Join top executives at WEF 2026 in Davos as they discuss why scaling AI is the hardest next step and what’s needed to unlock its full potential.
At Davos 2026, leaders from Philips, Visa, Aramco and Accenture argued that “scaling AI” is less a technical upgrade than an operating-model reset focused on measurable outcomes. In healthcare, Roy Jakobs emphasized giving “time back to the practice,” citing ambient listening and workflow automation that can return 10–15 minutes per hour to nurses while improving clinical quality. Julie Sweet noted a shift in executive expectations: “78%… believe that AI is actually helping growth more than productivity,” as pharma teams move from chasing approvals to ensuring content reaches the right physicians.
Aramco’s Amin Nasser quantified scale: 500 use cases, 100 moved from pilot to deployment, with multi-billion-dollar “technology realized value” and third-party verification. He credited trained subject-matter experts, a clear “kill, pilot or scale” cadence, and “data quality”—“If garbage in, garbage out.” Visa’s Ryan McInerney predicted 2026 brings native purchasing inside AI platforms, but true agentic commerce requires trust infrastructure: AI-ready cards, a “trusted agent protocol,” and user-set spending parameters.
Across sectors, the hardest part is adoption and talent. Sweet reframed governance as “human in the lead, not human in the loop,” while McInerney said breakthroughs came only after hands-on leader training. Jakobs warned reliability, safety and ecosystem collaboration must be designed in from day one.
Well. Welcome everyone. Welcome to our panel today on scaling AI. Now comes the hard part. This is part of the World Economic Forum's AI Transformation of Industries initiative, with with $1.5 trillion in investment in 2025, the race to capture the full economic impact of AI is on. But scaling beyond pilots remains a major hurdle for many, many businesses. It requires new strategies, new capabilities. We've got a great panel today. Who's going to talk about that? Joining me, Roy Jacobs, president and chief executive officer of Royal Philips. Ryan McInerney, chief executive officer of Visa. Amin Nasser, president and chief executive officer of Aramco. And Julie Sweet, chair and chief executive officer of Accenture. But before I start in with questions for you, I've got a question for the audience. How many people out here have tried a have launched a pilot program, an AI pilot program in your organization? Can I just see a show of hands for people who have? That's most of the that's most of the people in the room. That's definitely most people. And then if you have been able to scale those up, let me see those hands again. A lot fewer, but more than I would have expected. Okay. Okay. That's that's, that's that's about what I thought. And then one more, just for those of you who have scaled up, how many of you ran into unexpected challenges? Same number. Okay, great. So let's talk a little bit today about scaling up AI. And I'm going to start off by a lot happens every year. It seems like every year is ten years in AI years. And I'm wondering what's possible with AI in 2026. That wasn't possible a year or two ago. And Roy, I think we'll start with you and maybe I'll ask that question about healthcare. Are there, you know, clinical operational improvements that you've seen due to AI, that it's enabled in the past year or so?
Yeah, I think actually AI in healthcare is going really fast. I think the need is urgent because we just don't have enough people to take care of patients and people that need care. And therefore, I think the use of AI is finding its way and it's finding its way to improve clinical outcomes. For example, how can it help with, better diagnosing a patient when you take an image, when you actually measure people through a monitor? How can you do better decision support for clinical interventions to make them more accurate? To prevent any mistakes we make to be made. So actually augment the clinician with very fast and accurate data to support them in their daily work. And then there's a big piece which is also about operational improvement, because this is a huge admin burden in healthcare. And how can you actually take away some of those steps that nurses, technicians, doctors have to take that they really don't want to spend their time on? And that's actually automating some of the workflow of the care pathways. And that's where we see rapid improvement. For example, if you do ambient listening in a patient room, you don't have to actually have to a clinician taking notes, it actually can be recorded at once. And then actually, they can, sign off the transcript and then actually they can move on in their job. If a nurse that actually spends 15 to 20 minutes an hour on admin tasks can be relieved by 10 or 15 minutes of that, she or he can spend that time on the patient. So I think there's real meaningful progress. And you see rapid adoption as a result in healthcare of AI.
I like this example of the nurse because that's a concrete outcome. That's that's good for the that's good for the patient. Right? That's that's not just something that's good for the business. It's good for everybody. Is that is that the way you view it?
Yeah. I think so we also describe it as how can you give time back to the practice. Right. In healthcare it's all about how do you care better about the patient. The clinician wants to spend more quality time. And also having the conversation, having the kind of the aftercare from when you have a diagnosis, not only the hard fact of this is what you have, but also what does it mean, how are we going to treat you? And the same with a with a nurse that wants to have actually quality time with the patient on average is 5 to 7 minutes or three minutes that they can spend time. That's not a lot of time. And actually patients need better quality time for them. And that's where we can give time back. And AI and agents can give time back. So it's a different notion of you also improve productivity. You also improve clinical outcomes. But there's also really quality aspect in giving something back to care that actually we have been missing or that has become under bigger pressure because we just don't have enough people to do the job.
And do you go ahead, Julie?
I'd say Roy's point is really important because I think one of the things we've really learned is that we started a conversation around AI that was so focused on productivity and not actually the full outcome. So in a related industry in pharma, we're seeing the same thing where one of the biggest things when you take drugs to market is you have to comply all of the content to explain the drugs to physicians. You have to comply with lots of, you know, regulations around the world. Most pharma companies have a lot of different processes for that. And so we are working with a pharma company where we standardize the processes. We now have content. They can do what took months today, so they're able to get to market more quickly. But actually the most interesting insight was that the people who used to spend time saying, you know, how do I get legal approval are now spending time saying, who needs this? Was the content helpful? Because before, you know, if once you got it approved, the last thing you wanted to do was update the content, you know, and go through the same approval process. And so much like the patient outcome, they're spending more time thinking about how to get the drug to the right places, make sure they understand it, which of course helps revenue, but it's really helping patients. And there are many examples of that across the globe in different industries and different things where it's not just productivity. And in fact, our latest survey, just this last quarter, 78% of, of CEOs, the C-suite believe that AI is actually helping growth more than productivity, like in terms of like the value of it. So I think it's a really important part. And that's a learning over the last 12 months, as you've started to see a lot of these things begin to scale.
Thank you. Ryan, I'm curious about Agentic Commerce. We've been hearing about it for a long time. It seems like 2026 is the year that's actually really going to take off, or people are predicting it. And I'm wondering what, what you're seeing is sort of the next big frontier there. What are the what are the the challenges that you see coming with the, with implementing that?
So I think last year, most users, most consumers started to use these platforms and these APIs to shop for things, for discovery, to look for, you know, move off of maybe one of the search platforms or the commerce platforms. But then they actually when they actually went to buy something, they went to the native merchant seller site. This year, in 2026, I think most of us will continue to shop on our AI platform of choice, but now we'll be able to buy natively on the platform. The buy button will be there. I won't need to leave, whether it's, you know, ChatGPT or Gemini or Claude or Copilot or what have you. I think as we start to emerge from 2026 and look beyond, that's when you'll start to see a shift of real agentic commerce, not just me pressing the buy button on one of these platforms, but me empowering an agent on my behalf to go shop for something, find it, and then buy it on my behalf. But for that to work to your question, we need to invest in trust. You need to trust your agent that they're not going to go crazy and buy something you don't want to buy, or spend more money than than you want them to spend. Merchants need to trust that if an agent is showing up, you know, at their digital doorstep that it's actually there on your behalf and you've empowered it. And your bank needs to trust that when they get a request to authorize a transaction on your behalf, that you really wanted that to happen. So for that all to happen, we are deploying AI ready cards, AI ready visa cards around the world that empower users to set the parameters to so that they'll trust their agent. How much money can you spend? What's the size of the transaction? Where can you go buy? For how long do I want that open to buy to exist? Things like that. We've rolled out a trusted agent protocol so that a merchant knows that if a an agent is showing up on my behalf with a visa card, that it's a real one, with a real card that's actually been empowered with the data payloads that, you know, I just described. And then finally, we've rolled out a level of personalization so that all of us as visa cardholders, if we choose, can empower our agent to look at our shopping history on my visa cards and use that to personalize recommendations and shopping experiences and user journeys and those types of things. So I think we're on the brink of something that is going to be very powerful. I think it's going to be a great time to be a consumer. Shopping is going to be easier. It's going to be more secure. It's going to be more fun. It's going to be more efficient. And I think it's going to ultimately grow commerce and be a great thing for merchants around the world.
Can I follow up a little bit on that last bit when you when you're talking about having an agent sort of understand my purchasing history and transacting on my behalf, so would this be something like it understands what I'm shopping for at the grocery store on a, on a, on a weekly basis or a monthly basis and goes out and does that for me. It understands that, yes, it's winter time and maybe I'm starting to need a new coat. Like like how far do you envision this going?
Yeah, I think the, the, the collective data.
Of your shopping history on your visa visa card over time is very powerful. But historically, we haven't had the tools and the services to allow users to put that shopping history to work in a way that's very empowering for the user. So imagine a world going forward where, for example, you go to your bank and you're able to, you know, empower your agent via your bank to leverage your shopping history to tailor those purchases. What what types, what time of year do you do? You make travel purchases? What are the type of restaurant purchases that you make that allows, again, your agent to have much more context than what they're able to do from your normal user profile.
Amin. You were on a similar panel last year, and I'm curious, compared to this time last year, where you've seen the most meaningful change in what AI can support across your energy operations?
Absolutely. You know, we have, last year I talked about 400 use cases that we, came up with in Saudi Aramco this year. We're talking about 500 use cases. 100 use cases went from the pilots to actual deployments. We measure our progress in terms of what we call a technology realized value every year. And, you know, we used to have like 200 to $300 million in the previous years in terms of technology realized value. In 23 and 24, we achieved $6 billion. 50% of that is AI related. 25 we should publish our numbers next month after we finish the third party verification. And when looking at 3 to $5 billion, also more than 50% is AI related. We've seen the benefits of the huge infrastructure that we have built over 90 years, the 6000 talents that we trained on AI. These are the subject matter experts that come up with the use cases, not the data analyst. We have a couple of hundreds of these, but it is the subject matter expert who understand where are the create the pipeline for the opportunities. And the most important thing we established the operation model basically, capitalizing on our digital company, we created a digital company and the AI Center of Excellence, which established the pathway for taking ideas from the front lines to full piloting and then full deployment, establishing the processes and taking it into account helped a lot in creating the pipeline. Now the pipeline keeps increasing year on year. We were looking at 2 to $4 billion a year. Now the team is asking for a much bigger target considering the pipeline of the opportunities. So I think the development, the training, we found out that we can scale much quickly. And now the next step is working with the hyperscalers is how do we commercialize these outside Saudi Aramco to the market? Because this is very important and has a significant impact. Everybody talks about AI, the impact of AI, but where is the value, where is in dollar figures. And this is what we are able to establish. We want to turn the energy sector to be more intelligent in terms of capitalizing on AI. And we have the applications and the talents and the infrastructure. And the most important things in all of these is the data quality. If garbage in, garbage out, if you don't have the data quality and we have built data quality over 90 years, we kept everything because we had the infrastructure that allowed us to scale now and be at this level. Another important element that also helped in all of these is we have a venture capital arm. That venture capital arm have more than $7.5 billion, approximately $7.5 billion for investment. It find a lot of startups across the world. These startups, they want injection of funds, but the most important things for them to buy it and scale their technology. And this is what we offer not only funding, but we can scout for good, technologies, a lot of it AI related, and then help funding it and buying it, piloting it and scaling it up. And this is really helped a lot. Without the venture capital, we could not count around the world for a lot of opportunities and ideas.
Can you talk at all specifically about where you found some of these savings, where you found some of this, you know, the 3 to $5 billion you're talking about?
Okay, a lot of the savings the upstream is, is a high cost, but the opportunity is huge. For example, if we use AI in increasing productivity in the subsurface, you we have what we call the intelligent Earth model. We can basically increase the productive zone in a state of. We used to have over 80%. Now we have over 90% of the productive zone capitalizing on AI we run in life while we are geosteering the drill bits underground, the intelligent Earth model to understand where is the productive zone will be and we put place our walls increase the productivity in some wells by 30 to 40%. That is a huge impact. Not only your cost, it impact also your emissions and how much you are emitting because you need less. World's in. I talked last year about corrosion. It's a $3 trillion market and it helped us a lot in using AI to reduce the amount of corrosion inhibitor, the amount of, Pipeline failures increased our reliability. If you look at AI use in equipment reliability and predicting failures and predicting when we need to take equipment out of the service before it have a catastrophic failures, huge benefits in reducing downtime and increasing operational efficiency of our equipments. If you look at, pipelines in terms of inspections and how much we can save, you capitalize on AI a lot of downstream, which is not as big as upstream. We created a system where it allowed us to increase margin by looking at columns and maximizing the yield out of a column, instead of waiting for a console operator to make a decision, you can make the decisions in minutes and seconds now and change the yield. So there is a lot of opportunities that helped us, but each of them is treated like a project with a timeline deliverable and the impact. And this is what's important. This is where we were able to scale because we can see the benefits and we can prioritize every use case before we take it from violet to deployment and see, the real value. And when I mentioned the billions, this is we didn't even keep it for our team to decide. This is how much we brought third party. They finished their work. As I said, we hopefully next month they will give us the report on each item. And how much is the saving that is AI related, that is digital related or is it captured right? Or it's not captured right. So the real value is and of course, the most important thing you need to reward the people and recognizing them for all of these achievements. And this is where the pipeline now is. We just have to prioritize which one, that we can take it to a pilot and to deployment. Because the number of ideas that came from these 6000 that we trained is endless.
So I'm noticing a commonality here across all three of you, which is that there is this real focus on outcomes that you mentioned versus, you know, just sort of analyzing data and trying to find efficiencies where you can. It's interesting, Julie, you've got a unique view that's broad across many companies. And I'm curious, since last year, what would you say has changed the most in how Accenture and the large organizations that you work with are deploying AI at scale?
Well, you know, there's a lot of insights, actually, in what Amin just said. In terms of, you know, where where are we now in terms of scaling, because there's scaling projects like an individual one, and then there's scale really across the enterprise. And so Aramco is a great example of being able to scale across the company, not only just, you know, a few projects and you know, if you if you think about what I mean, just said and you sort of extract that to industries, what, what we're helping clients do, and I've got to see a lot of this firsthand with, with Aramco in terms of the discipline and what they do is that, first of all, they have the right technology stack. And so AI has been a catalyst for companies to really look at their technology. Aramco couldn't do what they did if they hadn't been investing for years. Right. And so the ability to look at the tech stack and say, what do I really have to do? And to do that at scale data has been something that's super expensive that for years people haven't wanted to do. And the companies that have done it early, like Aramco, think about a McDonald's who, you know, created their data foundation very early, and now they're surging ahead and how they're using it. So, but in addition, it's also about the operating model and the processes, which is what we also just heard. And so now we're working with a lot of clients who are saying we have to really rethink, move from a project to how do we have to operate, how do we have to have our operations? And then the biggest, one of the biggest barriers to scale has been the lack of discipline or willingness to say, I'm going to put a I have to get a value on this. I have to be able to see it in my PNL. I have to be able to embed it in the objectives of my leaders. All of these are like best practices where, as companies have said, this is real. I've seen the value. You know, I really want to move ahead with that. And those are major transformations depending on where you start. And then when you think about what Ryan said, I mean, every place, every industry has battles that they must win. So your best examples, I mean, we're all in your core operations subsurface, right? It's not the you're doing plenty in the, you know, corporate functions. But that's not where the real value is coming, right? If you're a consumer goods or retail agentic commerce is a must win battle. Right now, it's a brand new channel. It's rapidly evolving. It may be early because, you know, you can't yet buy a ton in that way, but it's going to move fast. And so companies are looking at and saying in many cases, there's a huge opportunity before you get to advanced AI right processes that aren't standardized to many spans and layers. But there are also must win battles. And so there's much more sophistication now. And really not talking about AI, but going back to basics, what's my strategy? Are the basis of competition in my industry changing, and if so, I've got to focus there while I build out the foundation.
All three of these companies that you're sitting to your left or gentleman sitting to your left have these rich data sets that they're able to, you know, to to dig into and look for places where they can, you know, find value. What do you say to companies that are maybe newer, that don't have those or that that, you know, what's the challenge for a company that doesn't have that rich data set?
Well, keep in mind, the stats are pretty stark. Over 90% of the data work that companies have to do when you look across the globe is still to come. Okay. So we we, you know, we have to get very realistic about where we are. And of course, the technology itself needs to continue to evolve. And so on. The data foundation, though, it's not optional, but it's also there's brand new ways using AI to build a data foundation. But that that work has to be done in order to scale. And so when people say, why aren't things scaling across the enterprise, you have to have the data, right? So, but there's new ways to do it. And we are working with companies who are building up creating investment capacity by doing what I said, like not focusing on the basics. I mean things like fragmented processes, too many spans and layers and management, like, you know, too many people. Those are things that everyone could have been addressing over the last decade and many companies have. But that's a big opportunity today to make sure you're doing that, to create the investment capacity to help fund this. And we're working with many companies where it's, you know, it's two speeds, like get the foundation in place, make sure you're using the technology. You have the AI that's already built there, and then focus on where you need to have the data and the must win battles.
Ryan, I'm going to come back to you, if you don't mind. I'm curious about Agentic Commerce, and I'm wondering what kinds of opportunities that's going to create for global payment systems that don't exist right now.
Well, I think, Julie talked about must wins for visa Agentic Commerce is a must win. You know, we, if you think kind of over the arc of time, like when e-commerce first happened, visa was a huge part of making that happen. Then you saw the rise of mobile commerce. And we put in place standards, technology to, to make that happen. And now we see this kind of agentic commerce as this big third wave of digital commerce that will happen around the world. I think it's going to be an opportunity for a lot of different players to innovate. But, you know, you come back to this notion of truly scaling AI. That's where we can play a unique role. You know, we have 5 billion visa cards around the world. We have 175 million sellers on our platform. We now have 13 or 14 billion visa tokens in the digital ecosystem. That's the kind of the basis or the platform on which we we build this. And, you know, I think most people in the audience traveled here on on Saturday or Sunday. They came up here to Davos. They didn't think about kind of how they were going to pay for stuff. When they got here. They knew that they had a visa card in their wallet, purse or phone. They knew that it would work when they got here. And they as they walked down the promenade and buy a coffee or a gift to bring home, they know that transaction is going to work and so does the seller. They trust from people from all different types of of countries around the world. That's the type of scale that we're trying to put in place with Agentic commerce in 200 countries and territories around the world. And, you know, you ask about the opportunities. I, I have a thesis that agentic commerce could be an amazing leveler and empowerer for small businesses around the world. Today, most of our commerce happens as users on a small number of commerce platforms or search platforms. I believe once we all are using agents in the way that I described earlier to go search the world's inventory to find the right item for my wife's birthday. The right price for that, that maybe airplane ticket that I want to take or the right combination of features that I'm looking for. You have small businesses all around the world that have the ability to make their inventory available to to make their services available. And, you know, I think there's a real chance that this third wave of digital commerce, this wave of commerce will empower small businesses to grow significantly, small businesses in countries around the world. I might not have found a small business around the corner from where I live in my town in Northern California that I might not thought of to go buy running shoes because my default instinct was just to go to one of the larger commerce platforms. So I think it's a very I mentioned earlier, I think it's a great time to be a consumer. I'm very excited about the rise of small and medium sized businesses in this third wave of agentic commerce.
Just to to to dig in on that a little bit. Are you are you imagining a world where like like I can understand the plane ticket example, you know, I, you know, I want to go to, you know, Ireland sometime in the early fall. Please help me find the best week, identify the best week and plane tickets. But are you also saying that maybe a generic commerce will enable purchasing things that I don't necessarily know about, like like to the running shoes example? You know, a brand that I'm not familiar with. You know, something I've, you know, would have even thought to ask it for.
Yeah, absolutely. I the speaking for myself as a consumer and I think on behalf of most consumers, our search window for the things that we buy is remarkably narrow, given the availability of brands and inventory around the world. You just think about your own user experiences and, you know, building on your running shoe or example or whatever it is, despite the fact of all these amazing brands and small businesses and products that are out there in the world, our search window remains remarkably narrow. With the power of AI and with the proliferation of these platforms, users have the ability to search the world's inventory in real time. And I do think it's going to be a rise of brands that you and I might not otherwise have been aware of because of the narrowness of our search window. I think it's going to be the rise of small businesses like I was mentioning earlier, potentially even different types of trips that you might take. You know, you talked about going to Ireland, which week to go. But as you start to widen the aperture significantly on the different places that people could go, the different experiences that they might want to go have, I think it's going to drive growth in commerce. I think it's going to drive growth in spending. I think it's going to drive, a much broader, if you will, democratization of spending of all of our spending on goods and services around the world.
Sticking with this idea of what's coming next. Roy, I'm curious what breakthroughs in in AI enabled clinical insights or complex tasks? Complex healthcare tasks that you see on the horizon that you think are coming.
Yeah, maybe building on a theme of of agents. I think the the power of of agents in the clinical practice will be really a breakthrough, because there are so many tasks that currently have to be done that, as I said, are under time pressure. There's not a lot of data that can be put together easily and quickly. For example, if you're a cancer patient, it starts with scheduling. Very mundane. How do I get into the system? What is the information that I need to get to when you arrive in the hospital? How does the nurse or the clinician have the holistic patient view pulling the data from all the different systems, not only the EMR, but then also the imaging data, then also the real time data that can be done by an agent in preparation of the next step in the process. If you need to prepare a tumor board, deliberation currently is an extremely laborious process. Actually, that's something that Asia can do. And I think, therefore, reimagining the future of healthcare is kind of how you work with workers and coworkers being agents together because we just don't have enough staff. We will not have enough radiologists, we will not have enough technicians, we will not have enough nurses. So we need to reinvent in the care pathway, which are the tasks that actually we can give to an agent to reliably support the practice. And actually that's already being done now in smaller parts. And what you will see, this will grow in terms of which complete tasks they can do. They can do outbound calling. When you're discharged and you're at home, you're recovering. There will be some in calling upon you. Currently, we don't have the time to ask nurses to call you every day, every week. An agent can do that. That's not a problem, right? And they can actually exchange with you. They can listen to you. They can actually feed that back into the healthcare system to also, if you then combine it with monitoring on your body, actually to have a more holistic view on how you are, they can check in, but they also see the measurement. Now that actually will trigger a complete different system that becomes much more proactive than reactive. So I think the future of healthcare will be one in which we can be really much more at the forefront, acting quicker, because we can actually put the data together in a better way and then act upon it much faster. But it does require you to reimagine the process, because we currently have. And I think that's the big change and the big challenge. We really need to redefine how we work, right. When you are going to adopt new workers in your workforce, you need to rethink how the team is going to play together to do the same tasks. And we have hardwired current tasks in processes, in IT systems, in standards, in job descriptions. So the real breakthrough in adoption will be actually adapting all of those, because then you change the system versus changing one piece of it and then injecting AI. And then okay, if the person is open to AI embraces it, goes with it, it will work. But actually it should become a new way of working. And I think that's where it's turning to. And that's a very, I think, interesting phase, but a hard phase because it requires all of us to do different things, breakthrough routines. And that's the hardest thing, especially also if you talk about patients because you don't want to break routines with a risk for patients, right? It needs to be reliable. It needs to be safe. So that's a component, which really, of course, comes into this whole, redefinition because you need to improve the practice from a reliability and a safety perspective. And I think you can. But then you also need and maybe that's the last point, to be fair to AI, what we currently see is we are not always fair to AI. I give you an example. On average, a doctor gets this diagnosis right in 82% of the cases. We ask from AI before we adopted it into practice. Often that needs to be 95% accurate. That 13% gap between the AI accuracy and the clinician's accuracy is a huge patient impact, right? So we need to find the trade off in terms of where do you put the boundary about where AI can be set free in a safe way versus kind of where you protect also patient interests, patient privacy. And that's I think where we are moving towards learning together, with an ecosystem and maybe building on what you said earlier, how can people that don't have access to the data set contribute? This is a real ecosystem play. As Philips, we cannot improve the health care practice alone. We are one of many actors in that system. So we need to work with open platforms. With open systems, we need to adopt others AI into our workflows, and we need to work very closely with regulators, with clinicians, with the healthcare systems, and with other kind of, companies that are innovating in this space to see what the best is, how we can move it forward in a more standardized way. And I think that's where we need to really collaborate. And that's, I think the other piece of AI where people often forget, okay, this is a technical play. No, it's a collaboration. It's really a collaborative play. And the power of AI really comes forward if actually rally around it together and not do our own technical stuff, because we can do that. But then actually we will not land because the system only changes if you change all the parameters.
I'm glad you came back to reliability so many times there. You spoke about it again and again, because I think that's one of the concerns that, as, as a healthcare consumer, I'm going to have, it's a natural thing that I want to make sure that these systems that, that you're implementing are reliable. How do you how do you how do you ensure that to to patients, to to healthcare consumers?
Yeah. I, I think it's extremely important we talked about trust and trust in Nye. So as an example we had the National Academy of Medicine, Academy of Medicine, building an initiative where we have worked across industry. Some are in this room as well. We had Mayo, we had Google, we had Philips, we had patient advocate groups looking together at kind of what code of conduct we should apply in, applying healthcare or AI into healthcare. Because what we also realize is regulation cannot keep up with the speed of technology. So we need to be ahead of regulation and self-regulate. So we need to have our own kind of rules. How we test, how we validate what is the level of rigor we apply, what is the kind of the practices, but also from biases that we take into account. So actually this is something that we own. You have your own accountability. You have an accountability as an industry. So the healthcare industry together. And actually by applying and living that self conduct, we can make sure that this actually lives by the practice of, of good data. I will also give the other example. I often give the example of say if you are an cancer patient, again, you're in phase 3 or 4. Data privacy has a completely different meaning to you than if you're healthy. That really because when you're there I can guarantee you you won't. Every data set in the world to be used to find the cure for you. So we also need to keep that perspective in mind of the patient, that he also has a right to get access to the data of the world so that we can innovate on those data. And you can do that anonymously. You can do that in a way that actually doesn't touch in any way the privacy. But we need to break through, and that's something we need to do in Europe. That's something we need to do in the world, because the power of that data play is where we will find really the next frontier of solving healthcare challenges in the world.
It's an excellent point.
Matt, can.
I please.
You know, one thing that, as you're thinking about what we haven't talked about is talent, because I agree with Roy, like the the point around process and having to reimagine things, the point around this collaboration. But to reimagine you have to understand the technology in a depth. Because if you're thinking about in healthcare and you say, I've hardwired this, I've done this for years, we know this works. The doctors have to understand the technology, right. The regulators have to understand the technology. And, you know, you have to have the leadership that is then going to embrace the innovation. Right. And, you know, when we talk about AI as hard, all of these things like rewiring the process is really hard, but you have to have a different level of understanding than you did in the digital era. And that is a big barrier for organizations. And it's also a barrier to building trust because it's hard to trust something, you know, whether you're a consumer, a regulator, a doctor, a leader of a health system until you understand it. And there's a lot more that has to be done kind of at all levels of the corporate world government, educational systems, to really build in that kind of level of understanding. And Accenture, we talk about leader led learning, like we started actually with our leaders because we said we can't rotate our business if our leaders don't understand the power of it, so that there is still so much ahead, to do at companies in order to ever be able to scale and to really get the value out of AI, and to move away from what's often very incremental. And, you know, companies I smell all the time, how am I going to operate in three years and five years? And I'm like, the focus really has to be, are you able to do something you can't do today, or do you have insights that you can't do today? And that requires a depth of understanding of the technology and its power that a lot of us, you know, are still working on.
I love that quote. I'm going I'm going to come back to that at the end here. I mean, I'm going to ask you the same question about looking towards the future. One of the things that we've covered a lot in the past year, year and a half are the energy demands of AI. You sort of have maybe a different perspective on that. And I'm curious where you see AI creating new opportunities for integrated energy systems in the coming years, to sort of take the I think a lot of what I see day to day, especially in the United States, is, you know, is concerns about data centers, concerns about about their power requirements. But maybe if you could talk a little bit about how AI is, is, is going to enable some new opportunities there.
I think the opportunities enabled by AI is endless. As I said, you know, the only issue is you need to continue to train more people in terms of understanding AI. And I always say, you cannot scale, without scaling the talents and the people. It's not about it's about creating value. It's not about eliminating people. And this is the model that needs to be used when you look at the future and what needs to be done. And I think, demonstrating is very important the use cases because a lot of people question what is the value. It's not about buying ships and GPUs, and it's ensuring that you have the system, the data quality is and people think that they can by bringing ships and GPUs and installing them, you can create value. No. It's about also creating the talents and making sure that you have the data quality that will help you. What we think of the future is that we will be working with hyperscalers, because the models and the use cases and the pilots that we scaled can be adopted across the industry. We can get to autonomous operation over in the future without losing control of our operation or safety, which is very important. We are running in a business that is very, accident could happen. So I think demonstrating the value that can be created while maintaining operation control and the safety of the plants, you know, we are running and taking that with hyperscalers and also our bland, investment now in humane in the Kingdom, which will help not only nationally but internationally in terms of data centers on all of it. I think, the use of AI in every aspect of operation is there, but more value is not in finance or in translation or in legal. The more value that we see is in real operation. When you look at the way you are doing certain things, this is where we put the billions of dollars in our operation. You know, if you think about it, our capital program is 50 to $60 billion today. I have $100 billion under construction. Imagine the integration and the value that you can create across everything that you do by adopting AI. But this is to create for the enterprise, for Aramco and Aramco Group. But for the future, you need to see that, intelligence in terms of running the operation scaled up across the industry. And this is I think we are better off working with hyperscalers who have the access to the global industries to scale, because a lot of what also we adopt and we scale is not only good for the energy industry, it's good for other industries. It could for so many other industries. But hyperscalers that have access to all of these industries other than our energy industry can play a role. But it is all about creating the right operation model, the pipeline, and also the decision making. In terms of kill, pilot or scale. You need to have a quick decision making. Otherwise, if you don't have the right process to review, prioritize, kill. If need to be killed, it's not good to, pilot or scale or even during pilot. You need to make decisions quickly, but you need to create the right operation model in your establishment for you to be able to create the value that we are seeing today. And this is what we are hoping through working with, hyperscalers in the industry. Accenture is a good example. You know, we are trying to group our use cases and see which industry might be the best user of them and how we can scale this across.
We only have a couple or few minutes left. And so I'm going to ask each of you to to maybe answer the same question. And I'm going to start with you, Julie, which is so what lessons have you learned, especially over the past year or two years, about scaling AI that you wish you had known earlier? And please do answer quickly because we just have a.
Couple, I think it's about human in the lead, not human in the loop. We will inspire people and we will run companies with people, and they will have a greater technology landscape. But we need to completely change the narrative to inspire people to paint the future. It is human in the lead, not human in the loop.
It's it should be a business bull, not AI or IT bush. If the business is not involved, you can boil it, you can scale, but you cannot have it across the whole establishment. So the business needs to be involved from day one for it to be, to capture the whole value across the establishment and the group.
I didn't realize it till ten minutes ago, but I should have hired Accenture. What? What? Julie said it hit me like ten minutes ago. What you said about leader lead is so true at visa. Like we went so fast by trying to democratize access to all of these LMS across the company. I preached it from the top, the whole leadership team. We talked about it for about 18 months, and we didn't really see the breakthrough until we got 300 of our top 300 people in the room for two days, and we forced them to go through hands on keyboard training, build agents, be supervised, be evaluated, and to your earlier point, like once those top 300 leaders had confidence to to build use agents and then lead their teams like that was the unlock for us and I wish I would have known it earlier.
There's still time, right?
I know.
There's still time.
I'm learning. I'm always learning. And this panel was no exception.
I would say spend at least as much time on the adoption as on the technology development. I think we're overexcited often by technology, and we are technology company. So definitely speaking to myself, but actually, if you want to have it really driving impact, think about how you get it being worked with. Which means that from the first moment you start to develop the technology, think about how it lands in practice. So adoption is ultimately where success is measured. And actually you need to design that in from the get go. And that is much less about technology, much more about understanding the practice that it will actually serve, and actually how you then indeed rally the humans around it that run that business or run that practice. I think that's something that we have been learning in the in the last few years.
Well, thank you all so much. I, you know, I think we kind of came back to the same place. We started about thinking about, about outcomes. And not just not just operations. And I, you know, you said something that I wrote down because I thought it was so great, which is that it's hard to trust something until you understand it. I think we've all come away with a lot better understanding about scaling AI and large organizations. Thank you so much, everyone. Appreciate it.
Thank you Matt.