh risk associated with it and the safeguards associated with it uh that is what uh might be very domain specific context specific and that is what a solution provider on top would uh provide so uh to the extent that you know you as a company who's providing solutions to end customers using underlying AI tools can manage that reliability for your customers um then you have a strong position if on the other hand you are relying entirely on underlying AI tools and you're actually getting excited that hey um we're moving from GPT three to four to five and uh there are you know there's all of these improvements in the underlying models that we can just use to answer our to to fulfill customers' needs better. And if you're constantly delivering your differentiation based on what the underlying tool provides you're essentially just becoming a channel for that tool. So you're not really creating sufficient differentiation on top. So that's really the key because the challenge is that when we get obsessed with what AI can do and the fact that the companies providing AI tools are so well funded, we fall into this trap of building around the advancing capabilities of AI and not building sufficient value of our own on top of it. So that's when you end up potentially in a place where you could get commoditized by the underlying tool provider because all your competitors will be doing the same thing. There's no way to differentiate all differentiation is captured by the tool provider. Let's talk about some differentiations I'm throwing a bit of a sort of a curveball probably because I'm still my my head is sort of still in the geopolitical space because some of the conversations I've had recently do you see this playing out differently across regions say the the US where the big tech dominates versus India or Europe where ecosystems are more more how would you say fragmented I think there's there's certainly um a a complementarity between um you know solutions that solve fragmentation at a fundamental level and uh or or or uh you know technologies that solve that fragmentation versus technologies uh that can create uh new intelligence on top of it so I'll explain what I what I mean by that if you look at for example India and uh 15 years back in 2010 India was extremely fragmented uh fast forward to today or you know what actually started five years later 2015 onwards uh when India started getting onto what is now called the India Stack a common identity layer with Aadhar and uh a a common uh um you know payments capability with UPI and so on uh a lot of that fragmentation has gone away and given uh rise to um different forms of standardization uh which in the case of India is not so much driven by the big tech but it's driven by uh what is called digital public infrastructure it's uh um a public-private uh driven initiative based on which um a lot of this fragmentation is getting solved so my point in in that case is that um to the extent that those fragmentation problems are being solved and and they are being solved in different ways across geographies the more you solve those fragmentation problems um the more you create the conditions for um uh an economy where you can drive concentration at layers about the underlying infrastructure that you create. So even if I take the India example the underlying infrastructure is being created as a public utility but there are companies that are building on top most notably uh the Lions Geo which has uh you know concentrated most of the uh mobile internet population across the country and associated services that they get deliver to those uh um populations um you you start getting these uh uh this potential for concentration on top of these underlying layers once standards are built at the underlying layers so uh I I I believe that you know all economies today understand the importance of moving away from fragmentation many are bothering the India model uh some have in the past tried to bother the China model um others you know in one way or the other uh get onto the um US big tech model uh so we are moving in a direction where economies are moving towards uh moving away from fragmentation and so the the the potential for creating this form of concentration or uh on on top of such an economic system uh progressively increases but to the extent that fragmentation exists today um you know there there are more fundamental issues to be solved before um you can have a a true AI uh any one economy okay um you classified AI integration into uh three roles tools engines uh or infrastructure so so so tell me tell me what would distinguish a firm that is merely using AI as a tool from one that is building AI driven infrastructure for um the ecosystem and why does it matter? Yeah I think the key point um that I'm trying to make with that is that there are two ways to use AI and this goes back to the task versus system distinction that I made initially you can either use AI to speed up uh um how you already work or you can reimagine your entire business and the logic of your industry around uh uh you know AI. And I'll give a couple of examples to illustrate this and to make this very tangible uh think of what happened with social networks uh an example that I use in the book um when um you know just before TikTok came about uh we were at a point uh in the evolution of social networks when Wall Street largely believed that the um dominant social networks uh Instagram YouTube Facebook etc could not be uh dismantled uh just because um they had very powerful network effects and more importantly because they were structured around the logic of the social graph which is that for a network to be valuable you needed your friends to connect with you uh your friends and your followers so the more uh so the the the more you could build the social graph around yourself the more valuable the network was and hence you would want to stay there which meant that it was very difficult to bootstrap bootstrap a new network when the incumbent networks like Instagram would not pre you know would not allow you to move your social graphs. And what TikTok did was it basically created an AI uh you know a social network using AI as an engine uh by making the logic of the social graph irrelevant. So what TikTok did was it instead created a behavior graph uh which was built around uh you know looking at what users were doing and identifying what they were interested in on that basis and accordingly uh linking them to other pieces of content that would match their interest and then over time if as a user I saw a lot of interesting content from somebody I would just start following that person and so the social graph built out over time. But in order to get to this uh they had to fundamentally reimagine what a social network was uh in order to uh remove the logic of a social graph and uh you know build out network activity from all the signals of how users were interacting with videos they needed to a um limit the video size so initially videos were only 60 seconds long uh they needed to um track every single type of interaction with the video you know what was being forwarded what was being swiped and so on um and they needed to capture all these micro interactions uh and hence the video size had to be chunked so that in a particular session a user would go through a minimum number of videos give all of that data to TikTok and on the basis of that the behavior graph could be created. Meanwhile Instagram and others were still stuck to the social graph logic and using AI to improve recommendations within the social graph. So that's a very clear example of you know Instagram using AI as a tool and TikTok destructuring social networking around AI as an engine and that allowed it to fundamentally change the basis of differentiation away from social graphs change the basis of competition that you could now compete you know uh as a social network even if you did not have users on board you could compete as a creator even if you did not have followers uh you could see value as a follower even if you did not follow anyone so the fundamentally the entire logic was completely flipped and that's the real idea of AI as an engine you need to reimagine around what is possible with AI thanks for joining me today on this episode of Heads Talk don't forget to subscribe to the show via my website ellaineprinkle.com forward slash heads talk wherever you get your podcasts finally I'd like to thank our sponsors guests and you for helping to make the show possible please join me next time where I'll be featuring more executives C suite leaders and heads of multinational Edital podcast with your host Elaine Pringle Schwitter