Realising the AI advantage for the UK

I'm Katie Prescott. This bonus episode of the Times Tech podcast is brought to you in partnership with IBM. IBM brings together technology, infrastructure, software and expertise and has consulting services for business across the world. It's got a long history of being at the forefront of tech and adapting AI tools and developing quantum computing. With the AI revolution at the moment, many business models around computer services are again in flux. So IBM has been putting together something of a roadshow to explain how it's adapting and looking at how AI may be changing what we see as productivity, asking what businesses need to do in the new age of AI and what's holding people back from joining in with this change. And to find out more about how IBM is meeting all these challenges, I sat down first with Leon Butler, who's their chief executive for IBM UK and Ireland, and then Dr. Juan Bernabé Moreno, who's the director of IBM Research in Europe. So to Leon Butler first, and his excitement about this new age of technology for businesses like his. AI really landed, I think in the public consciousness, generative AI really landed in the public consciousness first say in November 2022, which is almost exactly three, well, which is exactly three years ago this month. And now it feels like it's everywhere. And I just wonder what your reflections are on this rather wild three year period that we've seen and how things have changed for you and customers in that time. It's accelerated incredibly quickly, but the one thing I saw immediately were how many people died into AI, a lot of pilots were kind of kicked off. But people, I think, can see very quickly productivity gains, which was which was great across the world. I think in the UK and Ireland and certainly UK in particular, I mean the government see it, right? I mean, they've got their AI opportunities plan and they feel that it would generate 400 billion to the economic growth in the next well between now and 20. 2030 as well. So there's a huge amounts of potential there. And I think everyone, everyone really kind of grossed and sees it. I always got a joke, which is a lot of people who come around the house be quite bored when I first was talking about it suddenly everyone was to talk around the ice. I become a really popular person during during dinner. But I think everyone intuitively has used it at the next pieces in delving into actually how do you scale across enterprises and hopefully we'll get into that. Yeah, because when people use chat to be two for the first time, it allowed everybody to touch and feel and understand its capabilities, whether they were using it to write lyrics in the style of Bob Dylan or whatever it was, you say I'm sure people coming to see you suddenly really understood what you did. But it feels to me that there has been a bit of a disconnect between seeing the opportunity in AI and seeing its power and actually implementing it. And I just wonder what your reflections are on that. Yeah, I think from a user interface perspective, that's what opened it up suddenly everyone was using the technology and it became accessible everywhere. And I always say this a lot, I think we get into the point where people are intuitively using it, but it's great for maybe things like personal productivity. Not necessarily at the moment have people grappled with it when it comes to as a scaling it for enterprise and looking around that. So certainly when you look around things like domains around HR procurement supply chain. Things that you can really get good productivity gains for is where I think it needs to needs to kind of I suppose dovetail next. What are businesses adopting and what are you seeing working well from a business adoption perspective, I would say again what we're seeing so far. Has been basic summarisation of minutes as an example extraction of documents what we are seeing though our businesses now really grappling around I mean if we just take HR as an example, we have something called ask HR, which is a series of agents that actually do a number of different things, whether it's around talent management, whether it's around recruiting whether it's around. Task simple tasks around maybe getting a p60 to moving an employee and we had an assistant quite some time ago and we kind of launch ask HR to be able to use natural language natural interfaces to actually ask a question and get a get a process done. And that means that ultimately people would have spent a long time maybe trying to move an employee whereas now you can literally use natural language say I want to move this individual this person and it will tap into that back end system and do it. The key kind of aspect is making sure that you have a technology that actually goes into multiple applications because what you start dealing with is a process at scans or code goes across multiple applications multiple environments and not necessarily an agent that just taps on to one particular technology. And so by the time you and I talk around my talent management piece if I want to talk to my team around talent management that may be an email out to my team says let's look at the team that kind of comes back they have the ability to actually look at their team say this is what we want to do. It may cross over multiple applications but it comes back into one place to have that technology be able to do that can be complex on those particular processes and we have something called what's next orchestrates that actually orchestrates that layer so connects to multiple applications multiple environments and actually allows you to get an answer through what could be a really complex process. I should lots and lots of businesses would love to roll this out. I should reference that very famous MIT survey that went viral recently that 95% of projects in generative AI deliver zero returns what. What's your assessment of what is holding companies back from implementing say you know the HR model that you just described. I'll put it into a couple of categories actually there's okay I think complexity is one of them first of all so complexity in the fact of the average enterprise got about 1000 applications that's quite a lot of applications those applications. Well we'll we'll set across multiple public clouds private clouds on premise we think over the next three years will be a billion AI applications that we generated a billion applications that's huge. But it's quite terrifying for companies looking at. It is it is complex and again then when you overlay and the fact it could be a private clouds multiple public clouds are on premise that does become fairly complex are having the ability to actually deal with those. Applications across the environment is going to be really really important and. IBM's approach is very much around hybrid clouds are giving the ability to one automate and orchestrate across those environments so being able to connect multiple applications multiple environments as a step between private and public and on premise that's going to be really important. Automate the applications at the infrastructure being able to have control of those environments be able to put limits on it as well so you can actually control where those applications are. But how they talk to each other as well that's going to be really important data is another one so we feel that. About one percent of enterprise data is made its way into a large language model some form a description now enterprise data is a crunch of any organization. If you look at that enterprise data it's also unstructured so can be quite complex. So we need to kind of get into the ability to tap into that information again information can be siloed across organizations and it can be across those. Environments that I mentioned so private public cloud is an example so again having that ability to get that information be able to track things like the metadata changes all the way through so from the source of the way to the AI application is going to become very important so by time you make a loan decision or a hiring decision you know it's on accurate information. And you know I will say it's around trust and transparency as well and it is one of my one of my analogy so I say quite a lot of the moment there's a bit of an inflection point in the industry closed versus open. So you can have models that maybe are trained that you don't know what is trained on as an example and you can have open models which which are public sizes being clean and trained and. The analogy I do give and you may have heard it may not is the kind of the opaque test tube piece which is you've got a test tube which is opaque so you can't see in it. There's some liquid in it that may be the data that that particular models on. Then what's quite in it pour some more liquid in it as your own data shake it up you never drink that liquid. And it's the same it's the same with information and data mixing that information you need to understand where is the models are being trained on and you need to know how it's been mixed. So actually by the time you make that decision it's going to be clean and transparent and trusted. So I think that's going to be another piece on it and the third aspect to skills. If you look at skills at the moment I think by the way I think UK government's made a great step around the AI training for seven half million people we're part of that particular announcement as well. We have something called IBM skills built so free education that people can look kind of learn around AI or content. So having that ability and we're not the only vendor there are other vendors and and suppliers certainly do kind of have those kind of free skills out there but that's going to be really important because we need to make sure that we have the right skills. And that could be around training and also apprentices actually you know really leaning into making sure we have the right talent coming through the organizations but that's going to be really important I think as we go through. And then the last one is most trust and governance. You know trust and governance I think in particular is something that we need to make sure again when you're making the right decisions it's based on trusted and govern technology. And a lot of organizations I think certainly from the surveys that we saw 85% of businesses feel that they will tap into AI and trust and they think is really important. But really under half of those are actually making those steps towards actually having that real kind of governance around there and about 27% really feel that they're making any forms of great news around dealing with bias as an example bias around making a decision. I think people sometimes are a bit worried about that and therefore may not lean into a going from the pilot to production because they wanted to make sure that they actually have those tools in place. Again we have something called what's next dog governance that allows you to be able to monitor those particular technologies as well. How much of a problem are legacy systems and how much do people have to upgrade their IT infrastructure when they're looking at rolling out AI. I think the key thing is not starting from scratch. You know people have invested in systems and legacy can mean number of things as well. It could be just systems that maybe haven't been upgraded that could be upgraded as an example. So it doesn't have to be completely and again that does come back to kind of the open technologies technologies that are able to kind of connect to those applications and those infrastructures. And again when I talk to around that what's next orchestrates that has the ability to connect to a sales force or an SAP or an Adobe those different kind of backend systems so you don't have to rip and replace. It's really important to have the technologies that can orchestrate and span across those applications. The other risk around the open close piece which is you you know we are ones that are open so you do have other players out there that have close technologies and they have an agent that will connect or a technology to their own application. Which is great maybe for that application but again if you talk if you look at that set of complex processes I talked around even around those HR processes and they span multiple applications you need technology that can connect to those different applications as well as across private and public clouds as well. So that will deal a lot with complexity that is out there in the moment that customers are facing. So you mentioned agents let's jump in and talk about agente AI so these are essentially bots which can go away and do end to end tasks for people and are coming down the track. Incredibly quickly when do you think this will start to be part of every day life for business I think the boss of sales will smart venue off so you soon going to employ more agents than people. So what's your assessment of agente AI and its impact. So that firstly is really exciting throughout kind of consumer surveys about three quarters are pretty comfortable using AI powered assistance as an example. Now we get on to the kind of the agente piece on it as well an agent is very much a machine learning type model that is able to kind of span multi do a task across multiple multiple processes. Even a decision and more of an autonomous completely yeah so we'll make decisions with human oversight which is really important I think human oversight is definitely an important kind of aspects on it as well but ultimately automated process is all make the decisions across those complex processes as well. Really helpful for personal productivity and business productivity and you'll have multiple agents and you need multiple agents you won't have one agent that's going to do everything just like you won't have one large language model Swiss Army knife to do everything you're not going to have an agent to do everything. As well you have fit the purpose agents that will actually deal with a task the risk is you get agent sprawl as we call it which your agents everywhere and it's a bad thing. It's a bad thing when we talk around the complexity again if you if you are an enterprise with multiple different agents doing multiple different things. You need to make sure again that you've got the ability to orchestrate or control those particular agents so if they are doing an HR task if they're doing a sales task or a procurement task that you can coordinate them. You can coordinate them make sure that they're accessing the right information that is secure as well they actually they are making the right decisions which is around the technology and human oversight that's going to become really important otherwise you could see really get out of control so. Having the ability to gain orchestrate across those agents to coordinate it where those agents can tap into multiple applications is going to be very very important rather than the kind of the locked inversion of an agent maybe on one particular application. So what sort of things do you see agents doing just to explain to listeners what might an agent's task look like. Yeah maybe just going back to that kind of HR example that ask HR again where you know you could use it for as I said you know I would ask to move from my global job to my UK job I had to transfer my team as an example. Historically on a back end HR system that could take a fair amount of time to be able to do that actually try find the person try understand the process. Whereas I was able to use natural language to say okay I want to transfer that person is this person in their team are these other team members in your team yes would you like to transfer them who's the manager yes. And that process is then automated to be able to transfer those individuals to that person that would have taken a huge amount of time and an HR agent and assistant are ask HR assistant can do that. You know we had a lot of people at one stage that were taking phone calls from managers trying to move employees almost I call it that how to type phone calls. And that allowed us to kind of measure exactly how long people were taking to do those particular tasks you can then start putting time and dollars around it to actually say or pounds as to how much time you're wasting doing the sort of processes and I don't think any enterprise any organization wants people to be spending huge amounts of time doing that you can use that natural language to be able to do that so that would be maybe one example. Versus may be something that's dealing with talent acquisition or you know prospecting for a sales person or procurement as well so those domains specific agents are going to be really important so not necessarily tell it to an application. But actually doing some form of HR task or procurement task or sales for sales task and when do you think businesses more broadly are going to feel comfortable using agents in that way. And letting them go off and do well. Yeah, it's so we're seeing it already we are seeing it already there's no doubt about it as well so one of the examples we have and certainly UK references university hospital worries here and Coventry where they have something called people assist I think it is but it's a equivalent of a nature kind of assist agent that we have there as well and save them about 2000 days of effort per year already on it as well. So so we are seeing both private sector and public sector so so we're seeing that there and I suppose kind of going back to the other part I think you know genuine that consumer survey with almost three calls of people feeling comfortable using AI powered assistance that abilities there I think it's then how do you deal with the complexity around that as well and actually making sure you're picking the right technologies that are open to connect to those applications. So you did this incredibly extensive survey of three and a half thousand business leaders about their attitudes to AI I mean you mentioned sort of headline finding was there anything else that struck you from it particularly. Yeah, so we did a couple of surveys one was the consumer survey the 2000 people the other one was what we call the race for return of investment or race for ROI. This is a boardroom saying we need to implement AI how do we do it and what's the impact and truthfully the return of investment in it as well. So because that's ultimately if you don't get to a point of return investment and then quite rightly you're not going to scale across the enterprise because you're going to want to see some form of productivity gain. So I you know from those particular surveys and they were across Europe Middle East and Africa as well as UK and Ireland about 500 businesses in UK and Ireland and about 66% of those enterprises felt that they were getting significant productivity gains so that's a really good thing. 63% of those leaders felt that they were getting operational efficiency so again it is it is great from that perspective but actually and another statistic for you about 62% so those enterprises felt that they hadn't tapped into its full potential for productivity gains on that one as well. So this is where there is a lot more kind of gain that we can get out of those and that for us feels as if it's around those domains that I talked around so again having those agentic domains around HR supply chain procurements accounting so those domains that are not necessarily just for a specific task but for that entire domain on that one as well. I just feel like there's a lot of FOMO around AI I feel like I speak to business leaders all the time who say we really want to be using it but frankly behind the scenes we're not quite sure how to implement it but we're sure other people are doing it better than we are. And I just wonder what you think about that and how you can advise people to cut through that hype and focus on productivity gains and where the successes are there and whether you're hearing that FOMO. Yeah well I think it's there and it's real because you either have productivity gains and efficiency so you know that can be around dealing with cost or it can be around generating revenue so as a business you're competing I think it's really important on that one. You see that what what I said and certainly I saw it in the world why role I had was that again these pilots that were good science experiments but actually how do you get them into production. And I'd certainly say it's around picking the right business case and the right process and working back from there rather than just saying we're going to play around with technology so if it is solving something around talent acquisition as an example or an HR kind of aspect or a procurement aspect or invoice to cash actually saying right that said I'm going to deal with that particular process understand how that is that's dealt with and then actually pick the right technology to be able to deal with that that's going to be really important as we go through picking that outcome and working backwards rather than just say I'm going to play with technology so that I think that will be very well it is very important rather than those science experiments that we saw. You've been in the tech industry more than two decades yeah yeah ask you to do some crystal ball gazing then when do you think we will see AI implemented across businesses in a way that you know I guess the changes happen during the dot convument bus which is the comparison that a lot of people are drawing at the moment. Yeah as in kind of what's coming next yeah well no I mean I just mean more when we are going to see it adopted more widely we are only three years in I think it was sitting very very early days. Yeah look I mean definitely from from those kind of those survey results it is being adopted there is no doubt about it as well and I said it feels it's more around personal productivity then kind of enterprise on that one. I think as we're kind of looking forwards I think we are going to have to kind of grapple with some of those technologies to make sure it's implemented so that complexity that I taught around around public cloud private clouds on premise we've seen a huge investment in the UK which is really exciting. That investment is being very much around infrastructure numbers of GPUs something inhibitors around that as an example around you don't necessarily need a trillion parameter model to solve a query around HR as an example. We have the ability as well as others we work with our own models granite still granite is called as well as other trusted models that are smaller and more agile that take less numbers of GPUs or LPUs as an example. So that's going to be I think one of the one of the kind of the big I suppose accelerates is which is you don't necessarily need football feels full of hardware and GPUs deal with it which can be inhibitor I think the other pieces around as said open technologies I'm really passionate about making sure we have the right skills but also we start building up. Businesses because ultimately the you we taught around the 400 billion that I can contribute that doesn't mean making sure that businesses can actually create the applications and the jobs to be able to do that at that again is around open technology open technology means you can start utilizing codes and start applying it to your own solutions when you have open technologies that then starts allowing people to create applications those applications will then start connecting to those bits of infrastructure. The infrastructure is almost like a set of roads that we're building the UK and island at the moment is great but how do you connect them all together it takes time and but you can do that through making sure that people are trained in the right way using open technologies that then allows you as a said to create skills applications those billion applications I talked around you're not going to do that if you got close technologies if you got open technology that helps and you know we've had a number of kind of pieces I was in the invest summit in Birmingham recently as well. And there's some great resources across UK and islands who are really keen to site generating you business but again it's making sure you have the technologies to be able to do that. I was like from what you're saying that we might be maybe a decade away from it being embedded in our everyday life. Now I think we're I think we're here now yeah completely it's I think this is now tapping into the next level of those those gains that we need to get but it is those gains across the business and you know that is making sure that actually as as UK and island we are competitive as well so you know building the economy as well as competing with other countries. What do you want to achieve from today's event what do you hope that people go away go away with and I think it is moving towards actually implementing AI at scale not necessarily these small little projects now but saying okay I understand the technologies that are out there I understand the limitations and the benefits of them. But start leaning into actually those bigger enterprise why use use use cases and and pilots turning into production I think that's going to be really really important because again we can see the benefits I mean IBM wise we saved over last couple of years is about three and a half billion dollars of productivity gains and something called client zero so we do it to we we use our own technologies at the moment. And this is why I'm again I'm really on the fact that it can be done you can quantify those hard to do processes that I talked about can be automated and they can be automated across multiple different environments as I said whether it's private clouds public clouds. Being able to say okay I'm going to grapple that process because I want to get to an outcome and work backwards the technologies are there to be able to do it but we do need to lean in a little bit more onto that one say okay rights. I've maybe done one simple process how do I now deal with it from an HR supply chain or a procurement process as well because that capabilities there and we'll continue to do that as we go through that three and a half billion dollars is measurable is quantifiable and other companies can do that so I think I think that will be where we go next. The final question then I mean you're in such a unique position to answer this when you're when you're speaking to people out there or businesses generally what is the biggest question that you get asked about the rollout of AI. The big everyone want to know the biggest I think one of the biggest things I see and it's a cultural thing in companies at the moment which is again if you're if you're going to make a kind of move to make sure that you've got an agent that is helping you automate. Processes across things like maybe HR or supply chain you've really got to make sure that people trust to do it now. You could get to a point where you said okay I'll play with it a bit but I'll keep those systems in there and you need to make sure you trust that technology to do it and then also lean in behind it to make sure that you actually do it otherwise you end up having almost. Shadow it you know you get to a point where you've got the AI kind of abilities to do it but you still got the same processes there and going back to. I think what I said earlier on which is we had lots of people phoning the HR helped us saying how do I move that person they were doing it because it was easy and you know they had their local person they could phone and say right how do I do that. You need to make it easier to be able to do those automated processes but then you've also got to give people the ability to do it so we we have something internally called the what's next challenge which is our technology and we actually ask all employees across the world to go use our technology to go solve problems across there we vote on it we get the best ones that we go implemented. We make sure that generally all our employees utilize the technology become very familiar with it as well because that becomes really important so as we go through when I talk to around skills and using the technology that's going to be really important. Whereas I think sometimes people think actually right I'm going to put it in tomorrow and it's going to magically do it you still need to make sure actually people are afraid with those technologies and can implement it as well and can can use those technologies. Thank you so much and I think we're going to have to let you go you're a busy man it's a busy day. So Leon Butler went back to the IBM London event and now we turn to Dr. Juan Benabe Moreno who's IBM's Europe director of research and as we heard IBM has a long history of research and innovation and this is itself now shaping a our use. As well as the pursuit of quantum computing but first let's take the view from space. One could you just tell me a little bit about the partnerships with with NASA and the European Space Agency. Well we've been working with NASA for decades and we've been part of the Apollo missions so we had a team of 4000 IBM engineers in Hansville Alabama and we've been very integral part of helping NASA get the month to the moon and now we are. So we are enjoying a very very very fruitful partnership in terms of advancing science with AI technologies and not only with NASA but also with the European Space Agency and it's a very vibrant ecosystem at the moment. To tell us what for people who weren't really know what those partnerships mean what sort of applications you seeing AI having on earth but in oceans and space and scientific research. Let me go back to 2023 right the chat GPT moment yes we did have a moment as well work with NASA. And it was along the lines of look we see how this wonderful transformers architectures can do with language we see how chat GPT has changed the way people interact with AI systems. Can we apply this idea of this architecture to more interesting data for example the data we gathered with our satellites orbiting Earth. And by doing that we wanted to lower the entry barrier for the community to start working with Earth observation data because they are key if we want to solve the big problems that we have as humanity for example understanding this kind of a flooding. We are partnering with the Kenya government to help them in their station efforts working with this data is being a bit problematic because of the tones of data is like volumes of data with AI what we've done with NASA is we've created an a representation what we call a foundation model for this data. And we have also created the open source tools for everyone to work with this AI representation as a result we see a massive uptake. And by the open source community working on this type of data and creating their own applications for things like detecting legal dumping places or understanding the effect of planting man groups in terms of alleviating the problem of the formation of. Yeah, he died once in cities or I can go forever. Well, I think you've got one of the most interesting job titles of anyone I've met this year and in terms of the things you're looking at from space but also quantum. Oh, yes. Tell us a little bit about IBM's role in quantum and where we are in the development of the technology it's something that we always says tech journalist feels like it's always five years away. Yeah, no, it's not fusion. Yeah, so we started looking into quantum like 40 years ago, right, it's not something that happened, you know, from one day to another because it's complex. It's very, very complex. But we are very lucky at IBM to have a very strong legacy in terms of understanding hardware building systems doing great science. And if you want to make quantum computing a reality, you need all the elements, you cannot just be great at science but not have a very powerful engineering approach or semiconductor approach. We had all the package and they kept advance this idea of computing with atoms that's what that with quantum is right and we've made some to me a defining move. For example, in 2016, we had the first with a few qubit quantum system, but we opened that by cloud to the community. We started getting people, the scientists, the physicists to play with the first quantum system. And the role of the committee has been vital for us to innovate and to drive our quantum mission forward and further. That's one of the defining moments that we had. Another one is like we are from research and research that they have a certain allergy to commit to a roadmap. You mean you're not going to give us a date for. But in the case of quantum with it. So in 2016, also we put together our quantum roadmap along the lines of improving the hardware, the technology, but also making sure that more and more we have software layer and application layer. Where it's very clear what you can do with quantum and we had the community to also contribute that so we are creating a massive ecosystem. And we are very proud that we've met each and every milestone on this roadmap. Listeners who don't know what the what the roadmap is, what's the next milestone and then follow up like when when is quantum going to be rolled out in the way that we're seeing AI. So we are in a very beautiful moment and I feel very lucky to be part of this moment as a computer scientist by training. So the moment we are in computing is very special why because we see all this AI taking the world by storm. But also we see a completely new paradigm in computing and merging which is quantum becoming a reality. Right. I remember when I was working in my previous job before joining IBM in 2020. Right. We engaged with IBM. I was part of a young, I was the chief data officer for the energy company and we were exploring what quantum could do for energy. We were exploring is like where what are the top use cases, how could quantum disrupt the energy industry, right. But back at that time, we had a bit if question is like if quantum is going to ever come or become a reality, right. Fast forwarding to 2025. Right. We have a date already on the calendar is not if but when and we are making the right steps to hit in 2029, the first for a full tolerance quantum computer. I don't know if a full tolerance rings a bell but the problem with quantum is being, it's been very, very difficult to, you know, you can imagine we are computing in atoms to make sure that these atoms, they always compute in the same way. So there are no interferences and so on and every vibration, every temperature change could impact that. It's been very, very difficult to have something that works always in the same way, but we've introduced error correction means that by 2029, we can for the first time trust that the quantum is going to execute as we program it. But before going to 2029 for 2026, we have a big ambition, which is we want to bring up the first examples of applications where we prove quantum advantage. It's a term right. It's been there for a while. The definition is quite simple is like by combining hybrid as a by combining quantum and classic computing, you can solve a problem faster, cheaper or better than just with classic computing right and you can prove it. And it cannot be more exciting as you can imagine. So two dates for you need to come back on the Times Tech podcast on in 2026 and 2029 when these milestones are reached. Thank you so much for talking to me. Thank you. Well, that feels as good a place as any to end with a look to the future. My thanks to Leon Butler, the chief executive of IBM in the UK and Ireland and Dr Juan Bernabé Marano, the director of IBM research in Europe. This bonus episode of the Times Tech podcast was brought to you in partnership with IBM. [BLANK_AUDIO]