AI Reshaping Government and Business: Insights from World Economic Forum 2025

How AI is helping governments drive digital transformation

Experts at two panels held during the World Economic Forum's Annual Meeting 2025 discussed the need for greater collaboration and strong governance to help governments digitally transform using tools like AI. In an increasingly digital world, governments must keep up with the latest tech innovations, as well as growing demand from citizens' for faster, more efficient and intuitive government services. But the public sector has traditionally lagged private sector-development and implementation of technology like artificial intelligence (AI). In fact, the World Economic Forum's GovTech (government technology) Network has identified a $10 trillion opportunity as governments come under increasing pressure to modernize their systems and meet the needs of an increasingly digital world. Experts at two panels at this week's World Economic Forum's Annual Meeting 2025 explored the main trends that will be driving change for this critical technology. Governments around the world are now embarking on digital transformation journeys so they can provide seamless services for citizens, while also keeping people - and their data - safe and secure. "How to understand, master and harness technology is the single biggest thing for government to get its head around today," said Tony Blair, former UK prime minister, now executive chairman of the Tony Blair Institute for Global Change, during his opening address for the Governments, Rewired panel at the World Economic Forum's Annual Meeting 2025 in Davos on 23 January. While many governments haven't fully internalised the fundamental nature of the change new technology like AI is bringing about, Blair said that finding a way to properly harness AI will "transform everything". Governments around the world are now embarking on digital transformation journeys so they can provide seamless services for citizens, while also keeping people - and their data - safe and secure. "How to understand, master and harness technology is the single biggest thing for government to get its head around today," said Tony Blair, former UK prime minister, now executive chairman of the Tony Blair Institute for Global Change, during his opening address for the Governments, Rewired panel at the World Economic Forum's Annual Meeting 2025 in Davos on 23 January. While many governments haven't fully internalised the fundamental nature of the change new technology like AI is bringing about, Blair said that finding a way to properly harness AI will "transform everything". There are downsides to consider, of course - "technology can be used for bad and for good," Blair pointed out - but he also argued that it can transform healthcare, education and other vital services. "Government is about process, permissions, procurement," he said. But AI can bring about changes and improvements that will allow governments to tax less and deliver greater efficiencies to citizens, according to Blair. "Many countries are already doing great things but we are only at the foothills of this revolution," he said. Indeed, several successful government digital transformations were discussed by the panel of experts, which included ministers that have participated in these journeys. In Togo, for example, the government created a digital platform to deliver payments to its citizens during the COVID-19 pandemic. It took about 10 days to build this method of delivering financial aid via mobile and 25% of Togo's adult population received funds as a result, according to panellist Cina Lawson, Togo's Minister of Digital Economy and Transformation. The government encountered several challenges during development. The platform was built in-house since a procurement process would have taken too long, but the data analytics had to be completed externally because of a lack of government data science expertise. And while Togo's citizens were very happy to use the system to receive much-needed funds, there were "legacy" groups within government who were distrustful of new technology and afraid of failure, Lawson said. The key to overcoming this was top-down support from the president, which helped to align everyone behind the initiative. The United Arab Emirates (UAE) has also used top-down support to ease the way for its digital initiatives, according to panellist, Maryam Al Hammadi, Minister of State, Secretary-General of the UAE Cabinet. She explained that the UAE's wide-reaching digital transformation strategy is backed by its president, prime minister and cabinet. One of cornerstones of the UAE's digital transformation has been the reform of its legislative system, Al Hammadi said. The aim was to protect citizens, but also to address issues such as talent and skills development by establishing new visas to attract tech entrepreneurs and experts. Another panellist, Achim Steiner, Administrator, United Nations Development Programme (UNDP), agreed that governments need to be both enablers and regulators of new technology. As well as attracting talent, this should also include establishing public and private collaborations to accelerate digital transformation. "There is no question that we will be using AI, particularly Generative AI [GenAI], probably globally, by the end of this decade as the primary way of delivering government services," added Thomas Siebel, Chairman and Chief Executive Officer, C3 AI. He said this would result in services being delivered with higher quality, greater accuracy, more satisfaction and at lower cost. While this will help make the world "a better place", he also admitted that there are risks from bad actors. So, while public-private collaboration should be a key area for "rewiring" governments, governance is crucial. At Annual Meeting 2025 panel Can National Security Keep Up with AI?, participants discussed concerns that governments are not necessarily "in the driver seat" when it comes to AI since so much of its development has been led by the private sector. Nick Clegg, outgoing President, Global Affairs at Meta, said the biggest security concern about AI is that its foundation is being built by "a handful of Chinese and US companies". As such, he said, open-source technology is going to be "vital" to ensuring AI applications - good and bad - don't "end up in the clammy hands of a very small number of private sector operators". Clegg suggested learning lessons from the rise of social media, which has showed that governance should proceed in parallel with technology development, as much possible, "not as an afterthought a decade and a half later". Listing several AI regulations and multilateral groups established so far on AI safety, he said these efforts have been "relatively thoughtful" by not trying to second-guess downstream risks and setting up relatively light-touch institutions. In China, according to Xue Lan, Professor and Dean of Schwarzman College at Tsinghua University, the government's "agile governance approach" includes a national plan and governing principles on AI development. The aim is to encourage innovation while leaving room to develop specific regulations when problems arise. Governments can involve companies when developing these regulations, Lan added. One of the first acts by newly elected US President Donald Trump was to revoke his predecessor's executive order on AI safety, while also announcing private sector investments of up to $500 billion in AI infrastructure. Ian Bremmer, president and founder, Eurasia Group, said the US is ensuring it has "the sharp end of the spear" on global industrial policy and technology. He suggested other regions such as the European Union need to act collectively as a result - something they are doing successfully on issues like Ukraine, but not yet on technology. Europe does need to "close the innovation gap", admitted panellist Henna Virkkunen, the European Commission's Executive Vice-President for Tech Sovereignty, Security and Democracy. This can happen with innovation-friendly regulations and less bureaucracy, she said. "We have to make it easier for businesses to invest," Virkkunen continued, adding that public funding of AI is happening but that private market investments are also crucial. "We have a lot of potential but now we have to really make it happen," she said. Aligning with like-minded, trusted partners will be key and panellists agreed that this is a strategy that smaller countries, in particular, should also consider for a successful digital transformation. More global collaboration on issues like security could significantly boost GovTech innovation - not just with other governments, but with private sector tech companies too. Such collaborations could help governments accelerate global AI use, while also ensuring transparency, accountability and safety for their citizens.

Davos 2025 - should we worry about AI in the hands of autocrats? Tony Blair and Tom Siebel think not

Dario Amodei, CEO at AI firm Anthropic, is a worried man - and what he's worried about is something that impacts on everyone on the planet. He's worried about what happens if autocratic governments and regimes get ahead of democracies in their use of AI: Is AI stabilizing to autocracies and de-stabilizing to democracies? This in turn leads to other questions in his mind: Can AI allow people to debate and discuss with each other and come to the truth better? Can it summarize positions? Can it encourage better processes of democratic deliberation? Can AI be be involved in the system of justice? Again, we have to be careful with that, but can it be done in a way that enables us to give the same rights to everyone in a more uniform way? Can AI enhance the promises that democracies make to their citizens in order to to do a better job of delivering on the ideals that democracy represents? As for the potential threat to democracy, he goes on to argue: I think we can fight that [Among] the ideas I've had - and probably over the coming year, we as an individual company will work to try and promote these - is, 'Can I improve the quality of democratic governance?'. One area is public services. Many public services are not delivered in a technologically sophisticated way. Can't we use AI to re-invent democracy, to inspire citizens that their government is doing things efficiently, to use the modern language, but also delivering robust services and making their lives better? Can it do better at enforcing and augmenting people's lives? Can it lead to better deliberative decision making? That's a lot of questions. These are all themes and arguments picked up in Davos this week at the World Economic Forum annual meeting. Former British Prime Minister Tony Blair now heads up the Tony Blair Institute which works with governments around the world on policy matters. He told Davos delegates: My very simple thesis is that how government understands masters and harnesses the technology revolution is the single biggest thing for government to get its head down today and the challenge is a very, very simple one. A couple of years ago, I got the Institute to do an analysis of the 19th century Industrial Revolution and how long it took politics and government to catch up with the fact that, in the real world, this revolution was going on. Even allowing for the fact that life moves much quicker today, it took them a long time. The challenge, therefore, is that for a lot of political leaders and a lot of governments, they can see and feel it's important, and they're they're trying to get their heads round it in certain areas, but I still don't think the political class, the governing class systems, have fully internalized just the fundamental nature of this change. Politicians and government officials are also looking for a binary choice, he added, one that isn't there: Often when I have a conversation with the political leaders we work with, they'll say about this technology revolution, 'Well, is it a good thing or a bad thing?'. And I say to them, 'That is honestly a redundant argument. Yeah, it's a good thing and it's going to also be a bad thing. But the point is, it's a thing. In fact, in my view, it is the thing. Now, I don't mean to say there aren't downsides with the technology revolution, and a lot of it is general purpose technology, so it can always be used for bad as well as good. But the fact is, it should transform healthcare. It should transform the way we educate. Government's all about process in the end and about permissions, about things like procurement. It should transform all of these things, because you're going to be able to use Artificial Intelligence, particularly generative AI, to make changes and to improve efficiency in a way that then allows you to spend less, tax less, but deliver greater efficiency. That's the Holy Grail that people should aim for. For Silicon Valley veteran Tom Siebel, his blunt assessment is that fears that AI will undermine democracy are in some respects too late: Are there risks associated with AI in the way that it will be used by autocrats, in the way it is used by bad actors? Will it be used to undermine democracy? Absolutely. I mean, this is not a future; it's being used [like that] today. Anybody involved in Brexit? Anybody pay attention to the 2016, 2020, 2024 elections in the United States of America? People, this is happening today. We don't have to wait for it to happen tomorrow. But the C3.ai CEO is clear in his mind about the benefits AI can bring to both governments and citizens: There is no question that we will be using generative AI, particularly generative AI, probably globally by the end of this decade, as the primary means by which we are delivering government services. I'm not certain this strengthens autocrats and de-stabilizes democracies. When we get into the delivery of government services, whether it's benefits programs, whether it's public health, whether it's education, these services will be delivered at higher quality, with greater accuracy, into the hands of more satisfied constituents at lower cost through generative AI - hard stop. Let's look at the future, he added, and the positive changes it might hold. Gesturing with his iPhone for emphasis, he argued: In the not very distant future, we will have 6 billion people with the entire knowledge of mankind available in their pocket. We will have 6 billion people who are polymaths, like Voltaire, Thomas Jefferson, Benjamin Franklin, and they will have instantly available everything there is to know about healthcare, agriculture, animal husbandry. We're effectively raising the IQ of planet Earth by a substantial amount. There's a direct correlation between that and productivity [and] there's a direct correlation between productivity and GDP growth on a global scale. His overall conclusion about bringing AI to bear on government - on balance, it's more than worth it, he said: I realize everybody wants to bemoan the downside and terror associated with AI and how your smart refrigerator is going to take over your house. I get it. But there are some, very, very optimistic outcomes. A timely set of discussions to round off a busy few weeks in public sector tech transformation initiatives that have seen major announcements in both the US and the UK. It's clear that AI's role in government services delivery is one that has enormous potential, but given the track record of the public sector on both sides of the Pond when it comes to digital transformation, it's a potential that may need radical revisions to the established order to come to reality. Blair put it particularly bluntly here: I do not think there is any point in trying to hand all of this over to existing bureaucracies and civil services and say, 'Sort it all out'. A lot of it requires deep specialist knowledge, and there's got to be a completely new relationship between private and public sector, where we're harnessing a lot of the expertise and ability that resides in the private sector. Loathe though I am to agree with anything my erstwhile Prime Minister has to say on tech - this is the guy who personally signed off on the UK's NHS National IT Programme, a globally totemic public sector tech failure that cost the British taxpayer billions of pounds and later dismissed by political opponents with the line, 'You were the future once' - on this, he has a point.

Davos 2025 - will the last CEOs to deal with human-only workforces get to the office in robo-cars? Salesforce, Anthropic and Uber contemplate AI potentials

AI runs through the thematic DNA of this year's World Economic Forum gathering in Davos as it inevitably would, with discussions ranging from long term 'blue sky thinking' to the more immediately practical. One of the more unexpected examples of the latter came from the moderator of a panel session who said that his Davos app had told him it was time to go home after being out and about and having several drinks the night before. Thank Salesforce for that bossy app, it seems. Salesforce has run information management for Davos for over a decade, as CEO Marc Benioff pointed out: So when you're on the Davos app, and you're going through all the lists of all the sessions and who's attending and all that, and you figure out how to make the most productive use of your week, you're using all this Salesforce technology. But this year there's an added element in the mix and anyone who's observed Salesforce's pivot to agentic AI in recent months will not be surprised to learn what it is. Benioff explained: This is the first Davos that you'll notice that right on the Davos app, there is an agent, and this AI agent is there to help you, to be your guide, to be your partner in making sure that you're able to have a great Davos. What it does is it goes back and it looks at every session you've ever attended, because all of that is recorded as you badge through the conference, and every session that is available, and everybody that's here, and then that agent can say, 'Yes, this is the thing you need to do in Davos to be really successful right now'. It uses the best Large Language Models, but also the best machine learning, the best intelligence. It delivers the highest accurate result. It's delivering a 95% result for Davos attendees because it has access to so much data that has been collected here over so many years. It's a great example of very practical AI. Benioff added that at Davos this year he'd encountered others making similarly frontline use of agent tech, citing Dave McKay, the CEO of the World Bank of Canada: He's got this incredible wealth management business that he runs, and he's a leader in this area, and they've built this agentic layer. Agentic layer means he's built agents around his business that augment his workforce. So in addition to his sales workers and his service workers and his support workers, he uses an agent force, just like we're using here to run All of which leads him to another conclusion: We're going to be the last CEOs who are only going to be managing humans as our workforce. From this point forward, we will be managing not only human workers, but also digital workers...That's just fascinating to me, that we are at that moment that we're going to say, 'Yeah, this is when humans and AI are working together to create a higher level of success'. That's a scenario that resonates with Dario Amodei, CEO at AI firm Anthropic, which is currently working on what he pitched as a "virtual collaborator". This is something that might not be "smarter than a Nobel Prize winner" but: [It will be] capable of doing relatively high-end tasks in the workplace, that can open up Google Docs, that can use Slack, that can interact with its co-workers to perform tasks over hours to days, and that you only have to check in with every once in a while, like a manager would check in with an employee. We're trying to make that real. As to how long this will take, Amodei argued: The thing that most distinguishes AI from previous technological revolutions is the speed at which is happening. I've been in the field for for 10 years. I've been watching that curve for 10 years, and me and my co founders at Anthropic were among the first to document what we call the scaling laws, which is the the observation that as you pour more compute into these models, with small changes in the algorithms according to which they're developed, they improve very fast on cognitive abilities. And so, exponentials kind of start to flow, and very quickly pick up speed. I've become, probably in the last three to six months, more confident that we really are heading towards AI systems that are better than almost all humans at almost all tasks. There are bumps in the road ahead still to overcome, of course, he acknowledged: I think what is going to hold us back most in positive applications is the physical world and limitations on human institutions. So I wrote an essay about powerful AI [and] if this really does happen, what will we be limited by? The term I used was marginal returns to intelligence. Economists talk about the marginal returns to capital, to labor, to land. A way we're not used to thinking about things is, what if we have millions of copies? I called it a country of geniuses in a data center, of agents that are better than humans at everything? What limits [are there?] Do we just immediately solve all of the world's problems? I don't think that happens. For example, self-driving cars are difficult. So I think we're going to be limited by the physical world, by deployment cycles, and often by laws. That brings in Uber to the debate with CEO Dara Khosrowshahi focusing on how AI will have an impact on the physical world: Ultimately, the future of transportation is going to be autonomous. It'll be electric, and eventually will be shared as well. Congestion on the streets is something that can can be controlled,. Instead of having a human driver, you'll have a robot driver. We're working with [Alphabet] and Waymo and a number of other companies in developing this technology and really bringing it to market so that everybody can experience it. And experience it safely, he added: The ultimate promise of AI is literally to save lives, right? This year, there will be a million fatalities on the road as a result of human error. This digital driver is going to have thousands of lifetimes of driving to train [it]. It is not going to be distracted by texting, by watching TikTok, whatever. It will be a driver that's completely focused on the road. It won't have two eyes, it'll have 10 eyes and many different sensor types. So by definition, over a period of time, this driver is going to get better and better and better, and it will start to be more of a reality in our network. Our view is, we're working with a number of partners to bring this technology to market in a safe and responsible way. It's not going to get there tomorrow, and there will be a very long hybrid road where sometimes you're going to get a human driver, sometimes you'll get an AI driver or robot driver. But fast forward 15 or 20, years from now, the impact of AI in the physical world, and the impact in terms of saving lives and avoiding fatalities is going to be enormous. As we keep saying, practical examples of AI in action in the real world are critical at this juncture in the current hype cycle. The driverless, robocars future is all good and well, but it's things like the Davos app that people can 'touch' today that start to make sense of the potential of this most disruptive of technology revolutions.

Davos 2025 - Successful AI adoption demands 'show, not tell', strong change management, and HR re-invention, advise AWS and Accenture CEOs

A welcome realization about AI that we've seen more of in recent months is that successful adoption is about more than technology; culture change is every bit as important - and that will mean organizations thinking again about HR and change management. On Day One of this year's World Economic Forum Annual Meeting in Davos, two AI industry leaders, AWS and Accenture, picked up on this idea as their respective CEOs outlined their own experiences of 'eating their own dog food'. For AWS, CEO Matt Garman emphasized a by now familiar theme from the firm about Amazon's long experience in tapping in AI internally. This has made the current generative AI hype cycle something easier to manage in terms of winning take-up within AWS: We were fortunate to have some of the best technology experts in the world, and so they jumped in, and in many ways, were driving a bunch of the gen AI technology. Getting that set of people to dive into the technology wasn't really that hard. It's like kids in a candy store for them, really, they're quite excited to dive in. We have some of the best science experts in the world, some of the best technology experts in the world. So, that part of our business was quite easy. But there were other factors that needed to be taken into consideration, he recalled, with customers looking to AWS to be experts and provide advice/knowledge. This impacts on more than the techies and means that the likes of sales people on the front line need to be trained to be up to speed: Fortunately, we have that expertise in-house. We're training the rest of our team all of the time to be at the leading edge of technology. We've built that muscle over the last 20 years that we've been building AWS, so whether it's new technologies across the board, we've kind of built that mechanism to do training. Our customers look to us to train them as well and so we've actually built up a whole organization that's focused on training customers. It's free training. We train millions and millions of customers for free on AI because we feel like they're not going to get the most out of technology, they're not going to get the most out of what we do, if they don't. Again, the idea that successful AI adoption transcends tech comes into play, he added: It's actually not how you use the technology. That's not actually the training. The hardest bit is getting people's heads around what's possible and, and not even what's possible now, but what's possible 24 months from now, or 48 months, or two, three, four years out. I think people have a hard enough time realizing what's possible today, but as you're implementing it 12 months later, the technology is going to be much further along, so you really have to think about what's going to be possible and unlocking it. That's not necessarily as straightforward as it might sound. Garman explained: It really is unlike any technology that we've had before. It really is going to change every single job in every single industry in ways that I think it's really hard for people heads to get around. It's not just doing something 5% better, it's doing something 1,000% better or more, that order of magnitude. And it all takes time on the part of AWS, he noted: We have to sit down and help them understand how can their business actually get that much more efficient, that much more effective? How can we change customer experiences that way? So internally, we had to get our teams to think bigger like that too, and really push them to think what's possible out there. I think as soon as your employees get their heads around that it's not that my job's going away, or it's not that this is going to save us five percent of cost, but, that we're going to fundamentally be able to go deliver things weren't possible before, if you at least have the right sets of employees, most people get pretty excited about that opportunity. One way to convey this is to 'show, not tell', he advised: Lean into a particular use case that that takes work and toil off of the rest of your team, that unblocks something. One example we had of this is we built a product that automatically upgraded Java version, If you talk to software development teams, that is a task that not one person is super excited about doing. No one's like, 'You know what I really love is just upgrading Java'. It's boring, but it makes sure you have all the latest security, it gives you performance benefits etc. We built an AI system called Q Developer Transform that just does that for you, and so we decided to eat our own dog food. We said, 'OK, we're going to go do that across Amazon'. The teams internally estimated that to upgrade all of our internal systems was gonna take somewhere around 4,000+ man years to upgrade. We had a team of five people that did it in a couple of months across and so we freed up 4000 person years of effort to go build features and capabilities for all of our customers by taking away work that no-one wanted to do in the first place. It is a very good thing where you just lean in, you just say, 'This is a capability that we can go and roll out and and it's good'. It highlights that when you find that thing, it has a massive impact on your company. Accenture CEO Julie Sweet picked up on the new role that existing HR teams need to face up to: What this requires is a completely different thinking about how people work and what skills they have. In most cases, AI isn't replacing people, it's replacing tasks or parts of the processes, which means that in order to upskill people, you have to understand what skills do they have to begin with? At Accenture, we moved to skill-based HR over five years ago. I have a database of almost 800,000 people and their skills. We're systematically re-defining the skills needed at Accenture - who needs gen AI, what kinds of technology? - but also, as we are replacing some of the things that they're doing with gen AI, we're able to identify who could be upskilled. So you have to have skills-based HR. Enterprises also need to have HR can adapt. Sweet cited Accenture's work in introducing AI into marketing: We built an agent so [employees] can give feedback just like you would give feedback to someone you were supervising, or feedback to a colleague. That suddenly makes the AI very real to people, and they start to see how it can really help them. That's important when dealing with transformative technology, he added: That's a totally different mindset. We didn't have a playbook that says, 'How do we take people who are now going to use AI fundamentally in their job and make this be something that they really care about?'. Finally, HR needs to get its head around the importance of change management, she advised: The kinds of change management that you have to do here require you to understand the technology and understand the people. In most cases, companies use change management to train on a new system, or they use whatever partner they happen to be using. This is a skill set now that has to be at the center of your company, because it's a continuous re-invention. The amount of opportunity that AI is going to change in countries and companies is going to take decades. It's not a one and done. So building change management so it's not viewed as being something off on the side is really important. All of this is vital for successful AI adoption, she concluded: At the end of the day, the technology will only be effective if it's trusted and adopted, and you can then create new opportunities for your people. We have to think very, very differently about the incredible departments of HR, and the role that they need to play and how they have to re-invent themselves. Interesting perspectives from two organizations that are both users and purveyors of AI.

Davos 2025 - giving AI responsibility to CDOs results in great AI no-one uses, says Sanofi CEO Paul Hudson

[People] delegate AI to CDOs, and it's the biggest mistake you can make. At last year's World Economic Forum gathering in Davos, Paul Hudson, CEO of French pharma giant Sanofi, made the case for AI being a technology that would augment, not destroy, jobs and the nature of work, arguing: I think there is this journey to more meaningful work...We don't spend enough time talking about what this will enable that is not possible by human beings. Making sure that this message sinks home, Hudson was back to pick up the theme this year, expanding on how AI has changed the nature of roles, including that bit of 'tough love' for Chief Digital Officers (CDO) everywhere: I have a great CDO, but the nature of the change and the courage needed to change business process at scale, to better optimize resources and to go chase miracles for us, means that if you delegate to a CDO, you have automatically lost. You have great AI that nobody uses. Too many cooks spoil the proverbial broth, he added: When you're trying to change the business process, that's where the real win is, and that's why I have to lead it myself. It's not just about me, it's just I'm the arbitrator to make sure we get the adoption that we need. If I have a call to action, don't over delegate. There's also wider home truths to be levelled about the pace of AI adoption particularly among some of Sanofi's peers: I talk to a lot of my peers about the adoption of AI...and people are still moving too slow course. Part of it is that it needs to be sponsored from the top, because it's still unknown to so many people, and the scale of it, so many business processes have to be changed. If you're not going to do that, it's a sort of nice addition to the tools that you have. He added: While most big companies are piloting and dabbling, most CEOs are a little bit apprehensive and are running lots of proof of concepts because it's safer just to do that for a few years before retirement, We've decided that everybody will be in the same place in the end in healthcare, so being first and the broadest, which is where we believe we are right now, gives us a huge opportunity to beat others. And that's happening, he boasted: In big corporations, there's a lot of managers and senior managers who've made careers out of 'polish and slide decks' to make sure that they can communicate upwards what we should think about a situation. With radical transparency, we've sort of blown that away a little bit. People can turn bad news into good news with the right PowerPoint presentation. We're trying to really avoid that...I think in healthcare in general, people are a bit slow. I think in the race with the turtles, we're the lead turtle. And he can prove it, he reckons, citing examples of how Sanofi's adoption of AI tech has changed operations: We used to, when I joined the company, spend a week doing a budget process. Now we can spend about three hours because AI lets us know where we're heading, use it as a base case and adjust resources from there. We're deploying up to a billion dollars in real time, enabled by an agent to allow us to do that. Ai is also changing the nature of Sanofi's core business, Hudson said: It takes about 12 to 15 years to discover and develop and bring forward a new drug and Large Language Models are what two three years, at least, in practical application,. We spend about two to three billion to develop and discover and develop that medicine. At the same time, roughly a third of the discovery effort for the drugs that will launch in 10 years are being validated by AI,. People forget there are so many diseases that are considered undruggable and so many diseases that are unmet in terms of needed treatment that we're using AI to genetically validate targets, to make sure we're likely to bring forward a medicine a decade later. Validation is a crucial component of the organization's work, he added: We generally validate everything, and target selection is more than 90% accurate now, but we spent an awful lot of time putting together our data, making sure that we have it end-to-end. We've completed that. We're running validated agents now at scale, with more than 20,000 people using them on a daily basis. They'll be from as simple as inventory management, but the agent is correlating all the data real-time, almost like Instagram notifications. We're doing very much the same in terms of forecasting accuracy, more than 99% accurate in terms of use cases. Agent tech is playing a big part in all this, explained Hudson: We spend around €8 billion a year on development, and the key decision-making body, our drug development committee, begins its sessions with a validated agent recommending whether a drug should pass through a toll gate, whether it goes to phase three or goes to phase two. We do that because it's very sobering. The agent doesn't have a career at stake, the agent isn't wedded to the project for the last 10 years. The agent is dispassionately saying, "Don't go forward or go forward faster or go forward and remember these things'. We're not used to having somebody without a career at stake in the room at a senior level. What about the 'AI is going to take my job!' Paranoia? Hudson has one more blunt message born of experience: That's really not the question. Whether it's a message that those 'slow CEOs' he talks about are ready to take on board remains to be seen.