267. Can AI save the public sector?

0:01That narrative somewhat holds us back. There is a general risk aversion in the public sector and some of that makes sense because there are serious human risks associated with deploying AI in these contexts. But you're saying that nonetheless, despite the fact that there hasn't been a problem of that, so you understand the risk aversion? Probably one of the most consequential technology decisions that the MOJ has made in years if not decades, tested our technology, then told us they're building their own technology. The sort of the safety level is not just about, you know, which is better, which is safer.

0:34You have to hit this tipping point, I think, where people are so very, very reassured before they hit the accelerator pedal on it.

0:53Hello and welcome to The Rest Is Money with me, Robert Paxton. And with me, Steph McGovern. Now, today we are talking about AI being used in council services. We've got the CEO of Beam on Alex Stephanie, and he's really interesting because he's developed this business, which the tech of it is used by two thirds of councils to help them be more productive. So basically what this beam technology does is it documents meetings and then it kind of powers advice and insights for those working in areas like working with helping people find jobs or helping people who are homeless find somewhere to live and those are the types of people who are dealing with lots of information being thrown at them and not much time to then process it all and write the correct documentation needed to get them on the right track so it's basically doing all the admin work that often holds people back and it's a business that he's seen really grow and as I say two-thirds of councils are using it.

1:55But what's interesting about talking to him, Robert, is the carry on is what I want to call it when it comes to trying to get central government to take on this innovative technology. That's exactly right. And, you know, we've talked so often on this podcast and ministers talk so often in public about how the productivity of the public sector could be massively improved by adopting new technology, especially AI. So talking to somebody at the coalface about how easy it is to sell AI to the government is definitely a conversation worth having. By the way, we also point out that, you know, when you are implementing a new technology, particularly artificial intelligence, there are risks.

2:38And we have seen actually around the world over the last few years, you know, implementation, you know, for example, you know, in the States, facial recognition, It led to the wrong people being arrested in the Netherlands. The rollout of a particular set of algorithms, which was assessing whether or not people deserve to get certain kinds of benefits. And actually, a lot of people lost money they should have been given. So, you know, there are risks when it comes to rolling out AI, but there are also significant benefits. And so, you know, this interview is all about trying to assess whether the government is managing what will be an inevitable re-engineering of the public sector, whether they are managing that in an optimal way.

3:28So here's Robert's interview with Alex Stephanie, CEO of Beam. Our story is pretty unique. So we started out actually delivering frontline services ourselves with teams of caseworkers. We had councils all over the UK. ChatGPT was out and we realized, look, why don't we just make a technology that supports our teams to do all their documentation? That was what the infrastructure was built on in the first version. So they could record the meeting, press a button and hey, presto, they've got a really accurate case note. By this time, the frontline team had manually written, I think, about 600 ,000 case notes, right?

4:04Typing them out letter by letter. So we're very pleased to have this technology. and within a few weeks we could see that uh they were happier yeah our friends and colleagues were cheerful to not have to do this type of manual work that none of them really um were signed you know had signed up for um and they were also able to see far more people um with all of the time that that liberated for them um and so we realized that this is enormously applicable to lots of other frontline teams. So we began supplying councils. Fast forward to today, two thirds of councils use our tech for their social care teams.

4:42And it's used for about 130 ,000 meetings every week. Essentially, it's a transcription service. So it creates very specific types of documentation. So we've produced around 2000 types of social services or frontline services documentation. but it does a whole bunch of other things as well. So it powers advice and insights using what we call a knowledge base of guidance or legislation that's relevant to that specific organization. We're also providing AI voice agents. So that means that say it's the middle of the night and that lady that we mentioned before has become homeless, she can phone a number at three in the morning and she can get advice there and then.

5:28then what the technology does is it will assign the risk to that and then when someone gets to work the next day they can see okay there's been 18 conversations overnight here are the two top priority ones here's a summary of what the situation is and they can get on the phone as soon as possible and move them forward

5:52we're delighted to say that this year the rest is money is being powered by octopus energy and that means we get to welcome back the founder of Octopus Energy, Greg Jackson, now CEO of it. Time for a quick fire question. What's the one bit of advice you give a young entrepreneur? Maybe it's stuff that people don't normally talk about. Don't be arrogant, but do be confident. Almost by definition, as an entrepreneur, you're doing stuff that other people didn't think is going to work or that they've discounted or dismissed. so you have to be confident to believe in your own ideas and your ability to deliver them but you should still listen to everyone else and learn from them and not be arrogant i think that is the secret to being able to build something special but successful well to you is greg and thank you for powering this episode of the rest is money

6:47Nobody wants to get catfished by dating profiles or big wireless carriers with hidden fees. Your one true match, Visible Wireless. It's one-line wireless with unlimited data and hotspot for just$25 a month, taxes and fees included. Now that's a green flag. And it's fully digital, so you can switch as fast as you swipe. Tap the banner to learn more. Terms apply. See visible.com for planned features and network management details. Particularly with sensitive notes of this sort, there are always going to be anxieties that it won't be perfect, that there will be mistakes. How do you minimize those?

7:26Well, a few aspects to that. So the first thing is that this is not replacing frontline workers. So this is about augmenting frontline workers. So instead of having to create documentation from a blank piece of paper, they're able to get something out of the box within about three seconds of the meeting ending. That's nine out of 10. And then they spend their time reviewing and improving that documentation where they see any potential for improvements. So you don't have any products of that sort that actually replace, you know, you're not in the business yet of creating agents of an AI sort that replace people.

8:05No, we're about augmenting very overstretched teams. So you can be almost anywhere in the world and you can be speaking to people in frontline services and realise that they are struggling to hire people. They're struggling to retain people. And that's because these jobs are often very difficult. It's really, really hard work. It's really stressful. They're often not tremendously well paid. And as discussed, they're often characterised by very high levels of administrative work. I think you said you're working with someone like 70 % of local authorities. Do you yet have any contracts with central government?

8:41Not yet. So in terms of central government, what happened was some actual probation officers got in touch and said, this looks really great. We're smothered in paperwork and our jobs are really, really tough. Can we use it? So we did a pilot with the Ministry of Justice. Like most of these types of pilots, they're very loss-making for the companies. So we knew that it would cost us well over£100 ,000, maybe£200 ,000 to actually do this piece of work. And you do this at your own risk? We do it at our own risk. And we also offer to provide the technology at cost as well. Why are you offering it at cost?

9:24We'd spent years trying to work with central government and we thought we really don't want cost to be any kind of barrier. So probation piloted this tech. There was a glowing assessment of it, which found that it was leading to a 70 % reduction in time spent on admin. 100 % of people found improvements in their documentation. While the pilot was ongoing, MOJ requested lots of pretty commercially sensitive documents about how our technology is structured and how we adopt and train people on it. They then told us that they're going to be building their own version of it and also going to be heavily involved in evaluating which one they preferred.

10:14and we were kind of worried that it wouldn't be a completely level playing field, to be honest. It felt a bit like they were marking their own homework and that maybe it wouldn't be fully objective. So I do remember when I was talking to David Lammy about it, he personally very energised by the prospects of adopting AI to improve the efficiency of the justice system. In fact, he's now talking about it in terms of his court reforms, as you probably know. And I also remember when talking to his officials at the time that they absolutely insisted there was a sort of level playing field between you and their internal team.

10:49I mean, I guess my point is just a more general point, which is this issue around, you know, there are definitely strong public policy reasons why you need to encourage, I mean, I'm not making a case for you, but just in general, you know, it really matters that procurement is a tool of industrial policy in this country, and that means using procurement to actually create a more vibrant, growing tech sector. Well, I guess on that point of procurement, it's probably worth saying that there wasn't a procurement for the technology for probation officers. It's probably one of the most consequential technology decisions that the MOJ has made in years, if not decades, and they didn't go to procurement.

11:37I think it would have been a good idea if they had. So what was it then? If it wasn't procurement, what the hell was it? They tested our technology. They then told us they were building their own technology. And then they told us that they wanted to use their technology, which still hasn't been rolled out widely. And we've tried to understand some fundamental things about how they chose their technology. Obviously, we spent the best part of a year working on it and it lost a lot of money doing so. But ultimately, they don't answer the questions. We did have a meeting with them, which they cancelled at the last minute.

12:14So it's difficult to say how fair it was. One hopes it was, of course. Alex, tons more to talk to you about. We're going to go to a quick break. But after that, we will really drill down into why there is so much resistance within the government to backing homegrown British tech talent. Today, we helped a latte for Sam coffee shop get an insurance quote simply and easily and made sure a floral delivery van was able to make someone's day. We're the Hartford, with decades of experience insuring millions of unique small businesses. When it comes to your small business insurance, Thank you. one size absolutely does not fit all.

13:04Get a quote or find an agent today at thehartford.com slash smallbusiness. But it seems to me there is another issue, which is one we talked about quite a lot on the podcast, which is, you know, we do have in this country a really very lively technological sector, mostly sort of small, medium, growing companies. And if you look at, you know, essentially faster growing economies across the world, one of their characteristics is that you have public sectors that are imaginative and creative in favoring successful local growing businesses. Actually, for years, the British government has talked about trying to procure from businesses like yours.

13:54But then in practice, it doesn't seem to happen. So why is it in the end, do you think? I mean, you've done very well with local government. But why do you think central government routinely favors big tech, routinely favors the likes of Microsoft, for example? but a company like yours struggles to get a foot in the door. Very happy to speak to that. But maybe let's just take a step back and ask the question, why does this matter? Which I think is not obvious always. And we are what you would call a govtech company. Now, I'm sure your listeners have heard of fintech. We're govtech. So what does that mean?

14:38It means that we supply largely government customers. We build technology for government customers. Now, I think you want to have a vibrant GovTech sector for at least three reasons. The first is that the more companies you have building technologies, ultimately, the more competition and the greater likelihood that people who are delivering these crucial public services are going to have good technology. That's going to be driven by competition and everyone benefits when people have good technologies. Ultimately, we're all paying taxes for that. The second reason is growth. Now, GovTech is really interesting because what we're doing is fundamentally creating economic growth because we're a group of businesses, but we're also allowing the government to be more efficient, to find new productivity.

15:29So you can think about GovTech as a kind of double whammy for growth. And that's why, you know, we need to really encourage the sector because growth is what's going to pay for our schools and hospitals and so on. And then the third reason is I think it's important that we have internal capabilities when it comes to these fundamental technologies and services. If we are, as you say, just relying on foreign built technologies, that doesn't put us in a great position in the long run. And I think you're right that it can feel as if UK technologies are not first choice. In fact, maybe they're not even second choice.

16:13But is it just because the officials who are in charge normally of the procurement process are just risk averse that, you know, you know, broadly, it's quite hard. You know, if you're buying from Accenture or if you're buying from Microsoft, everybody knows the brand. It's sort of tried and tested. the risk of you being criticised for it going wrong, even if it does go wrong, is sort of minimal because then they'll say, hang on a second, how was I to know that this particular contract was going to go wrong given the reputation of the company? Is it just that they are just terrified of a company like yours, which they've never heard of before, failing and therefore they're not going to take the risk?

16:54On the one hand, you have politicians who authentically want UK AI champions. I have no reason to doubt that is the case for the Prime Minister and many other people and many, many people in the civil service. So why isn't he able effectively to push that down the chain into actual contracts? You're right that the reality of companies like us is sometimes that first choice is US tech. Second choice is let's build it ourselves. Third choice is, OK, if we must, let's work with a UK startup. Now, why does that happen? I don't think it's a universal thing across the civil service, but I think that there are some pockets of the civil service that are just not on board with that.

17:38Is it that they resent the fact you're probably going to make a fair bit of money out of it? Is there an anti-commercial thing going on? It's difficult for me to know. If they'd chosen you, you could have rolled that out to their 9 ,000 probation officers, the ones who haven't got it at the moment. They'd have it now, would they? Absolutely. Yes, they would have had it in weeks. As it is, millions of pounds of taxpayer money has been spent to replicate technology that already exists. But the main thing is that every single day that passes, millions of pounds of taxpayer money is also squandered because people don't have modern technologies.

18:12So, you know, you will have to ask them why they made that decision. I think we and every other company has absolutely no right to any government contract. I want to be clear about that. But what we all do need as a country is to make sure there's a level playing field so that the right decision is objectively and fairly made. And by the way, I think it's also important to note that some of the time it's exactly the right thing for the public sector to build their own technology. It's certainly not always the case that they should be relying on the private sector. In terms of what your technology can do, anybody who's used to go to a hospital recently or go and see a GP or indeed had a deal with the police, We still see in those organizations astonishingly clunky processes for making notes, taking witness statements or writing down the symptoms of a patient.

19:20Is your technologies, in a sense, flexible enough and accurate enough to be usable for law and order or health? Is that something you're looking at? Yes. So it's something we already do. So really our speciality is frontline services. So that covers all these types of things. And we have hired people who are probation officers, work coaches, who have left those professions because they were kind of didn't enjoy the reality of that work doing admin. So definitely, it's all about how you take this core technology and then how you tailor it to very specific workflows. So we have just had a team of people flying around the US, working with government bodies there in California, in Colorado, to customize our technology to specific workflows.

20:14And one of the things you see when you do this enough is that kind of people are people to a degree. So the services you would provide to an elderly, isolated person in London are very similar to what you would provide in Paris or Boston or wherever. But there are all kinds of endless nuances in terms of rules, guidance, guidelines, etc. And so you have to heavily customize the technology. And so we use what's sometimes called a forward deployed engineering model where people will actually go inside of these organizations and really co-design the software with people doing the actual work. And as you will know, with things like Claude Code and Cursor, it's now becoming so much quicker and easier to actually build technology that is highly customized to partners.

21:07Just wondering if you've had any conversations here, you know, with the police service or with the NHS. We work with various different NHS trusts. And we're also speaking to different police forces in the UK and Australia as well. But yeah, you're absolutely right, Robert, that this technology is very malleable to different frontline services. Employment services is another big area. So working with employment services, those are often private companies that the government outsource. to support their frontline teams. And do you have any data on how much more efficient an individual becomes, what the productivity gains are?

21:46So it would typically be around eight hours a week. So you save, so a typical person using your technology to record notes and create reports, they'd be saving about eight hours of admin per week. Yes, exactly. And it's that stuff, but it's also general workflow. So if they need to make a referral, write emails, anything to do with their sort of organizing their time, getting advice, which they usually would have to do from a manager, they can now get support from the technology. They can ask it a question and it will tell them what is the policy around this within our organization. So it's a whole load of, you know, it's like a chatbot type thing where they can talk to it and they can say, I need an email for this.

22:34I need some advice on that. And the technology will give it to them. You know, just in crude numbers, this is like a 20 % productivity gain, freeing up of time for what we all want more of, which is, you know, a public servant actually spending time with the patient or I hate the word customer in the context of the public sector. but basically spending time with people who actually need their services. I mean, that is such an enormous saving in the context of a government which is really struggling to fund public services at the moment. Are you surprised? This is not supposed to be an advert for you, But are you just surprised by how long it is taking to roll out these sorts of services throughout the public sector?

23:26It's been very swiftly adopted by councils. And if you look around the world, are there public sectors which are well ahead of the UK? People always talk about Estonia, for example. I mean, are there public sectors around the world which are way ahead of us in terms of digitising and improving productivity through the use of technology? I mean, I think overall the UK is doing reasonably well. I don't think there are many examples of governments that have raced ahead. I think that there is a general risk aversion in the public sector. And some of that makes sense because there are serious human risks associated with deploying AI in these contexts.

24:09So, you know, you don't want to necessarily be the first to roll out these types of technologies. But you're saying that nonetheless, despite the fact that there hasn't been a problem of that sort, you understand the risk aversion. Yeah, I mean, I think people are just nervous around risks that are less familiar, right? You know, and you see this in self-driving cars. So ultimately, self-driving cars are profoundly safer. Like we know this, we have enough data, but people would still be more worried by getting killed by a self-driving car or, you know, one death would be disproportionately covered in the media.

24:47And so the sort of the safety level, it's not just about, you know, which is better, which is safer. You have to hit this tipping point, I think, where people are so very, very reassured before they hit the accelerator pedal on it. But I think that the UK can go faster, can be a global leader in the deployment of AI in the public sector. I have not given up hope, partly because of the success we've had with local government. And, you know, our headquarters is still in London and we're still regularly talking to government and, you know, hopeful that we'll roll out DWP and maybe we'll get an opportunity to provide our technology to the thousands of probation officers that still don't have any.

25:30So we will have to see. I mean, particularly within frontline services and social services, you know, we have a very proud legacy in the UK. If you go back to the Beverage Report and after the Second World War, we invented concepts and we rolled out programs that made an enormous difference for millions of people in the UK. And by being copied around the world, for probably hundreds of millions of people around the world as well. If there was just one decision that a prime minister and chancellor could take right now that would make all the difference to, you know, I'm not sure I love the expression govtech if I'm honest with you, but we know what you mean.

26:09To really put Britain at the forefront of govtech, what would it be? I think there's lots of stories of investments. You know, we've attracted£5 billion for this,£10 billion for that and so on. I think people are a little bit numb to these numbers that come at them all the time. I think what people need are stories and they need to see actual services that they use improving. The story that the prime minister, by all accounts, keeps coming back to is the story of how our technology is being used by social workers up and down the country. You mean your beams technology? Beams technology, yeah. From the Orkneys to the Isle of Wight, it's being used all over the UK.

26:53And that is a good story because it's about how services are getting better for vulnerable people. And we need more stories. So I think what I would say to the government is things have been going somewhat slower than you probably wanted. it's probably time to roll up your sleeve and to try and take a look at why that has been happening. And to take probably a more active role in the procurement of these technologies and make sure that procurement is fair and fast and that the best possible technologies at the best possible prices are getting into the hands of people delivering public services.

27:33You know, one other thing I would say to that is that there is this narrative which is kind of procurement is broken and everyone knows it and everyone's trying to fix it. It's just taking time because it's so complicated and so on. I think that that narrative somewhat holds us back because the reality is that procurement may be broken for UK startups, but procurement is working beautifully for lots of incumbent interests. And so there are plenty of stakeholders that actually don't have the the incentives to change a thing about the way that we buy technologies for the public sector. And I think we need to actually lean into that, accept that, and work out how we change that if we want to be living in a country where we're getting the best technologies at the best prices to people who need them.

28:24Really good to see you. That was fascinating. Thank you so much, Robert. So that was my conversation with Alex Stephanie. Annoyingly, Steph was doing something probably more exciting, although I did really enjoy that conversation. But nonetheless, you weren't there. You have, though, now listen to it. What did you make of it? What I found fascinating was what Alex has done with the business is clearly found a solution to a big problem in terms of productivity for people doing really important things. You know, we're talking about council workers who are trying to find people homes, trying to find people jobs.