Shared planning and AI Meeting Recording_8 Oct 2025

five, six years ago where we um some of us sat down and discussed um opportunities well the challenges in the planning service and opportunities and it's just been astonishing the um those ideas have um um come a long way and I think we're we're now an exemplar service within this council and the shared count uh joint planning joint councils but also across the country. But one thing I'm very aware of that it is the accelerating use of AI and um how it could I just feel that it could almost become out of control and we very much want to stay in control of um how we use this as a tool so it doesn't swamp us. So, and one thing I'm aware of with conversations with Bill is about the necessity to understand where we want to go to. So forming um an AI strategy that will that fits with where we are now and where we want to go to get to and um uh to ensure that you know we we it's but also um without a doubt it's a tool for all of us to use um dayto-day task to task. So it's not just so it's I'm sort of aware that we could just pick up uh what's readily available and just use it without really understanding how to make the best use of it with the tasks that we've got. And I know Bill's been talking about uh these issues for many years. And I just think some of his thoughts and philosophies might help us maybe pin down a strategy and make sure we're using these tools on a day-to-day basis as well as the bigger uh platforms and that we've developed. >> Yeah. Thanks. Is it worth saying who I am in relation to you Katie so understand exactly why >> so Bill is my husb Bill Bill is my husband and we have many many conversations about a >> and I I then have these conversations with Heather and um the kind of digital team um so it's a great relief for me that uh that you can uh listen to Bill now and ask him questions. Thank you. Thanks. >> Might be worth putting in context. And if I shout loud enough, you'll hear me on Katie's microphone. >> You're in the same room. >> Separate separate rooms. We've learned this over years of the COVID lockdown. Don't ever try in the same room. It never works. >> Um >> um so Bill, just you know, Katie very um kindly introduced then. So I suppose um just for Jane's maybe um perspective as well, we we do have a very small digital team in in planning now that sits under Charlene who's on the call. Um and that was very much because we recognize there's a lot of you know initiatives out there that are happening and um we were very fortunate that we're successful in a number of funding bids. Um Toby is is um sort of in charge of one of those or um part of that one of those um funding bids. Um and and this is around how we can help to speed up the processes and you utilize um AI effectively. Um so we're kind of on that journey, but it would be really great to hear from you and understand, you know, um some of the sort of great work. I I I um I know that you were working on the BBC AI strategy. So, it' be fantastic to hear from you and then um like Katie said, open it up to questions as well. >> I'm I'm very happy to do that and really pleased to have an opportunity to to meet some of the people I've heard about because I have heard about the great work you're doing and did read about it in the papers. Um perhaps I'll start by sort of just giving a bit of background to myself. So, um in my career I've sort of had two strands to my career. Um I'm a technology journalist. I've written for the papers, done radio for many years and report on what's happening in the world of everything sort of computers, internet, AI, quantum, whatever you've got. I've been doing it for several decades now. And then I've also had worked inside the IT industry. I worked for local software house um here in Cambridge. I worked for AON computers back in the day when the ARM chip was being developed. Um so I remember the armchip being launched and um went to the Guardian newspaper where my sort of two strands of my life came together and for the past decade or so I've been at the BBC research and development. I sit in the research and development team at the BBC and our our mission is to help UBC sort of prepare for and indeed shape the future uh when it comes to technologies so that we can continue to deliver our public service mission. Um, I also still write, broadcast, blog and and hang out. And several years ago, nine or 10 years ago now, it was clear that um what was then called what was called AI machine learning was becoming important and started a program to have internal conversations and build an internal group or the AI machine AI ML community within the BBC really to make sure that we were as an organization thinking about these issues. And then three years ago when generative AI exploded, we were sort of ready and in a position to have some influence over the BBC's approach to it and it strategy. And I'm still very involved with that. And now as the BBC looks to the renewal of its royal charter in 2028, we're starting to think about, well, how do we think about getting the language about AI into the charter? And I've been going around saying the term AI must not appear in the BBC charter because it has to run to 2038 and we won't be talking about AI by then. So it' be very dangerous to to tipping our hopes on it. Um so my background is as a computer scientist and and also as a reporter and I think that's given me a helpful perspective both in appreciating the hype um understanding what's really going on and certainly within the BBC helping us think about the possibilities and and within my team in R&D we've helped with many of the AI trials we've done but then I've also watched with interest what's been happening in the rest of the world and in particular thanks to to to Katie's position, what's been happening with with planning and obviously the work you've done with with plan AI. Um the should we say slightly um restless coverage of government announcements about how massive percentages of your time is going to be liberated by tools like extract and other things like that. um you know um and the perhaps complex framing of the the use of AI within local government and central government and elsewhere. Um one of my projects at the BBC has been establishing our what's called our technology hub up in uh Newcastle in the northeast and our building is a mere seven miles away from a building called Cobalt Park which is going to be home to the first AI growth zone. that we're also now getting directly involved in the ways that both economic policy and if you like planning on the ground is being shifted and some of you may have been um seen a paper yesterday that came from the Center for Progressive Britain about AI growth zones and how all planning regulations need to be torn up to make way for these vital parts of infrastructure. And in that 40page report, the word water appears six times and each time it's coupled with energy and it just says energy and water need to be sorted. There is no attempt in it to actually address what we know are very key infrastructure issues. So I try to put forward a a realistic perspective on what the technologies are capable of doing to think ahead in terms of what we want to achieve from them but also to be very grounded in in the reality of the fact that it's not fairy dust it's not magic in the end it is transistors and silicon and power and buildings and data unless we recognize that you know it's not the cloud it's just somebody else's computer. Unless you recognize the cost of infrastructure needed to deliver and deploy these technologies, then we could find ourselves in a difficult situation where it becomes impossible to do the things we want to without enormous cost to people's lives or the environment. So, I'm hoping I'm balanced. I'm not a planning expert. Many of you are. So, I will always defer to your better understanding of what you're trying to achieve within the service. Um, but I'm very happy to talk about some of the work we've been doing in the BBC and also perhaps to reflect on if if you're interested in what government seems to be doing and what they might do. Um, I've worked for the BBC, but I'm not speaking for the BBC today. I won't say anything that's not in the public domain, but it's not really intended to be shared on social media or spread beyond this group. Useful context. I'm happy to take the conversation in whatever direction you find helpful, Heather or indeed Toby or Charlene if you want to lead off as well. >> I was really interested when you said that um using the language of AI using that term AI because by I think was it 2038 you said or I can't can't >> Yeah. Can you explain that a bit more? Cuz that's really I mean I think I do think you're right. I think I cuz I'm not I'm not an expert in this field. I think we use AI um digital digitization you know we've all got these terms that are just kind of floating around which all mean something different to individuals. So it' be good to to understand that I think um that would be helpful. >> I'm very happy to to talk a bit about that that broader context. So personally I I really don't like the term artificial intelligence or AI. Uh I'll settle for machine learning. As as somebody pointed out to me, it's only artificial intelligence if it comes from Silicon Valley. Otherwise, it's just sparkling statistics. Um as in it's it's a term that has become so broad it's effectively meaningless. It it was coined in 1956 as a marketing term for um a a summer school at Dartmouth College in the States. And professor Stuart Russell has said calling it artificial intelligence was a massive mistake because it set things up in a particular direction philosophically and it's been impossible to get away from these ideas about these machines have become intelligent. artificial general intelligence. The supercomputers are going to take over the world and that distracts from their actually utility as a different way of making computers work and solve a spec a sets of problems in different ways that might be useful to us as human beings. And he prefers the term computational rationality. Can you write that one down? Computation rationality CR. If that had caught on in 1960, we we wouldn't have been in the same place we are here today. say um it feels to me that what you do when you call it AI is you you flatten everything into people who don't really know much about the technology a sense of this this magic that just solves problems don't think too much about it and you then avoid getting in into the specific details of the different types of AI for example I said earlier generative AI is very different from the rulesbased systems that predated it and stuff like that. Things that we in BBC have been working with for 20 or 30 years now. I did a did a undergraduate dissertation about machine vision systems in 1980. You know, this stuff is not new, but some elements of it are new. And if you call it all AI, you lose that sense of it. There's also the sense of that. Okay, the parallel is the word content. You must have heard the word content. Content on social media. I'm a content creator. Whatever. Content flattens everything. Nobody went into the cyine chapel and said, "Wow, Michelangelo made some great content, didn't he?" Okay. And within the BBC, we use the word content. And it it almost diminishes the quality of what we make. It's a television program. It's a documentary. It's radio. It's the specificity becomes important in understanding the characteristics of the thing you're making, the creative output. And I think specificity when it comes to the computer systems you are building is also quite important. Help people understand the range of capabilities perhaps the limitations the applications it's suitable for how it might work the data needs and things like that. So there's a very big different difference between all these systems. And the generic term AI appeals to me could have a shortish life because over the next five or six years all of these capabilities are just going to get built into the tools we use and we're not going to think of them as AI and talking about adding AI into something just won't won't make sense. It won't it won't feel coherent. Um any of you who have a smartphone I like that's all of you. There's so much like mo sort of neural networkbased capability. Don't want to call it AI neural network based capability embedded in your phone at the moment. Um my Apple emails, you know, there's the little summaries that appear now um before you open the inbox. They're all being generated by, you know, a machine learning system that sits on the phone. It's running on the phone, not even in the cloud. So the technology is becoming so pervasive that having this term for it, AI, doesn't help. And in many cases when it comes to trying to persuade people about the systems you want to develop, saying AI only does not help clarify what you're trying to do and puts the people you're talking to in the same position as a, you know, as a desperate government press officer who's trying to shine up some new announcement by saying with added AI, but it doesn't actually tell you anything useful. So in the context of the BBC charter, we're trying to take out all of the technology specific stuff so that we don't get constrained to carry on using things when they're past their sell by date and also so that we are forced to be clear about the ways in which we want to use technology to help BBC deliver its public purposes but have the freedom to choose which technologies. Um, I remember saying to somebody a while ago, it's one of these predictions that didn't turn out to be true. But if you have the word digital in your job title, you should be worried. Turns out the word digital has actually carried on a good five years older than I think it should. Um, because everything's digital. I did actually have one success with the director general of the BBC. We had a um a revised mission was going to be the world's we were going to be the world's first global digital public service media organization. sounds good. And myself and some colleagues persuaded him to drop the word digital because the last time anybody in the BBC actually edited a piece of analog tape, you know, on a tape recorder was 30 years ago. The BBC has been digital for a very long time. It the word is now abused and I think that AI as a term is being abused as well. And I would say to all of you, you find a way to describe the thing you're doing so that you don't have to use that term and you will find it has clarified your thinking enormously. It's worth the effort. think how would I describe this if I wasn't allowed to say AI >> as a as a as a so you going to say something too >> I was just going to say kind of my observation about a lot of the improvements that we've made over the over the last kind of couple of years really have been around kind of management of what we hold kind of digitally and putting the kind of tools in place to extract that information and kind of use it for whatever purposes you performance monitoring for example in planning but we're now kind of moving into this bit more uncomfortable space where you know there is the ability for example to use a large language model that Liverpool University are developing for us to summarize things and using kind of machine learning and that that is both exciting but also so potentially kind of terrifying that and and you know and and and also I guess from some of the concerns that I've got are just around you know the public perception of public of public officers working for the council relying on machines to summarize um sometimes very personal representations that they're making on um planning applications. I don't really know what question I'm asking here, Bill, really, but I'm just kind of >> thinking. Yeah, >> it's a it's a really good question. So you you and it is that element of okay somehow the the there is a almost an aura around large language models largely thanks to GPT's marketing and the way tools like anthropics claude represent themselves as being somehow empathetic real intelligences instead of what they are which is statistical machines and that people feel somehow that there is a level of engagement with the machine that comes if you use an LLM that you wouldn't get if you said you're using Excel to analyze >> and so you're having to cope with that when you want to what you want to deploy is a technology that you understand very well and you understand its limitations very well and you've also designed a system which has suitable safeguards in it to make you confident that it will achieve achieve the outcomes you want. And it turns out actually we're really good at designing systems that include elements which are imperfect, infallible, and don't do what they want. They're called organizations with people in >> every person listening to this call is just as flaky as an LLM. And yet we are an efficient machine for working and delivering. So, you know, we know how to do this. um we just somehow don't apply that sort of critical thinking to when we're using LLMs with their degree of uncertainty. And I do think a lot of it is about you know the desire of people like Sam Alman to become very very rich and powerful has led them to create a situation within which there is deliberate confusion about these tools and these systems and you sadly are at the sharp end of coping with that. So the the answer the only answers are on BBC. So I'm going to say it's obviously public education helping people understand better. We happen to have a public service broadcaster that will help you with that. So please increase the license fee contractual obligation fulfilled but but more seriously that that that degree of understanding but also clarity and transparency from you about what you're doing. And I think that publishing what you're up to, publish, yeah, as much openness about the systems you're using and how they fit together is important because the the the phrase I've I've been using is you want to be the the the human in the loop, not the hamster in the wheel. So you you want to be the person who is controlling the deployment of these technologies and that means using all of the systems thinking that you're good at already. All of the ways you've designed computer systems in the past all the ways you've designed processes you know I know about the transformation of shared planning over the last 5 years by becoming more efficient and more effective by organizing yourselves in the appropriate way. All of those techniques applied as you deploy these technologies, but these technologies come with an element of of public concern that hadn't existed before that you're having to deal with and an element of sort of, as I say, deliberate confusion that makes it harder. telling people what you're up to, acknowledging where you know you're going to have to intervene to keep it keep the machines working correctly. All those things will help. And then perhaps there's the other sort of more political thing which is to make sure that people like I don't know you the cabinet member for planning don't go around making unwarranted claims in public in press releases and things like that because they think it makes them look good and might help them get reelected. you know, you need to control what the politicians are saying around the capabilities of your systems to make sure that the public have the right expectations. And that applies all the way up to central government. You know, one of the issues I've had with people like Peter Kyle >> is not that he's a bad politician, but he's saying things which are quite hard to justify in terms of what the technology is capable of and then all the way down the system, people having to scramble to make what the minister says come true. Um, a significant part of my world, this is not for repeating, is trying to make sure the director general of the BBC never says anything actually stupid about technology. Okay? That actually you control what is said at a high level so that you don't you don't end up having to deliver promises that >> end up distorting what you're doing. >> There's a politics to this as well, I'm afraid. >> Yeah. >> Yeah. Sorry, Toby. Go on. >> That's okay. I I was going to um ask a kind of separate question actually, Heather. So if you want to >> Yeah. So I was just Yeah. No, what's really interesting um Bill there is I think is from what what you were saying and what you were saying earlier as well is that obviously our our um powers that be in government are are making some promises around the fact that we've got um let's call it AI digital, you know, I'm going to use those terms in this space. you know, it's necessary alternative yet. >> No. So, this kind of the promises around that seems to be, oh, you know, everything's going to be miraculously quicker, faster, cheaper, more effective, etc. Whereas from what we're saying is the reality is that that isn't necessarily going to be the case because there is going to be tradeoffs here and there are going to be that human intervention. It's it's not something that the that we we know that the point of us doing this is to make us more effective in what we're doing rather than uh you know, >> let's be honest, cut a load of people out just because, you know, we we we're going to do things in a lot, you know, a lot more a lot faster than we could before. That isn't the case. It is just in a different way. And one of the pieces of work that we're doing is looking at to looking and trying to understand what skill set do we need as a as a kind of um planning service to if you like if you just take us as planning to to be able to navigate through that and achieve what what we need to achieve. So I suppose I've kind of trailed off into something else now. But I thought it was interesting when you said we totally support that viewpoint that you're not just going to save a lot of, you know, money by just be just because of doing this, you know, doing this with with um um automation, digit digitization, machine learning, all the rest of it. It's just that all it does is make you more effective potentially if it's used in the right way. I think that's right. I'll come back to you, Toby, but there are there's a couple things embedded in what you said there which I think are really interesting. Um that point about it's not about saving money, it's about being more effective. What I found certainly is it's something you need to put into every document. You must never allow a space within which it can slip to being about costsaving because once you do that, they'll come looking for those costs savings. And some of my work has been involved is being basically, you know, at the point where somebody senior says, "So, how much are we going to save?" Being able to push back and say that's not the point. Yeah. It might end up reducing our cost, but we're not doing it for that reason. So, we can't properly anticipate that until we've got to this level of the trial until we've rolled out, but don't hold me to a number. Hold me to the numbers about your KPIs about effectiveness, about delivery times, and things like that. I will do that and try not to let it get built into the set of assumptions and that's a real challenge at your level Heather right that's that's where you're the one who gets to see and understand it and with support from your staff you can do that but try not to make that the conversation is is been what I found most effective you will lose some of the time but at least you can make it not the premise and that also I think helps reassure both your customers you're not doing this to be save and also the staff. Your point about skills is really important. The other side of that which we're really come up coming up against and I think a lot of other people are is around um career career progression particularly for junior staff. So there is a question about what are the p what's the pathway to being senior and effective and does it involve doing stuff work that is actually quite boring and dull and if that work is then done by a machine does the fact you haven't done it make it impossible for you to progress you know if you're a lawyer does the fact you haven't read through hundreds of cases and served briefs and worked for a barristister mean you can't actually progress because that that gives gives you the core skill and understanding that you need. If you're a software developer using copilot and other tool to write your code if you haven't written it from scratch, does that make it impossible? We do not actually have an answer to this question yet in many professions. And it's worrying and again for me it's the sort of thing which needs to be addressed face on to say to junior staff. Okay. They've all had you all had the issue around you didn't actually work face to face with people because COVID happened five years ago and now you're in a position where you may not be developing the core skills needed that have got you know Heather to where you are Toby to where you are because the machines can do some of those things and the question is therefore what do you need to have done? >> What do you actually h what is that new career path to progression? And I don't think we're giving enough thought to that at the moment. We are with our journalists. >> Yeah. So, fortunately, there are elements of reporting like standing out in the street box popping people that are unlikely to be done by a humanoid robot for a while. But, but it's like thinking what are the things they need to have done and how do we build that in? I actually have a sneaky feeling that the the growth of apprenticeships might give us a way to do that for a lot of professions because apprenticeships are really adaptable, can change quickly. Um, and so there there's something in there and again one of those things to think about how does somebody get to be me when they won't have had the experiences I've had over the last 20 years because the machines are doing quite a lot of it that um know if that's helpful. >> Yeah, thank thanks Bill. It's really interesting isn't it? I mean, I kind of, you know, I'm thinking back to you, you could apply this to anything though, couldn't you? In respect of the skill set and and what you need actually need to do. And I'm thinking, sorry, I'm going to go to my farming background here. And the way we do things now is not the way that was done in the past. It's very, very automated. You know, uh, when we feed our chickens, our 48,000 chickens, it's all done automated. It's all based on the size of the egg. It's all calculated. That's all done by a machine for us. We don't do any of that anymore. If they need, you know, their water adjusted, it's all done automatically. So, uh, it's incredible um that there's a skill set there that isn't required. So I think for me there's a difference between >> if I was to take that analogy what is actually required to enable us to become good planners is what I'm getting from you but what are those bits that we could leave >> easily for a machine learning or whatever which way and say well that bit you don't need to have a skill in that but in some areas you may need to develop your skill by experiencing it by actually doing it really really interesting food for thought there Toby I think you wanted a question I'll come back to Toby said actually because you just said something which >> um echoes so look at tool like extract which is going to digitize planning documents yeah how important is the skill of being able to look at two documents and understand they don't map to someone in their career maybe there needs to be time maybe there needs to be training to do that even though most of the time it's being done automatically so that you can bound check the machine so you understand the issues. Again, it's starting to design career progression and training courses around working with the systems rather than just letting them replace the skills in some areas. >> Okay, going on too long, Tony. >> Yeah, >> I don't know if this is the right analogy or not, but it does involve farming. Heather, there I was listening to Mark Steele last night. He was in Rutland and it was a Q&A session. Well, his his his show about Rutland and when the um reservoir was kind of formed, one of the farmers was saying, "Well, um the only thing I'm going to be able to farm is uh hamsters." Um because all of his land had been kind of consumed by water. And I was thinking when you were talking about kind of water earlier, a couple of my daughters actually said, "Did you know, daddy, that whenever you use AI to design your um beer bottles, you're use you're using a a bottle of water for every request." And it does kind of get me thinking actually kind of around the kind of environmental impacts of what we're doing. um if we're relying on machine learning or a large language model to process 7,000 planning applications a a year, you know, do we need to kind of rethink that strategy? Particularly given that we're kind of I know we're, you know, we're in an area of complete kind of water drought, but the I don't know where the the the machine is that we're going to to find the answer. If they're in if they're in Rutland extracting water from Rutland reservoir, then are we okay? I mean, I just I'm I'm a little bit lost around. >> It's a good concern. Um Jane, you got your hand up. Do you want is is it building on this point or should I deal with this and then come to you? >> Deal with that one and then come along. >> Okay. So, I grew up in Corby, so I actually remember that reservoir being flooded. I remember it um um arriving and you I've been I've been I've been at the bottom of that reservoir when it was a village. Um so it's quite quite a good analogy. Um which view of the future would you like? Okay. So there's there's there's the the the uh the doom laden scenario um which is that the entire world's energy supply is going to go into data centers um and all the world's water will go into data centers and it will bring about the apocalypse and the emergence of the antichrist and this is potentially a bad thing unless you believe in the second coming in which case it's a good thing eventually. Um, at the moment a lot of the statistics around energy and water use are out ofd and exaggerated, which doesn't mean it's not a real problem, but there's a danger that it means the problem gets dismissed and we're not being realistic enough in our assessment of it. So water doesn't get used, it's just that clean water becomes dirty water and has to be filtered again. Water doesn't go away. So the design of the systems that could be more efficient is is feasible and is starting to happen. And the one thing that gives me hope is in fact this because the computing power on this, the iPhone, this is not an iPhone 17, but the new one is perfectly capable of processing your several thousand planning applications a day. What we're starting to do is to offload a lot of the stuff into trained models onto the edge. If you think about the design of the systems, the Apple intelligence was the first tool to do that, which is basically if you give a query, it's like the phone says, if I can do it, I'll do it. If I can't do it, I'll offload it to Apple's data center. If Apple's data center can't do it, would you like it to go to chat GP to GPT4? So, they have this layered approach to it. And as the capabilities of the edge devices increase, and this is unless accidentally not water cooled. Um so um that the total compute capability of the planet is increasing so rapidly thanks to these powerful devices that again it's one of those things where with political will a bit of intelligence we can solve this problem >> that will and intelligence doesn't seem to be there at the moment so you know to say I was optimistic and if you look at the plans for people like Nvidia and open AI to roll out Stargate data centers and things like they seem totally unrealistic um in terms of the environmental impact. So you have to hope that they're being done to impress shareholders and increase share prices, but they don't really plan to do it. Be a vain hope. Again, there's there's a loop round to some political pressure and realism. And perhaps within the UK, as the government promotes the UK as a center for as as a base for these things, getting some intelligence. So you don't just do the Matt Clifford let's roll over the planning system. actually say let's have a planning system which balances these things because in five years time we're not going to need those data centers anyway in the same way because of the rate of development. >> I've contrasted a lot of things into that answer. My ultimate answer would be that it is not the it is not the most pressing issue when it comes to the environment at the moment. Okay, it is an issue and I certainly don't feel guilty about running queries on Gemini and stuff like that because actually if you almost everything you do online involves quite a lot of servers spinning up in in various places and at the moment Genai is not as bad as people say it was two years ago it was much worse. >> Yeah, thanks Bill. >> Try not to but engage. >> So Jane, you had a question. Oh, thanks Phil. This is really fascinating and I'm going to take a slight diversion as well. My first training, my first degree, my first work was as an archaeologist. And I remember from that learning about when iron came in and the way, you know, early iron tools probably had to be hammered out every time they were used. Um, probably almost less efficient than the ones before them. But, you know, you see where we are now with with the use of metallics. Um, >> yes. And I wonder whether we're there's a there's a real challenge for us in local authorities and I'm going to go a little bit broader than planning it because there's that what there's one thing about the people recognizing that these are tools and and getting away from that kind of woo space. The other one is are these tools in a how far along are we on that journey and are we the right I mean planning you're doing a great job but for the rest of us is it our space to be in there as innovators how much you know and how when is the right moment I think some of the challenges you've put in about how people build their careers and other things I'm taking away completely because it's really important but there is a question for me in there about what's that right moment when we're not going to have to hammer it out every time we use it. >> Lovely analogy. That's a lovely analogy. No, not yet. Absolutely not. Um to give you a very concrete example, um we built a system inside the BBC uh to support the local democracy reporters. You know, the LDRs who sit in local papers, they they file copy for the local paper, which is very good. And some of their copy is rewritten for the main BBC news website, but the copy you file for the Cambridge Independent is different from the BBC News website in style, in detail. You know, you you'd say there was a fire in Cambridge, you wouldn't say it was a fire at the end of Stan Street, those sorts of things. And so we trained a model on BBC News content to do the first part of that rewriting. And it's very good. And that was 18 months ago. and we made it and we're turning into a it's actually a tool that's now being used in the eastern region and in Wales and we're starting to see some impact. So now instead of two or three stories an hour being rewritten for news website, we can do 20 or 30. Massive increase in productivity. News stories that were local get a bigger audience. Everyone benefits. The reporters get more credit. A win. The model we trained was better than GPT4. Now it isn't because the quality just of an off-the-shelf model like Gemini 2.5 Pro has increased so much that with a bit of prompt engineering you can make it do the job we actually had to train it for and so we are still hammering out our tools. I think I'm going to use I'm going to borrow that one with credit >> in future we are not there yet. Um Charlene do you want to comment about this particular point or another one >> it sort of flows onto that. Okay, let let me just let's just make one other point u to to to to Jane. Um >> the one thing we do is we talk to other media organizations and other broadcasters and find out what they're doing to figure out whether our pace of development is about right. >> We don't want to be at the bleeding edge, right? Some of our stuff in R&D is, but you'll never see that. We want to be, you know, offering public value and stable and able to feel confident that what we're delivering is, you know, is valued for money in terms of the cost of develop versus the benefits that are achieved. >> So, we talk to other people and they show us their stuff and we show them our stuff. And then when somebody senior says we're not moving fast enough, we can say, well, actually, Guardian, Reuters, whatever, they're about the same as we are. You we're doing okay. And I think those sort of conversations say through the LGA through other professional bodies could be really useful. So that would be my advice. Yeah. >> Can Charlene there is a risk which you haven't done in the planning. There's lots of offtheshelf but quite expensive products being offered and I wonder if they are a bit like the tool that you've just talked about and in fact in a year's time we'd have done better just waiting and chat GPT would do it for us. It might >> sorry and I was just going to add to that. I think um one of the things that we reflected on because we we did a digital directory um presentation yesterday uh with Laura Terry and um James Fisher. But one of the things that was clear from from that is that there's lots of councils doing different things. So I think the point that you're making, Bill, is we can't do it all. We can't do it all ourselves, but if we're on the journey for the pinch points that we are experiencing and then another council is on the journey for their their you know the bits that they're experiencing, hopefully we can share that knowledge. I think that's that's for me is is critical as well. So sorry Charlene I just jumped in. >> I agree. And finally before Charlene will come to you. Um the question to ask anyone offering a bespoke solution is how is it updated and how do you change the underlying model and if they say you can't then they are a bad idea you their system should be should have be built compartmentalized to do that. >> Thanks Charlene. >> All right just making note of that because that's really useful. Um, yeah. I mean, sort of what Jade and Heather have both been saying is it's that for me it's trying to get to grips with technology changing constantly and it's that not wanting to be at the very leading edge and the very pointy section of it because there is a risk involved at that point. But it's knowing when, >> you know, almost to step into that treadmill and not being too far behind. You know, I've been with the council a very long time and I can remember less than 10 years ago, we weren't doing direct debits over the phone, you know, to set them up. And it's like, >> I'm sorry, what? How can we not be doing that? You know, we were that far behind >> for something that everybody else was doing. And it's that for me it's trying to pick that point of when to step into the treadmill and the other point you made about being able to change the underlying model. And actually, it almost goes back to the, oh, you're saving time on staff because they're not processing the stuff, but what you've needed to do is shift it to developers to keep you at the forefront to get that development in place to progress so you don't become stationary product like you were saying that 18 months ago your product was brilliant. It's now behind again and you've got to have that. And I hadn't considered that element either that you've got to have that development constantly feeding into the background. >> You you do with these technologies they are so not stable. Um and and indeed it doesn't look again perhaps a good analogy is the people on this call. Your staff are not stable either. People get new jobs other places. They want career development opportunities. They decide to have career breaks. They go up and do things. And yet we manage systems in which there's a a take a a churn, you know. Um I've had people who left my team and I'm really pleased they've left because they progressed to the point where we couldn't offer them anymore and they were ready to fly and they've succeeded elsewhere. Um so we expect our IT systems to be more stable. Well, maybe with these tools we need to start shifting those assumptions slightly and coping more with the uncertainty and the change. And that's a big sort of almost an emotional investment I I fear bigger question than than this conversation. The practicalities of it though are that in a again from what I know about the transformation of shared planning over the last few years. You now have the capability to take your heads up to look up from the desk from time to time in a way that you didn't. This is a journey we're currently going on in the product group inside the BBC. Right. Until recently, the amount of sustain effort from our developers was in the 80s 90%. Yeah. Most of the work we did was just keeping the thing running. Minor changes. And what the product team are trying to do now is to create space for innovation to create that headroom as as as our director of engineering calls it within which we can innovate. And the better you run the systems that you've currently got, the more capability you have to think those deep thoughts at every level in the organization. It's not about senior staff having that. It's about a junior planning officer, a software developer, whatever having the capability to look around and think actually this could be different and I have a little bit of time to explore and I have an organization that's receptive to the thing I might say and it might have a little bit of time to explore and we might find something. If all you're doing is processing applications, if all you're doing is heads down doing the job, then quite rightly you can't do anything else. It's just it's impossible. And then the constant rate of change becomes really threatening. And by being a better department, you also create the possibility to to cope with these changes. And I know it's easy for me to stand here and sit here and say that because I'm not you. I don't have to do it. But that is the thing that seems to make the difference. is the thing that seems to make the difference. >> I would absolutely support that bill. I think we've been on that incredible journey that you do go on as a shared service where you do the change management and you you know you drop it's natural. there's a natural curve to it and you come out of that and then when you do that's when you you have that head space to start to really gain traction and then your trajectory gets a lot faster simply because you've got that kind of head space and and the we're fortunate we have the will of our um members Katie but also our leaders Jane you know so that is really really helpful to have those um that that support and know we've got that support is is um is brilliant. So yeah, thank you. Totally chime with that. >> You become part of a community of good practice as well. Again, this is what we find. We there's a thing we call the Genaii salon that we run once a month where some senior people from media organizations come media organizations come together to talk about our experiences and what's happening and share you know good and bad experiences and that means also you don't have to sadly to Toby's point earlier you don't have to do all of it you focus on the things that matter to you and other people doing what they're doing and you learn from them and then when you have the time or capacity to think oh we'd like to do that thing as well You've got friends who have done it who can help share the their their battle stories or whatever they may have been or the success stories to help you do it better. So you're not trying to move forward on everything at once and change everything at once which for anybody working in the organization is is really challenging. You're focusing um and and doing things. So I think you know the way you're currently doing things particularly the the trials and then scaling them up and being very measured about the claims you're making about the difference it will make is the right way forward. And again, it's one of the things we did inside the BBC was early on in the process, we established a structure that had steering group for generative AI of senior people like the director of news and members of the executive committee and the head of our AI research team, whatever, which was looking at everything they they had ultimate responsibility. Anything that involved Gen AI and the BBC, they should be aware of and could if necessary change or kill that. The power was put in one place. Underneath that was something called the central design authority which took a technical perspective on trial. So if any if you wanted something to progress from trial to to you to being funded to to to scale, it had to go through the CDA. So again there was one node of technical experts from across the business who would just evaluate it and there's no there was a center of deep research in R&D which was we were building our own models we were doing the hard work we we built a deep fake detector for images that outperforms the commercial models because we trained it on BBC news material. So you we we are really good at this and that was a center of deep expertise. I used to say foundational expertise but people confuse foundational with foundation models so I stop that deep expertise and then we also built an AI hub which was a team of people this is just getting formed now who will drop into the business in particular places and help solve specific problems and then walk away again. So if you are if you're a data scientist in our audienc's research team and you want to start using an LLM to do analysis of of the information you've got to help us understand a particular demographic then John Howard's team can come and spend four weeks working with you to build the tooling you need to get it all working to make sure you're happy with it and then they'll go off and you now have a new tool and and that approach um has been effective in allowing us to move forward in the measured way and to be sort of to to get backing internally for the things we're doing. And then in public, we made it very clear that the core statement is we're the world's most creative media organization. Human creativity is the heart of what we do. We will never compromise that. Well, we're the country's best planning department. Effective service to customers is what we do. We will never compromise that. You know, there are things you can say out loud about the principles that you will apply that are both reassuring and heartening for people. those two things together having the internal structures that allow you to feel you're in charge of what's going on and in particular you had to have visibility you know just you know really made a difference you know the fact is that you know the executive commission of the BBC was comfortable with this because there were three exco members on that genai steering group there was the right level of visibility and control then we could get on with it and actually do things quite quickly that made a difference as well so again I'm not suggesting what to model it exactly on us. But those are the things that worked for us. >> Thanks. Thanks. >> I'm conscious of of the time, so I don't know if anybody else would like to ask any questions or make any points. I'm very happy to do that. Laura, >> just wondering if we could book another session in with you. I've just find it the whole thing really really fascinating and um yeah I just thank you so much for your time. I'm my first thing on my task list is change my job title so dear >> I don't want to be digital project manager anymore. Um but no I mean I I suppose you want to pick your brain about what BBC plans to do in the next you know years few years coming up. What is their AI strategy? It's just all that sort of because what you've said about even the structure is really really interesting and something that we certainly can bring in. Um >> yes. >> Well again you you'll find a structure that works for you but having a structure in fact it's one of those things where just having made having decided helped. >> Yeah. Not letting it drift. We decided we would do it this way. And at times it was difficult and complicated but we stuck with it. And so yes I'm very happy to continue the conversation. I'm sure Casey will facilitate that next time she tries to do it in person. Um over a cup of coffee. Yeah, other people can go on it as well. Um very happy to to share more. There's quite a lot of stuff about this on the BBC website. Um I'll get Katie to send in the link. So there's stuff you can you've we've blogged about it. We've written about our approach. Again, we're trying to do it in the open. Um so we just done a trial recently uh which is really popular. Actually, this one might work for council meetings, Casey. Um, we have commentary on football matches on local radio and people can't listen to it. So, we've now got a tool which transcribes the commentary in real time from our local reporter and turns it into social media type posts and then post them out up on our live blog every few minutes. And that works really well because you've got a constrained vocabulary. You know the names of the players so you don't get them wrong. And there's really quite a limit to the sort of things that might be going on, you know. So if the machine says alien spaceship has landed, that's probably not the case. But you know, you can constrain the level of of hallucinations quite well. And people who can't listen to the commentary could just pick it up on the phone, really like it. And again, it's another added value service. We would never pay somebody to do this, but we can do it automatically. So I can imagine, you know, why not why not live tweet council meetings by doing it automatically? get more listeners than on YouTube. So, that's just me being random at the moment. >> And that would be great because we put all of that effort in and I think we could probably Sorry, Katie, but I could probably name the people who are watching it. >> Yeah. >> And one of them >> and a lot of them are staff. >> Yeah. >> Oh, but you told me that the local democracy reporters all watch on catchup to get all their quotes. >> Yes, they do. It's a valuable public service. It's a valuable public service for my fellow reporters. >> I I re I really I think that's there's some so many good ideas there, Bill. Again, I think we can we can have a further conversations and we'll arrange that with Katie. Um I really like the point you made there just to kind of before we wrap up around making it clear what our proposition is. I think that was, you know, really positive that we say these are the areas we're not going to compromise on as part of our kind of offer or our journey or our sort of strategy moving forward. And I think I think that's a really um critical element that we need to take away and consider and and um yeah, I thought that was really another bullet golden bullet there I think for us to take take on board. Well, um I'm glad you I'm glad that's helpful. Um separately from this call, I'm very happy to share some of our internal documents with you, Heather, as a just as a as a basis for that. It's not material which we wouldn't make public, but you might find helpful. Let's have a further conversation. The city council shared planning matter a lot to me and not just because of my wife. I live in Cambridge and have done for a very long time and what you do is really important to to the city I love and for the county I live in and to the world I live in. So I'm really pleased if I can be at all useful and if this has been of all of interest. So thanks for having me. >> Thank you. >> Thank you very much. >> Thank you for inviting me. >> Thank you Bill. That was really useful. >> Yeah, it was very much. Take care. Thank you everybody. Five.