$100 Trillion Question: How should enterprises consume AI?

0:04 Very good morning. It's wonderful to be here. And thank you, Arya, for for the invitation today. For those of you who don't know me, I'm Becky Anderson. And I've been here for about 12, 13 years, coming in and out of the region for the last 25 years. And I think last week's Donald Trump trip to the region, was remarkable for me as a journalist and I think remarkable, for this region, it was a pivotal opportunity to reimagine US Gulf relations in a new era.

0:37 And that era is an era that will be dominated by AI and advanced tech. The message, I think, was absolutely clear. Technology is no longer an industry. It is a pillar of national power and a lever for global influence. Those who act early get a foothold in building the AI infrastructure that will define what comes next. That is the big development bet. What we're going to talk about today is the deployment story because AI's influence on the economy depends, of course, on businesses integrating it into their workflows and customer facing platforms.

1:20 And The Economist had a really interesting article, which I want to allude to, just this week, suggesting that for many companies, the promise of generative AI is giving way to what they describe as a vexation of making productive use of the technology. And they quote a number of organizations, data organizations, suggesting that, the share of companies actually abandoning their projects, their sort of early projects is actually going higher faced with sluggish progress The economists suggesting that the bosses are sliding into this sort of trough of disillusionment and for our McKinsey friend we know that there is the famed hype cycle for consultancies that comes after the euphoria generated by tech.

2:12 So I think is going to be really useful to get your sense of where we are at, today. But I think, you know, the idea of this is to really do a deep dive on where we are at, you know, what enterprise needs And then I think, Arya, of course, you know, this is an opportunity for you to provide a key solution for this audience. Arya, of course, as a founder and CEO, you will know, him here.

2:39 Eddy Farhat is the Executive Director of E& Capital. Faheem, my friend here on my left, is the Group CMCO of G42. Vinay is the partner at Quantum Black AI, McKinsey and Co. Guillaume Princen is the Head of EMEA for Anthropic relatively new, as I understand to the EMEA file for Anthropic. So welcome here and I know that this opportunity for you is going to be a really good one, just to sort of shoot the breeze with so many people here.

3:16 who’ve got a sort of deep understanding of what's going on. I want to take two minutes out of what I know is going to be not enough time today to have the conversation about enterprise but I do just want to spend a couple of minutes with Faheem, who there is nobody better served just to give us a debrief on what happened here last week and the historic announcement on the 5GW AI infrastructure park. The elevator pitch, if you will.

3:47 Will you just give us a debrief about just how significant and consequential this is going to be? Thank you, Becky and morning, everyone. We are extremely proud of, what we announced here in the UAE with the 5 Gigawatt US-UAE AI cluster. And to get that into context, I think it's very important to understand what we're dealing with here. We believe that AI is perhaps the greatest human invention that we have today With us, after we found fire and this is not incremental it's going to pretty much change our social development as civilization And to do that, everyone's kind of looking at AI as this new matrix like thing.

4:35 That's very cool, very sexy. It'll change stuff. And the threat that comes from it. But if you actually look at a lot of stuff that Arya and his team does is at the lowest common denominator, using it for efficiency and transforming that. Now, to do all this, what we are focused on, what we are focused on is building the infrastructure. So imagine this as the electricity grid, right?

5:07 That's what we are building for enterprise businesses to function because minus that infrastructure, nobody can use the true power of AI. And UAE is going to be one of the global nodes to be doing that especially to feed into the global. So it's a massive bet. It is a massive bet. It absolutely is. It pretty much redefines, who we are as a country And, what we bring, to the global audience from an economic perspective and also prosperity.

5:39 Thank you. And I hope that was useful. I know that, you know, those of us who live here, those of us who travel in and out of this country, those of you traveling in and out of this country. Well, if you know, I've been across the, the enormity of what happened last week. But I think it's, really useful to get your perspective, on it as well. All right, so we're going to spend the next 18 minutes, talking amongst this group about why enterprises to a degree remain gridlocked by legacy systems.

6:10 And what I know Arya would, consider outdated processes, despite AI’s promise Let's just get there immediately. I want to talk about Human-AI integration. What does effective human-AI collaboration look like and how to enterprises get this right? And I want to talk about future-proofing. How can enterprises ensure that AI strategy remains relevant in what is a rapidly evolving technology landscape.

6:42 So once we've done that and we'll chat amongst ourselves on that I want to take a couple of, I hope, very targeted questions From you guys. This is an opportunity for you to get some perspective from these guys. So please take that opportunity as well. And we should close this out, I think, at the top of the hour. So let's start. Why is it that enterprises I started this by saying there is a bit of a lull going on.

7:12 There was a bit of disillusionment going on after what has been some early mover activity on enterprise. Why is it that people are getting a little bit stuck at the moment? Why are we seeing this disillusionment What needs to happen next? Thanks first for having me. And indeed, I, I'm very excited by what's happening in this region along the lines of what you said. And that's why I'm here today. To the question, I think that's, it's a very broad question.

7:43 And I think there are many reasons that we're going to talk about today. I think I, I might dive deeper on two specific reasons. One is this notion of trust. We actually are in a business of trust here because this is a very new technology, evolving very fast, and it can seem scary at some point, including for enterprise companies who might be wanting to take big bets. But can be afraid of what's going to happen. I've heard customers saying, what if?

8:14 What if this thing breaks my bank? What if it takes all the money from my company, like some things are You know, you could think kind of ridiculous, but people are somewhat at looking at this and being careful about what it is. And so I think that I'm saying this also because that very much ties back to the first principle of what we build at Anthropic which is we build frontier models that rely on trust and safety.

8:43 And that's been the case since day one. And we talk about it and we act on that day in, day out. So just an example. Every time we get a model out there, a new model out there, very powerful models we will have a set of policies to test them out. And if they enter one of the categories that we deem being potentially very harmless for the world we will actually do hours and weeks and months of extra red teaming and training to make sure what release on the market is entirely safe.

9:20 And that is very key to what we do. And that is something we're pushing across the industry as a standard. We like to make it a standard because it is not the case today. But we think for the benefit of adoption, but also in some way, humanity, we better make sure that the models that are out there and more and more powerful are extremely safe. That's point one. Point two is what we're seeing in the enterprise adoption curve is that enterprises have very much adopted what I'd say the what I call the the early adoption, relatively simple, mundane use cases of simple internal efficiency-driving use cases.

10:04 So I'm going to use this tool for data analysis. I'm going to use this tool for marketing content generation. I'm going to use this tool for internal chat bots. Those are relatively circumscribed low risk. What we have not yet seen that much is large companies entirely transforming their business processes and core processes. And the reason- one of the reasons and there are others. One of the reasons we're seeing for that is, is actually the we've talk a lot about agents, but we're on the road to that.

10:41 And that's what enterprises need. What is an agent? What is a virtual collaborator? A virtual collaborator should do two things. One, it should have the ability to work autonomously for long periods of time because that's how you actually get some real work done. Two it has to have the power to take action. So ability to work autonomously and power to take action. Today, I mean, we're evolving very fast. I predict that in very little time we will have all the autonomy side models who can work hours on their own.

11:16 That's where you're going to start to see the real impact and the real very wide possibilities for enterprise to be able to transform their core business processes. I believe. That's fascinating. And let's talk about what E& is up to. What's your experience to date? So, you know, for me and capital standpoint, you know, we invest in technology startups, some of them being AI. Value in investing in AI is going to derive from the value that the AI provides businesses.

11:46 Enterprises use systems to manage a problem, which is complexity of scale. If you have 3 employees, it's different than you have 30,000. The way reassurance was derived is through rigidity. Actually rigidity of existing system is a feature for, you know, compliance, finance, predictability of what the customer gets. And so there was this trade off with how fast do I evolve with how well do I give trust to my management and my shareholders that it's not going to act funny in production?

12:15 I think with agentic AI and efforts that are being done with some of the fellow panelists, especially Opus is remove that false trade off of how can I give you the reassurance and how can I help you evolve every quarter how you serve your customers? And that's being done in two ways a bit technical. First, you need it like 12 months to redesign how a process would work to serve a customer. Today, with digital collaborators, you can cut that design time and collapse it to weeks.

12:44 The second, and there was a very nice paper yesterday by DeepMind The amount of delegation you can give to agents is highly defined by how well they can explain back to you. Right. And if you think about it, in a human, if I get an intern, I will build confidence and I will grow their scope of agency based on their ability to evidence good execution and doing it. And I think that adoption into enterprises is going to follow this ‘First helped me design faster and make sure that actually you execute well.’ And then as that evolves and we can compute for hours will progressively grow the scope of operation of agents.

13:21 And so companies who are working on either of these are where we believe there will be value in investing because they're going to be material driver for OpEx constriction but also strategic advantage kind of refinement for their companies. And so they bring value to their customers. They make money. We make money. Right, Vinay, I just want to get your insights on this. Governance, relevance of enterprise AI initiatives, trust, agenetic AI Where's your head at this point?

13:53 Couple of points. I'd say, just to add on what some of my fellow colleagues have said. If you spend a dollar in an AI to get it right, you need to spend $5 in change in the organization. So a lot of institutions that have tried to go down, even before a genetic AI- GenAI even traditional machine learning models a lot of institutions have taken this path of I have a technology, I'm going to try to implement that and that's going to work in terms of financial value.

14:27 So we find that's not the case. 92% of institutions have not realized the value, but there are the 8% that have. And the ones that have one element is this looking at change management. How does the organization look different? How will the processes look different? How do I upskill my talent to actually engage with this technology in the right manner address some of the biases that they may have but also realize the benefits that obviously shareholders and the management team are looking for and customers are looking for.

14:58 So that's one point that's helpful. As I look at the leaders, there are a few lighthouses. There are a few in the region also. What they've done is they've taken a business-forward approach to applying AI. What are the business objectives I want to drive? If it's on a call center if it is around a credit-risk processing for the bank if it's for an insurance claims processing, what do I want to achieve? And then figure out what is it- what technology is it that is going to help me.

15:27 It can be AI. It can be agentic AI. It can be even just basic digital process automation. But if you take a business forward approach, you typically do find some some value being realized. Given what we've just heard, What are your sort of first thoughts? So, well, we've been wrestling with this question for a little over three and a half years on how should enterprises consume AI? There are kind of realities and issues with the tech and kind of how you can build it and how you can have security, how you can have trust, etc., but that a lot of the issues are actually non-tech related, which is What the business needs? What are the processes?

15:57 So a solution needs to have all of this incorporated to actually be usable. And, and A lot of the lessons we've learned were actually in the non-tech domains. There is, for example, a major principal agent problem where there's an antibody reaction in any enterprise of people sabotaging it because it's their livelihood on the line. Right. It's not a joke. It's kind of their job loss. People fear change, but the reality is clear that every enterprise must rewire to survive.

16:27 It's not even like an option. So, how can you consume this pill? One thing that we really wanted to do when we were tackling this problem is to build a dream team of people that can really look at this and make an impact. So with each person here, we have a partnership. The only example that comes to mind is electricity. And oftentimes when we're in the room with the infrastructure point, We ask an enterprise What is your electricity plan?

16:53 They don't have one. It's just every single thing you do is electricity. Right. And it’s the same with AI. You should not have an AI plan. You should be business forward. What is the business objective that you want to accomplish? AI will enable that. AI is not an end in and of itself. If you think about when we invented electricity, which wasn't that long ago, by the way, If you think about when we invented electricity, which wasn't that long ago, by the way, what that entailed Imagine you were living with an enterprise that didn't have power, and somebody came and said We can put a wire in.

17:23 What does that look like? And what are the decisions and what are the safety? It can kill you if you get electrocuted. Is it dangerous? What happens? It's that rewiring that needs to happen. It is interesting and I want to bring you in here because we talked about the big development story. And you sit at the very heart of that with G42 and its group of companies. You are of course, in and of yourselves organizations that need to consume AI.

17:52 You want enterprises. So I'm fascinated to hear just the challenges that you faced, perhaps, and where you see the opportunities. So, just to comfort everyone. We also have the carpenters door syndrome. Just because we're G42 doesn't mean that we're AI-native just yet. So I also have a day job, where we run the marketing and communication for the entire group. And for the last six months, my mind has been very focused on the fundamentals of what needs to happen.

18:28 Ever since our evolution we have used the power of communicationvto influence but we've only used it to influence ourselves as human beings. But today, the game has changed. If you just remove yourself, from your jobs and look at your daily lives, we are pretty much run by apps, right? What do you want to do? Your grocery, your laundry, your taxi.

18:59 Everything is apps, and they pretty much decide what you're going to do, what you're going to eat, where you're going to go. It used to be Google many years ago, but it's the apps. Now, imagine this. On Anthropic or ChatGPT In a few weeks you'll be saying: I've got a meeting in about an hour's time I need to have the best miso soup in town in 30 minutes. And I want it not to be more expensive than 50 dirhams.

19:33 The soup gets delivered, you like it, you enjoy it, and you tell your chat. Save it for the next time. I want to have my miso soup. What happened to those brands that you were selecting to pick the restaurant? To pick the delivery? It's all gone. So from a marketing perspective, suddenly you now need to influence not the human, but the API that's going to influence the decision.

20:06 It's a whole paradigm shift, right? What we are struggling as an entire industry in marketing is looking at AI as a small tool that can do incremental stuff. Can I do better creative? Can I do all those little functional stuff? But that's like having ten Einsteins in the room with you and then you're getting him to clean the dishes and wash the clothes. Just an absolute waste.

20:37 So one of the fundamental things that we've done is in January 25 human capital has now got evolved to be labeled as augmented human capital. And we are looking at What is the role of humans and AI to coexist as colleagues. Right. And within the realm of marketing, my brief to myself is in a year's time, I shouldn't have the job that I'm doing now in its current form, I should probably be on a beach somewhere, enjoying my life and nudging and having the impact of what I do today with maybe 18 hours at work, right?

21:23 So we're looking at marketing from a department into a marketing intelligence engine that is at the heart of driving the business, and it's objective rather than a function of fulfillment. Save me a towel on that beach, please. And that takes us seamlessly into what I wanted to do. The second part of this, which is look at What effective human-AI collaboration is going to look like. And how do enterprises get this right?

21:53 So I think in the first round We tackled trust, we tackled technology adoption. And the third one was competence, right? I think the fourth one. And that's a great segway from Faheem is that is an identity redefinition for the workforce and the enterprise. But to draw from a lesson like a decade ago, I was in a much more technical role introducing the first AI tooling in the in the advertising industry. And the biggest friction was not the C-suite, not the management, but the ad ops person saying But what about my role?

22:24 And so everybody becomes a manager today. If you can describe a task, you can inspect the outcome and you can provide feedback. Actually, you're kind of AI ready to some extent. Everybody becoming a manager is the mindset change required to say I can derive personal value in an organization without necessarily being the maker. We'll talk about the other aspects But for me, fundamentally, change is going to be about redefining how we get to work done. And do we need to be the ones doing or can we just orchestrate, manage it and inspect it?

22:56 But that's a significant change in the way that a company is wired and its staff go going. That's the new job. Yeah. Who leads that change management? I hear this all the time. It's going to free people up. You know, people can get more creative. People are going to be given more time. That's great on paper. Is it realistic? You know, because the many, many people in these organizations, many of those people may be responsible for, you know, investing in the sort of technology solutions that you're providing.

23:29 There is going to be an aversion. There's going to be reticence, right? This is a fundamental change that's coming along. I think. Where we've been focusing on, on this question of human-AI interaction the role of one or the other is, looking at What is the actual usage of Claude of Anthropic today by job. So we actually did a major mapping by job, by activity amongst each job, etc., of what that looks like.

24:01 And we came up with what we called an economic index of AI use on Claude. Right. So that's the bias. Two really interesting things. I think one is, the majority when you look at Is AI augmenting or replacing human work? The majority of the AI usage today is actually about augmenting, today. less than half less than the majority is about replacing point one across jobs.

24:34 So I think I think like that is one element. Another really interesting element is that depending on the jobs you're talking about, that ratio of replacing versus augmenting changes dramatically. And I think the the one major example today that is arguably the highest use case in the world, well, at least it is on Claude From what we're seeing Over a third of Claude usage is on coding.

25:05 Coding jobs have a different, a different split. And so it's roughly 80% of coding jobs is AI is replacing, 20% is augmenting. Now, to your point, the really interesting part is even in the replacing part, there are feedback loops because you just can't like today, you can't just let the thing run. And so and so I think, you know, what is my- what is our opinion on that. It's, it's not that we have an opinion, it's that we try to lay the facts out.

25:37 And what we do with enterprise is we we try to we say, listen, these are the facts. How do you as a company want to play with that? How do you as a company want to go faster on this or on this? And how do you as a company want to transform? But these are the facts of if you use these tools, this is where it's going. You keep saying today, which takes us seamlessly to the third part of what I wanted to discuss today.

26:02 The question for so many businesses and to enterprises is How to ensure that their AI strategy is relevant today and tomorrow and the next day when the technology is rapidly developing. Nobody wants a system which is going to augment but not replace staff. If the competitor has a new system, bought tomorrow, which is going to replace, you know, sorry, Khalid, but you know- it just makes more sense.

26:33 The level of aperture is quite interesting when you think about the human-AI integration. And we go kind of deep, you know, if you go really philosophical What is the role of the human now? Right. Like there's a lot of indications that maybe the age of science is over the age of mathematics, the age of the mathematician is over, right? Where there's no human that can do the calculations per second than the average guy. So what do we do?

26:57 We took a lot of pride in, you know, our engineering. If you look at it from first principles, the way we see it is there's a lot of roles humans do that are soul crushing, that are kind of repetitive work that I think is the first thing that we can automate. And humans then start working on the protein. Right. The edge cases, the things that are quite interesting, the parts of their day that gave them joy becomes more and more and more and more, and, and but but for us, where it gets even even more complicated and we'll go through that a little bit later is.

27:28 For some reason, the economy that we inherited from our forefathers. Unitised the work of man by time and and the kind of, you know, there's a lot of time on, on future proofing. We think about what will be a commodity and what won't. Like where do you put your chips that will stay relevant and if you're not careful, you know, with the next iOS update you're out of business, right? It's like so fast. So, We think ultimately economics is the allocation of scarcity and what is scarce is time.

28:01 And that is the kind of fundamental thing that will stay. So why are we selling it for an hour? Whereas, you know, I don't care if it takes you an hour to do something or two hours, why should I pay more for two hours? I'd probably pay more if you do it in 10s. Right. So. So the unit of time has to be rethought. And that is a fundamental question. So we think that labor or human will move towards an output-based pricing model.

28:25 It happened in Uber right where it used to have salaries as a driver. And now it's up a pace. You do a ride to get pay. You don't do a ride, you don't get paid. So we think knowledge work you know which is this a billion human beings earning 12 trillion a year will get repriced from from time based compensation to output-based compensation. And that has massive, ramifications. And then where we see the human stay, which is future-proofed, is on both sides of a piece of work, which is what is the intent, which is what is the prompt we call it.

29:02 But what is the spark of the human? And then on the last part, which we call supervised automation is is kind of like somebody needs to hold the liability, if you get sued. And that is kind of- there's a minimum fee a human will would require to give you their signature and their liability. Yeah. So, I think it's on these two sides. I know Eddy has a lot of thoughts. I think two sentences. If getting what you want becomes so cheap It's very important to know what to want That's one.

29:34 And I think it will cause a lot of existential crisis. And two, I think, How do we also check that we got what we want. There is always going to be a role for that. If today I was an executive working remotely, I have no social life. What I send through email or back, is it a colleague next to me? Is it a colleague in another country or is it an agent? At the end of the day, there is a point where the buck stops and there is a validation.

29:56 I think these are two things. Maybe just to answer your question on institutions What happens if new tech comes tomorrow- future-proofing? I think my view on this one is It raises a bar on institutions to think of what is their true value proposition? If the value proposition was I have an Anthropic model that is using, you know, let's say public data to do something for my customer. That value proposition may not remain not because the model is wrong, but because the value proposition that they are going to consumers with is the wrong thing.

30:29 So because with the future, the way you think of technology, the way you think of your talent, your people, the way you think of your customers and distribution, all of this has to be rethought because what is your true value proposition? Why would that person order the miso soup from your app versus any other app? So that's where the question of future proofing has to be. It can’t be this technology is the right answer for me to future proof.

30:52 It's actual holistic view. What is the customer need? What am I bringing to the table that's unique? And how am I going to change? Because change is going to happen. Today, the customer wants miso soup. Tomorrow, he might want nothing related to Japanese cuisine, right? He might want to go to Italian. So how are you going to have the model flexibility in your entire organization to address that and the people behind it? This has been amazing.

31:16 You got five experts here on how enterprises should consume AI. Let me take a couple of questions from you guys. Thank you so much, guys. This is, really exceptionally interesting. I really enjoyed this panel. Probably one of the best panels I've seen in 2025. My name is Mohammed and CIO of Noia Capital. Dubai and Belgium based multifamily office focused on growth. We are on the CAP table of AppliedAI. The question I had is I wanted each of you guys to make a prediction because I love the fact that you had bold statements and I think the examples you gave are really, really great.

31:52 I wanted to make you, ask you guys to make a prediction to say You all mentioned that augmention versus replacement right now Is it 20, 80, etc.. What's your prediction in five years from now? Dasically of what that will look like Very good question. And yes, can I just take yours as well? Hi, I'm Sonia, I'm actually from Cleveland Clinic, the sister company from G42. We like to call it now. I'm the Executive Director over there for RCM and Finance.

32:19 My question is, pretty much like, you know, how do we your customers are your future customers or how do we partner with that change management? Because we didn't- we not only saw this, from, you know, generative AI is like next level. But even at the RPA level, we did see human sabotage happpening and a lot of fear and disengagement. On prediction. What I can say is what I'm working on now, currently, I will have my first AI full time employee in about 12 weeks within my team.

32:54 And that's the pace and scale with which we are moving. And I anticipate that happening across the board. But the way to do this is not look at this as a technology transformation, but we are going through a social renaissance. It's a human transformation that is happening. And technology is just the agent that is catalyzing that change. If you look at it from that perspective, it is not about what you need to change in your business, but will your business even hold true going forward?

33:30 Because there will be businesses that industries that will not see this but there will be a lot new ones that will come. So it's all about going through that transformation and finding a place. In 12 weeks, you'll have your first AI member, a staff in the department. Does that replace headcount that currently exists, or does it replace a series of headcount that might otherwise exist? I mean, it would, yeah, it would. It would redefine the role of what I have to do and my colleagues have to do it. Yes. Right.

33:59 Fascinating. I have to be at the beach. I'll take the two horizon answer on this one. Short term, In the next few weeks and months, I think we're going to see models that can work on extremely complex tasks, using thousands of data sources for hours in a row without needing human interaction. Like, that's kind of where we're going in the short term Five years from now...

34:32 Five years is long time number one. Three years from now...I like to refer this as there will be countries of geniuses inside data centers. That is clearly our prediction. And mine. What is that going to mean on the work side? We need to figure it out. But there will be countries of geniuses in data centers in this world. I'll make a selfish prediction. I currently answer 9-10% of my emails.

35:05 And the next 12 months I hope to answer none. Like, ideally, I'd have a digital Eddy for 99% of the questions that somewhere in my knowledge base, it can handle on my behalf. Right? So more beach time, uninterrupted. The second one, I think it's not the first time, we introduce technology. I think the more the industry is regulated the more that is formal verification the more you can build verifier agents without a human. I think that's where we're seeing actually the counterintuitive entry of innovation is that the industries in between where the human needs to verify it's a bit- Good, is not defined narrowly, you need a human checker. but actually where Good can be defined very precisely, you know Like drug design, it's not an easy thing.

35:52 It's actually on the other end of the spectrum The benefit of it is either it binds to the molecule or not. So so you actually have fortement. And so, you know, when we look at startups, we've seen quite a few and very rigid domains actually achieve things that we would have thought were five years down the road. It's the contrary. On the human side, The best analogy I always give is if you had a free intern that doesn't need to take weekend breaks, what would you have them do?

36:17 There is an entry point. And even augmentation is a path to automation because you need acceptance first. And acceptance is: ‘How can this be useful to me?’ at the lowest level of the organization. So and I think the role of technology and HR teams is How can we actually make our employees the first very happy customers of AI? And that's how you bring it in right now. It could be a Trojan horse on the long term, but that's the starting point.

36:44 Fascinating. So yeah. So on the prediction. Well, I can just say one thing that whatever prediction we say we will be wrong in a few days This is from McKinsey. And actually, so we did some research, at MGI, which was what is the level of automation? And we had done this seminal work a few several years ago that you can actually achieve just through RPA, etc. at that time you said 50% of activities can be actually automated, and that was several, at least 4 or 5 years ago when the report came out, when we looked back at some of those numbers and some of those, predictions that we had, all the numbers are, frankly, you know, understated and we’ve clearly missed a boat on a variety of those activities.

37:30 So that's the challenge we have, right? The bar of what we believe reality is going to be and adoption is shifting rapidly. Now I think there is a world however, with a massive stock of institutions rigid- have no idea how to move. Those institutions are going to struggle over the next several years. Right. And so the question for us would be, not only from an institutional standpoint or also from a societal standpoint how do we help people move and shift in the right direction?

37:57 Because I do agree that there going to be a lot more Eddys who are going to enjoy the beach, and but there will be a few others who haven't figured out how to use the digital Eddy. Right? So how do you figure out and help that particular person? I think what we've seen in our own data is we've increased the productivity per person around 8x, right? I think by the end of summer, it'll be around 10x.

38:18 which means you need 10% of people to do the same amount of jobs but with a higher accuracy. Right. So our accuracy has gone up 18%. So it's not the old trade off of productivity of like faster or quality. This is a parallel shift of the productivity curve. So with the caveat that as Vinay mentioned, I think the human mind is incremental and we kind of cannot see what what things look like. We think there will be a massive amount of increase of production, a productive that you can do. And how you allocate that is your choice.

38:49 So you can either keep, you know, the staff and do ten times more revenue or you know when when you have churn, you don't replace that kind of graduate. With your question. This has been the bane of our existence. And it's where the reality hits the road. Our experience is really strong executive sponsorship. So when you have that sort of internal politics there's a moment everyone fights and there's sabotage and it's very smart sabotage. They're like, upload erroneous stuff to show the AI is wrong, to make the case in a way that's like they collaborate because it's their life, right?

39:26 But there's a moment when the ship has sailed and they know it's happening, that the politics shifts. Well, now they want to be heroes, now they want to fix it. Now, they want to be the first to use it. So the sooner the directive from the top is not We're exploring a POC or a pilot to see if it makes- There they have a chance to kind of sabotage it- But it's happening. You're either on a ship or you're not.

39:46 And then, you know, it removes a lot of that, gamesmanship. That's fascinating. I, like you had a you had a, comment to make on this as well. Yeah. Just two things. One is, a pharmaceutical company that adopted AI very early on and, it did it in its, human resources team. They have about 8000 people in the organization the most trusted person in that human resource team is the AI bot.

40:16 It has no frame bias, no agenda, none of that. But what it has done over the last five years as it has increased productivity in the organization by 37%, but innovation has gone up to 72%. So all the stuff that was getting blocked by line managers and different politics, as you say, has completely vanished. So engineering that change where you focus not on the product but your own people is a good starting point.

40:48 And I'll build on Arya’s piece. Humans have always used information as the primary tool for power. So if you can commoditize that information and then focus only on the exceptions that you need to build in, the whole cultural transformation will take place. We actually build systems, not on exceptions, but on blocking every information and then just opening up a very small funnel. It's all human made in our heads.

41:20 None of it holds true. And with that, I'm going to wind this up. This has been absolutely fascinating. I hope this has been valuable, to you guys. I'm seeing a lot of people nodding here. I think it's been, amazing. I was at the Culture summit, a couple of weeks ago, and we were there being asked to consider what it means to be human in a post-human world or in an age of unprecedented innovation. I put that question as a prompt into ChatGPT.

41:49 Sorry. And I got this quote In a post human, being human involves navigating a landscape shaped by technology, ethical complexity, cultural evolution. It challenges us to reconsider identity, connection, and responsibility, and to find meaning at a continually transforming existence. That was the kind of, you know, the sort of very ethereal way of looking at this. The culture summit, where everybody's got a sort of, you know, big lofty ideas. I think this is sort of really been an opportunity to hone down and to your point, how do you manage this transforming existence.

42:28 So much of what we've discussed today speaks to the work that you've been doing in working out why people are still working on legacy projects how you integrate AI with human interaction And then this future proofing story for businesses is pretty much where it's all at. Thank you very much indeed to you all.