Science, Innovation and Technology Committee
Oral evidence: Digital centre of government, HC 790
Tuesday 25 March 2025
Ordered by the House of Commons to be published on 25 March 2025.
Members present: Chi Onwurah (Chair); Emily Darlington; Dr Allison Gardner; Steve Race; Adam Thompson; Martin Wrigley.
Questions 1 - 45
Witnesses
I: Laura Gilbert CBE, Head of AI for Government, Ellison Institute, and Visiting Professor in Practice, London School of Economics; and Richard Pope, Director, Richard Pope and Partners.
II: Rachel Coldicutt OBE, Executive Director, Careful Industries; and Joe Hill, Policy Director, Reform think-tank.
Written evidence from witnesses:
Witnesses: Laura Gilbert and Richard Pope.
Q1 Chair: Welcome this morning to the Committee’s first evidence session for its inquiry into the digital centre of government. I am very pleased to welcome our first panel, which is looking at the achievements, challenges and lessons to be learned from previous Governments’ efforts in this area. We are very pleased to be joined by two witnesses with great experience.
I will ask you each to introduce yourselves when I put to you the Committee’s first question. It is a simple question. I will ask Laura Gilbert to start and then Richard Pope. What would you highlight as your biggest achievement when working on digital transformation in government?
Laura Gilbert: My biggest achievement would be bringing in genuine tech experts at the top of their field from major tech companies, sometimes from universities, and building an almost purely technical team that has been able really to deliver improvements in public services.
Q2 Chair: Can you say a little bit about yourself and your background?
Laura Gilbert: My name is Laura. I have a varied career. I was initially a particle physicist. I had a brief stint in defence intelligence, three years in quantitative finance, and nearly 10 years as one of two directors and chief technology adviser for a medical tech start-up. We brought it through SME to acquisition. I then joined 10 Downing Street in September 2020 as the first director of data science. That was trying to transform the way decisions were made in the prime ministerial office. We went from having lots and lots of Excel spreadsheets, and occasionally photographs of tables, as the data source to work from, through to about 8,000 live data feeds. If you go into No. 10 now, it is expanding across the whole of government. If you were to enter “data 10 DS” on gov.uk, you would see live data on how the Government are performing and what we know about all the services that are being run. There is also a lot of live interactive policy-modelling for decision makers.
In late 2023, I was asked to set up the Incubator for Artificial Intelligence and its technical team. It was building on the 10 DS, 10 Data Science, expertise in data science that brings in engineers, software developers, cloud experts and so forth. What we have done with that team—I am very proud of it—is to go for a really different model of how we deliver improved services. The tagline is radical transparency. The team puts code out into the open.
We are very transparent about everything that is being built, trying to use it as an opportunity to have a two-way conversation with the public, reassuring them that the kinds of things the Government are building are supposed to be for the public good and are in their best interests, and to give them better services and faster outcomes. I always say that what the team is trying to do is make government more human with artificial intelligence. You should get a service that responds to your needs. If you need to speak to someone, the kinds of tools that are provided get you to the right person faster and free up people’s time so they can do that.
As of January, I exited No. 10 and I am now running an AI for government programme, jointly between the Ellison Institute of Technology in Oxford and the Tony Blair Institute that is still advising the Department for Science, Innovation and Tech one day a week. That is the brief background.
Q3 Chair: Thank you. You spoke about bringing engineers and technologists into No. 10 as being a change, and then you spoke about radical transparency as also being something new. How was the Incubator for Artificial Intelligence received within the Prime Minister’s office, and by other Departments. Are there lessons for the new GDS in that reception?
Laura Gilbert: It was technically based in Cabinet Office, but reporting to No. 10. Then it was MoG’d, moved over, after the election into DSIT as part of the digital transformation.
On how it was received, it is very similar to the data team: mixed responses. The civil service has a certain way of doing things. They also have some preconceptions about who technologists are and the kinds of things that they do. Some of the first advice I received, very strong advice from policy experts and advisers, was that you need to hire—I remember a direct quote—“at least two Whitehall sherpas for every technologist to wrap around them.”
Q4 Chair: Two what?
Laura Gilbert: Whitehall sherpas.
Q5 Chair: Whitehall sherpas?
Laura Gilbert: Generalists, because they would be needed to translate what the technologists were doing. My pushback on that is that, as a technologist, I think I am a reasonable communicator. I also think I can hire people able both to engineer and to communicate the value of their work, why they are doing it, how it works and do that really well.
I think you see a split. Right at the top, particularly on the political side, people are very excited to see what you could do. Maybe they have had experience of seeing transformation before that has gone well. In the lower ranks of the civil service, there is a significant number of people who want to learn. They want to see things change, and maybe more if they were tech.
We have a programme that I kicked off in January 2023 called Evidence House, where people could sign up to a mailing list and they would be invited to hackathons. We had a big conference where 500 people brought laptops to learn to code in Python, from everywhere in the civil service; usually a little bit more junior. They would come and work with the engineers. Some of the products the teams built, such as Redbox, which is now our answer to ChatGPT internally, came out of a hackathon originally. Members of private office suggested the idea and built a prototype with engineers, and it has been taken forward and rolled out.
There are also people who, very naturally, struggle with the idea of data and tech people wandering into their space. If you have been working in a certain way for a long time and you believe that you are the expert in how to do that, it can feel as though people coming in with very different skills and a very different approach will disrupt your workflow. It might make more work for you and it can feel like a real risk and threat. We did get pushback in that space.
The other thing about the incubator and 10 DS is that they are central teams. You are trying to reach into Departments. You are trying very hard to work with them. It is not in our best interests to work against them. You really want to work in an integrated way and you want to support them and for them to have a good experience, but sometimes there is a lot of natural pushback against the idea that someone from a different team—who has maybe come in on the basis of political rather than civil service backing, potentially—has come in to try to change the way you are doing things. That is not always comfortable for people.
Q6 Chair: Absolutely. You raise a key point there. We will probably want to explore the difference between political and civil service backing and how we can achieve both in the future.
Richard Pope, I want to ask you the same question and give you an opportunity to say a little bit about yourself. What would you highlight as your biggest achievement when working on digital transformation in government?
Richard Pope: Probably working on gov.uk back in 2012. I was the first product manager on gov.uk as the original GDS was being set up. From what we were building around gov.uk—the team we built around that—we scaled into what became the Government Digital Service. We took a lot of the learnings from that into the Government service standard, which I was also involved in, and which has subsequently been applied to digital services across the public sector.
Q7 Chair: Could you say a little bit about yourself?
Richard Pope: Yes, sure. As a civil servant, I worked on various programmes across Government. Having worked at gov.uk, I acted as a kind of digital product mentor in various Departments, including during the reset of universal credit. I also worked on land registration, apprenticeships and various others. Since leaving the civil service, I am now a freelance consultant and work with Governments nationally, regionally and locally around the world, helping them with digital.
Q8 Chair: You mentioned universal credit, which was a big programme, and the transformation of an essential public service. What are the particular lessons that the new GDS team could learn from your experience with universal credit?
Richard Pope: There are three broad learnings from UC. There is one around delivery and ways of working, which I think is being adopted, hopefully, with the test and learn programmes coming out of the Cabinet Office. You need to start small and grow, grow the team and grow your understanding of the service. The thing that is being tested and learned about is not abstract; it is a service used by real people.
The second thing was around the design of the services. In digitising things, you change their properties. You cannot just digitise what is there already. For example, when you have something as complex as universal credit, which uses data from lots of different sources, calculating people’s earnings in real time, you need to create towards that, to help people understand it. You cannot just digitise existing process. As the Government are thinking about using digital to create efficiencies, you cannot just digitise what is there; you have to redesign it.
The third thing is around transparency. That is something that we didn’t get in early enough to universal credit, but, following it externally, I think the team will be doing more. It is being transparent about how the service works and how it changes so that civil society organisations can really understand. Public services only get better through feedback loops, and digital can be quite opaque. We need to make sure the transparency is there so that the whole of society can help steer digital services.
Q9 Martin Wrigley: Thank you very much. In a former life, probably six or seven years ago, I was a non-exec director at a DFT agency. We had some interactions with the then GDS, so we are talking pre-pandemic. That was painful—really painful.
What you are doing in No. 10, Dr Gilbert, sounds very interesting. How do you make that work outside the personal space, where you are trying to reach across the wider civil service to implement GDS, whatever GDS is these days? I am not entirely clear where it is right now, but it sounds as if it is in a better position. How do you get that going forward, rather than having the painful interactions that we had some years ago?
My background is as an IT architect in major telecoms systems, billing systems, all that sort of stuff, and putting transformation into that sort of environment, so it is something I am quite familiar with from a professional background. What was there then was clunky. What you are doing now sounds quite radical, which is good and impressive, but it requires, from what you describe, personal influence to make it happen. You cannot do that across the entirety of government. How do you spread out into those areas—sorry, this is a fairly clunky question—learning from the previous GDS clunky working into a much more 21st or, dare I say it, 20th century way of doing digital transformation?
Richard Pope: GDS in its early days was creating influence through delivery. It was definitely hard to engage with other parts of the company because there weren’t necessarily the people with the digital skills there. That has radically changed now. Every sizeable Department has really good digital people, certainly at the practitioner level, in a way not seen around the world. Central Government in the UK has an amazing number of brilliant digital designers, technologists and researchers working for them. It is an amazing resource that was not there a decade ago. That is one big change.
Looking forward to what is in the blueprint, the commitment to have senior digital people in every Department at exec level and to have dotted lines to a GCDO could be quite powerful as well. All those things were not there a decade ago. It was very much show through doing. Sometimes that worked; sometimes it was a bit clunky. You said it was very clunky to start with. There was initially a very difficult relationship between GDS and DWP, but out of it came probably the best digital delivery team in the public sector.
Laura Gilbert: I am slightly less optimistic, and I tend to be a very optimistic person. I think it is going to be very hard. Certainly in GDS there are a lot of technical skills and there are pockets of real excellence through the system. I come across a lot of people in government, and there are many brilliant people but there are a lot of people in probably quite senior digital data roles who I do not consider are technologist, data people. I wouldn’t hire them. It is very difficult to hire technologists well. The way the civil service hires is not suitable for this sort of purpose; it is really hit and miss. It is very easy to come in with a CV that has all the buzz words and all the jargon and all the programming languages, but the system does not have a way to hire that assures that people can really do their jobs. Until that changes, it will continue to be very difficult. In many places, there is a standard of digital expertise that is far below industry and far below where it ought to be.
The way I hire in the incubator and, in fact, data science teams everywhere, is through hands-on testing. You would very much agree to learn by doing and the doing has to be in the hands of the user. To be in the incubator as an AI engineer, you have a four-hour coding test: “Here we go. Here’s some data. Here’s an API. Here’s a problem to solve. You’ve got four hours. Some kind of app. Go nuts.” You can solve it any way you want because we want people who can solve problems. When we come in four hours later, these people—their CVs have the experience on it; we have assessed them and we know they should pass the bar—will be selected based on what they have done, how they have done it and how they communicate it. We look at their code. We try to figure out which model they used and why, and they need to be able to answer questions on it. When I hire a private secretary, it is exactly the same thing: “Here’s five briefs. Make me a day pack because that’s the job I want you to do.” While the civil service is still doing behaviour-based interviews and trying to apply that to technical roles, we will continue to have quite a problem.
My worry with the blueprint is first—there are lots of great things about it—that they said 10% digital skills. Will they really be digital skills? Would Google hire them, because that is the standard we should be going to? If you are going to run a country really well, you need the kind of people who you would hire in Microsoft or even a tech start-up to be at that level. Seniority is an issue as well. Very often, it is still the case that the people in charge of technical delivery are existing civil servants who have been the director general of something else. There is a role free, so the system says, “This is available. You can go and do that.” There is still a very strong view that they should be fungible, that people should be able to do anything. In fact, I have had a lot of very good really senior civil servants.
I was told to apply for the director’s leadership programme, which is a great programme. The last interview question I had to get on to it was, “If you want to be a director general, you will need to be able to run policy teams. Are you prepared to do the upskilling necessary to be able to go and run policy teams?” I said, “Well, I like to learn. I probably could have a go at that but I’d be terrible at it. I’m very good at running digital and data teams.” I think the thing that needs to change is the system rather than me.
Q10 Chair: Specifically on hiring people to the civil service who would be hired by Google and Open AI, say, wouldn’t the salary differences be a barrier to that? Are there other motivations that the civil service can play on to hire those kinds of people?
Laura Gilbert: Yes, I think so. The incubator doesn’t pay industry salaries; the pension is attractive. You have to be able to give people salaries in a range where they do not feel stupid for taking the job. That is basically some advice I got from Emily Lawson. If people are taking a third of the salary and they have to explain to their husband or wife why they can’t have a holiday and the children aren’t getting as much money as they could, they can’t do that. When people are younger and they have a dream, they are prepared to take a lower salary, if you can get it to within the 30% level and people can see that they really can make a difference. This is what we did with the incubator. We sold a vision: “Come here and try to really change people’s lives.”
There is something in the tour of duty idea. When I joined, I thought maybe this could be a career, but it is very difficult to go up as a technologist in the civil service. As I went through, I thought I could do it for a certain number of years and walk away, having done something. If we could get a lot of people who believe in that tour of duty idea, and if we could communicate to the tech community and say, “Come in for two years, fix this and then you can go back to industry, and that’s a good thing to have done. It will feel good on your CV. It feels good for you personally. You get experience of something different,” there is a real sales pitch. We are starting to use it but we could beef it up. I think there are a lot of people in the technology space who want to work on public good initiatives. A lot of the time, they have a bit of money on the side.
Chair: Thank you.
Q11 Martin Wrigley: Moving on from there, is DSIT the right place for the digital centre of government? Will that work?
Laura Gilbert: It remains to be seen. It is a big transformation. DSIT before the MoG was 2,200 people, of whom 100 claimed some degree of technical skill. It was formed from moving policy teams from other areas. In some cases, the reason for that change was that those teams had not perhaps done the job that people wanted them to do where they were originally. It is a difficult basis for setting up a Department that you would then have to pivot to delivery. It is a very big culture shift.
You have to be careful with people bandying around the word “upskilling” a lot. You can upskill from any background to be an AI engineer, but you have to have aptitude, you have to want to do it, you have to care about it, and you have to understand data and statistics. If you just try to upskill a Department, you end up with people who are not at the level they need to be professionally to do it and that is when you end up with mistakes. That is when we have AI that is unethical. People never do that on purpose. They do it because they are not skilled enough; they have done it wrong.
It is very difficult. The Secretary of State announced to the Department when he arrived that he wanted it to be a delivery Department. They understand that that is the drive, but it is difficult to do. What you need to do with the digital centre is see it as a separate enterprise somewhat, which is under DSIT, rather than expecting DSIT as a whole to be very integrated into it. If the digital centre is very tech-focused and hires the right people, it could go very well. There is no reason why DSIT could not support that, but we perhaps should not expect DSIT to entirely change its species, as this goes along.
Martin Wrigley: Thank you.
Q12 Chair: Do you want to add something, Richard?
Richard Pope: Yes. I’m less worried about delivery. GDS can deliver; it showed it can deliver and it can do more of that in the future. What I would like to say on whether DSIT is the right place is that digital is as fundamental to running a modern digital state as a finance capacity is. GDS, the new digital centre, will need to be able to influence across the public sector.
Chair: Absolutely.
Richard Pope: Some of that will be about delivering strategic things. It will also be making sure that we are getting good digital governance, that we are able to prioritise things that create value across the system. Most digital projects in government at the moment are attempting to do something that a Department already does a bit cheaper. It is hard to look across the system and go, “We’ve got duplication. We’ve got the opportunity to do something that has a multiplier effect.”
Is this HM Digital Service a counter to HM Treasury? I don’t know. At the moment, it feels like it is a smaller entity. It will need to be able to show that it can exercise influence across the system. If GDS needs some backing, will the inter-ministerial group give it that? When it says, “We think, on paper, this business case is not a business case, but strategically, for the UK, we think there is an important bit of work to do here,” will it back it?
Laura Gilbert: That is exactly right. If it gets the right backing, it does not matter where it sits.
Q13 Dr Gardner: I have a couple of quick questions. Martin used the word “clunky”, which made me think that, in the state of digital government review, key challenges were identified. One, obviously, was legacy systems. Also, cyber and technology resilience risk was there. In such a radical transformation across various Departments, how sure can we be that you are not laying this radical transformation on shaky foundations, and do we need to fix some of the basics first?
Richard Pope: You are absolutely right. There is a lot of legacy technology, although I don’t like using that term. It lets people get off free—“It was here when I got here.” People talk about unmaintained technology. There is a lot of unmaintained technology that the Government and the public sector rely on to do their job and it is not good enough at the moment. There is obviously a job to improve that. We need to do that at the same time as delivering value for the public and value for public servants.
There is a bit of a gap in the blueprint. Firmer foundations are pitched as being about cyber and about legacy, but actually we need to be thinking a bit more holistically about digital public infrastructure. What is the digital public infrastructure the UK needs to deliver the public services of tomorrow? How can you invest in common capabilities that cut across individual bits of legacy technology? I have seen a couple of reports recently, which are worth replicating, where researchers have looked at the data infrastructure in particular areas, such as the court system or in the health system, going down to the database level and saying, “What data is recorded? Is this thing fit for purpose?” We need to do that on a systematic basis.
Q14 Dr Gardner: On the flipside of that, you talked about unmaintained systems. When you are building the radical transformation, and you are talking about ethical development, what governance structures are you putting in place so that the new systems we put in place are maintained and you have proper process, including updating decommissioning? What governance processes are you putting in place or should be put in place?
Laura Gilbert: It is a detailed question probably because the governance needs to relate to exactly what you are doing. There are well understood ways to do data governance: software transformation, the ways that you do your penetration testing and the ways that you record how you are using the data. I genuinely think—because it could be quite a long answer—that it comes down to the professionalism of the people implementing it.
The reusable point is very well made. You have lots of legacy tech and lots of Departments. We spend a lot of money maintaining or, in fact, not maintaining it every year. What we probably need to do is make sure that the digital centre is able to build something on the side that is unrelated, which you can then lift and shift over and is reusable across lots of Departments.
If you have a very skilled technical team with the power and the money and long time horizons in place, when the budget will not get cut next year or be used for something else, you can do that; it is achievable. If you leave the system the way it is, it is not going to fix itself. It requires massive investment. A lot of that investment needs to be put in very carefully, ensuring you have the right people in place and then believing them when they tell you the answer.
Q15 Chair: Thank you very much for that. We now come on to some of the costing issues. We have seen the figure of £45 billion as the prize to be saved in the digitisation of public services. We have not seen very much detail as to where that £45 billion is coming from but we have seen a few bullets, as it were. Can I ask each of you to say briefly how credible it is that £45 billion can be saved in digitisation?
Laura Gilbert: Wonderful. I have seen the methodology for producing some of those numbers. A few different organisations, including one internally, have come up with roughly that figure. The methodology is perfectly good. If somebody said to you, “How can you estimate what the savings could be?”, it is a reasonable guess, but of course it is a guess. It is very difficult to measure.
The thing with that £45 billion is that there is a front-loaded part where there are easy gains that you can pick up and say, “If we did this thing, if we automated this and there are this many people, there is this much time saved or money saved on quite big things that do not rely on legacy tech that you can just go in and implement.” An awful lot of that is a long tail where, if you invested this much for a year and a half, you would make a 1% saving, but actually there is an opportunity cost in doing that. You would need a lot of people focused on it and they could be off doing something else. I don’t think there will ever be a point when somebody goes, “Here it is. We’ve done the £45 billion”.
What I would like to see around that is a lot more detail, in that there is a list of 15, 25 things we will achieve and we have an evaluation of what that will do, and that is the target for this Government. I would like to see a different number. The £45 billion is perfectly reasonable if somebody had, say, come up with a ballpark estimate of what you could automate across the public sector, but I don’t think it is the right question.
Chair: Thank you.
Richard Pope: There is this paradox when it comes to savings and digital. It is really hard to digitise an analogue or partially analogue service and not save money. At the same time, services or projects that are predicated on saving money tend to create outcomes that do not work for the public particularly well. That is why the learning from universal credit, the idea of starting small and growing something, is definitely the model to replicate. You will get the savings; Government will get the savings. But if you start off by going, “If we can turn off all of those and all those and then we replace this system, we, hopefully, end up with a service that the public can use or public servants can use,” it does not tend to work. Starting small and growing works.
The other thing is about savings. I wonder if Ministers have slightly painted themselves into a corner with AI. It seems to be that AI has become synonymous with savings, and that is part of the picture. There are almost certainly opportunities for efficiency and savings through AI, but there are through more traditional digital processes and better design as well. There is a really interesting project that MHCLG have been backing, creating new open source planning systems for local authorities. They have saved loads of time because they designed something that works for council officers. I don’t think there is any AI in it. They have just designed a really good system that works for people. We would need to make sure that is in the mix as well and that we are using the full scope of digital to try to find those sorts of things.
Q16 Chair: There are opportunities that come from AI but there are also opportunities that come from generally being more efficient, and digitisation of existing processes and different services.
Something that surprised me is that there is no mention of data cleansing in the blueprint, building from the point that Allison was making. There is an old saying, isn’t there, “Garbage in, garbage out,” on the quality of data? What is your view on the quality of data—you talked earlier about the data infrastructure—particularly in the context of the Sullivan report published last week, which highlighted the challenges in collecting and sharing data on sex and gender due to a lack of common definition of what each of those meant, and also when each or both of them should be collected? Sex and gender are contested terms in certain quarters. Doesn’t it suggest that the Government need standardised common definitions for data? Where is that work going on, if it is at the moment, that you are aware of?
Richard Pope: Going off and writing standards that end up just published on a website is not the answer. You get to those things through delivering common infrastructure. There are a few too many euphemisms and metaphors in the blueprint when it comes to data. We have the data library, we have a digital backbone and there is something else as well. We are not talking about it for what it is. What we are really talking about is data exchange systems for the public sector, the sorts of things that Estonia, Ukraine and India have really invested in. If you build that, again starting small and growing it, you bake in the data standards, you bake in the ability for one Department to trust that it is getting data from another, knowing the format it will be in, in a predictable way. You get the transparency because you publish somewhere what data is held and what standards it meets.
I was having a think on the way here today. There are five separate data exchange projects or initiatives that have come out of GDS or central Government as a whole in the last 10 years. I assume that the blueprint is signalling another one. These things are very much long-term investments. They take decades. The Government need to pick something and stick to it.
Laura Gilbert: I agree with all of that. A lot of these things are technical, including AI ethics—it’s how you build it—and the data standards are the same. The health service did fire as a data service standard internationally; if you want to share data you have to conform to a fire standard. It’s very practical. You can see what it looks like, so for the technologist when you send the data it is a technical piece. We have to fit the data into the slots—otherwise, it won’t go—rather than a piece of paper or a website that people struggle to look at or engage with, particularly when they are busy. That practical implementation is very important.
I think there is a real lack of clarity on what people are trying to do in the system. I was quite concerned about the data. I communicated that to a lot of people. Since the four years I have been there, I have seen roughly—I don’t have the exact number—£250 million being spent on data infrastructure programmes. Their value for a large part has been low and not at all far-reaching. I do not want to see that happening again. My concern with the data library and the language around data sharing very similarly, quite metaphorically, is that what needs to be very clear is why we are doing this, and exactly what problem we are solving. The data problem you could solve might be data exchange inside government. We are not talking personalised data; we are talking about where the money is spent, what it is doing, what are the outcomes, where the people are. Even that sort of thing is currently not easy to access or share. It could be about public data, in the same way that transport data is released to the public and generates a lot of companies that improve your travel experience. We could do that with a lot of Government data.
It could be about personalised data. There is huge value in connecting people’s health records with their tax records and their welfare, but the pushback is that there are huge risks as well. If you are to do that, as Richard said, you need a 10-year plan; you need to be confident that you are communicating exactly what you are doing and why. It is not just writing about data. The first thing that happened when I joined Government was that I was given a 200-page data document to sign off as the data strategy. I said, “I’ve only had a day to read it. I don’t want to sign it. I do not think that in five years this will solve the problem,” and it hasn’t, so I would like us to stop doing that.
Q17 Adam Thompson: I want to come back to some of the conversations we were having a little while ago about the new Government Digital Service, and drill down on a few interior points on that. First, do you think that the new GDS has the right remit?
Laura Gilbert: It is a difficult one to get right. If you are going to fix the whole public service digital infrastructure, it is a very big ask that would require more people and more investment, so it needs to be bounded, which means that you need a sense of prioritisation of what you are doing. On paper the remit is largely good. It may be not the remit but the implementation that I would be worried about. I think that is the answer.
The other thing I would be worried about is how we track whether it is delivering on that remit in a way that does not involve lots of written documents but involves actual outcomes. How much money and time has been saved? What is the public’s experience of improved digital services? Are we seeing the outcome we want, which is that people feel that things are much better than they were? The remit has to be bounded by the budget and the people available, and it has to be reasonable, but I care more about the implementation.
Richard Pope: It is generally good. Bringing together lots of disparate groups across Government into a single place makes sense. There needs to be a strong centre of digital within government. They have highlighted a few practical things. There are a few positives and a couple of questionable areas. There is the focus on the administrative burden, measuring that across the system so that you can choose whether it makes sense; the commitment to digitise paper credentials, not just driving licences but the sum total of things that people need to prove eligibility for things; and the ability to mandate Departments to create APIs so that you get interoperability. They are all quite bold system-wide things.
That said, maybe some of it is down to GDS’s prior history. It is still quite focused on the idea of citizen-facing services. I have some questions about that and the focus on personalisation. I am not sure that is necessarily a good place to start, but there is definitely a focus on citizen-facing things, as opposed to businesses and the wider economy. You see lots of other countries investing in things like paperless invoicing, data exchange between Government and companies and digital identity for companies. That is not clear at the moment. Similarly, the relationship it will have with local government is evolving, but how is it going to interact with devolved government if we have some UK-wide digital public infrastructure? How is that going to work functionally? There are a few things to play out.
Q18 Adam Thompson: That is really interesting. DSIT has talked about its proposed account management structure. What does that phrase mean to you both, and do you think it is the right approach for the new GDS to take when engaging with different Government Departments?
Richard Pope: I have not come across that phrase; maybe Laura has.
Laura Gilbert: I don’t really understand what it means either. I imagine it is supposed to mirror the way Google has an account management structure, where they have customers and manage that relationship, so you have a degree of one-to-one mapping where you have a relationship with a Department. It seems sensible. It could involve a lot of overheads. That goes wrong in government when it cuts out the reusability point, so you would need to be very careful about that.
We have an account management structure effectively in Treasury for spending reviews, where a spending team works on a Department. The result from that can be that, when the Budget comes out and you scan the whole thing together, you can pull out—I won’t be specific—incidents where clearly this thing in this Department is about the same as that, and there would have been quite a big economy of scale in doing them together but no one spotted it. I do not think it is a bad thing to do. Relationship managing will be crucial because people will be very uncomfortable in Departments. Again, it will be in the implementation. Does it cut out reusability and a broader eye? The alternative structure would probably be product-specific where there is a data improvement team, for example, that builds APIs across Government and then reuses them, and you focus on the technology rather than the account. I probably have a slight preference for that. I don’t know how you feel, Richard.
Richard Pope: The centre needs the capacity to spot emerging patterns, almost like mergers and acquisitions of teams and things like that: “There’s a team over here and a team over there, and we’re going to take them out of the Department and get them working together.” There is a risk if the relationship is centred only on individual Departments. That makes it harder to create delivery capability that cuts across them. If you want to create services that meet an end-to-end need for a new parent, or someone leaving prison and going into the probation system, that will cut across multiple Departments. If your lens is always on a Department-by-Department basis, you end up reinforcing the Conway’s law problem. The missions may be the way out of that. We will see how those teams start to evolve, but if you can do that alongside creating teams that are cross-cutting, and they have the backing to deliver outside the bounds of Departments, maybe it could work.
Q19 Adam Thompson: Thank you both very much. Chair, maybe there is some scope for seeking clarity on what we mean by account management infrastructure, because I am not 100% sure what it means. Thank you both for those answers.
I have a couple more questions. Are the initial kick-starters that we have heard about the right areas for the new GDS to prioritise?
Laura Gilbert: The gov.uk app and the wallet feel very sensible to me, and work is reasonably well under way on those things through GDS. It should be deliverable and impact people’s lives in a positive way. It is a good signal of progress as well. Some of them are a bit vague, such as the collaborations on long-term health conditions and disabilities. One would want to see specific programmes on that. It would be interesting to see what the early targets look like and how they will choose them. It needs good targets, but it is very non-specific. Gov.uk Chat has been under way for some time. It is a very normal thing that an organisation would do. How can you use AI to interact with our system efficiently? Why not?
The AI accelerator upskilling programme slightly worries me for the same reason as before—helping digital professionals to become machine-learning engineers. I do not quite know why you would do that. If you are trying to deliver more AI expertise, it is important to find people who have the aptitude and want to do it. If you have somebody who does excellent user research, they should continue to do that because it is very valuable. There is a risk of weighting everybody away from core digital skills, as Richard was saying, that fix the underlying infrastructure and some of the more easily digestible problems and some of the more annoying problems, and telling everybody, “Let’s all move to AI.” If you are to move to AI, I prefer the model where, as we have done at the incubator, we steal people from Microsoft and IBM and bring them in, because they are state of the art. We are not trying to get people who have a casual interest and turn them into experts. When Government have delivered that kind of upskilling before and tried to do it as a sheep dip, the levels of improvement you see are really not measurable. That is the problem.
The final one is the Government vulnerability scanning service. I do not see why not. It is a solved problem really. Vulnerability scanning is a very well-understood science base. Richard might know more about this than I do. If you know where all the services are and have a list that goes through them—the penetration test and the whole lot—why not have a central system that is assured and owned by the Government that people really understand? You can keep that internal and it would help you to create a data list that says, “What are our reusable problems that we can solve where there are vulnerabilities across the system?”, and try to fix them in one go. That should be achievable. We know how to do it, and implementing it is a good idea. There might be gaps in that, but I think it is just a piece of work that is ongoing as the digital centre structure comes out. People do not yet understand the scope of the opportunity. We expect to see that evolve.
Q20 Adam Thompson: Thanks, Laura. Do you want to add anything, Richard?
Richard Pope: I would like to see a much bolder list. It looks a bit like some things that are already in train anyway. I do not know whether it tests some of the bolder ambitions in the blueprint. On encouraging things, the wallet app could be quite transformational if it can be positioned as a capability for the whole of the UK, not just a central Government thing. We will see how the gov.uk app itself plays out. I hope they are doing the work to understand what Departments need from that. I would like to know what existing digital services could—not be switched off, but not have to run their own infrastructure any more because the app will provide a bunch of functionality around forms and notifications. What are the first few things that get turned off? We know there is real value there. I hope that in its initial phase they get to revisit the kick-starters and focus on that a little bit.
Chair: We have to pick up the speed a bit.
Q21 Adam Thompson: I have one last question, Chair. Laura, you specifically mentioned the incubator. From your time working on the incubator, what are your thoughts and feelings about how it went, whether there are any failures and how we can learn from that moving forward?
Laura Gilbert: I am very pleased with it, but I would be because I built it, so you should really ask other people. It showed that Government can do things very quickly and that hiring the right people is possible and is an effective way to go. It is very fast. When you get people who really understand data and tech, and can pick up a computer and build, you can solve specific problems incredibly quickly. Every Department needs that. There are things the incubators could do just to make Departments function better, ignoring the wider public services to start with. You just need a quick app to sort out that this team needs a dashboard or that team needs an app that does something quickly. For anything they are repeating, you can build that and shove it on everyone’s computer and off you go.
In terms of solving problems in the system, the things that worked well were when there was ministerial backing—where you go in and insert an algorithm to replace a non-automatic process at the moment, or when you build something that is stand-alone. There is something called Caddy, which we built with Citizens Advice. They use it to improve the advice they give people. The turnaround is positive; case resolutions have gone up by about 60% using that system. Lots of it is really good.
As for where it has fallen over, there are two things. One is that strong political backing means that when an election is called, there is immediately a hiring freeze. That hiring freeze took about eight months to unblock. The team was supposed to be originally at 70. That put a lot of pressure on people who needed big things to do. You needed specific technical skills that we had not managed to get before the hiring freeze. One specific thing from which we learned early on arose when the Prime Minister wanted it done, the Secretary of State wanted it done and the director general SRO wanted it done. The digital and data team in the Department did not want it done, and they won. You got all the way through pilot evaluation to a system that would work but was never implemented, and probably will not be.
There is something in that for a future GDS where you are building stand-alone systems. I have very high confidence in their ability to deliver something like the app and the wallet. They can do that. It is a stand-alone service they can do themselves; they do not need so much investment from the rest of Government, apart from permitting the data to go through. That can go really well. Transforming Departments is harder.
Chair: That is really important but we are behind time. Thank you very much for the evidence you have given, but we need to pick up the speed a bit.
Q22 Steve Race: I want to talk about the “Tell us once” principle. Can you help us to understand how that approach might change the doctrine of public services? Can you give examples of what reusable data might be? If it had been implemented at the time across the Government, might it have had an impact on your work previously as well?
Richard Pope: There are various forms of it, but it is the idea that either data is stored centrally in one place and is accessed by multiple services, or you have some kind of event-based system where a Department gets notified and it tells everyone else when, for example, somebody’s residential address has changed.
How does that start to change services? First, on the types of data, if you look on gov.uk, the pages for all the services have something that tells you the eligibility criteria. If you copy and paste those into a spreadsheet, as I did once to help some people, it is obvious stuff; it is address; it is age; it is qualifications for other benefits. Services and policies as designed today tend to use those sorts of things. There are also some from outside the public sector—for example, things like proof of earnings, right to work and proof of savings. It is not just a central Government thing; some of them live in local government as well. Again, it points towards a UK-wide data exchange system of some sort, or certainly interoperable ones. That said, once you have that kind of infrastructure, the policies and services you design in the future start to change, so the data we use today will not be the same as the data we will use in the future.
On how it changes services, it is quite fundamental. Universal credit is a very complicated, data-driven service designed by policy people to use lots of different data points about people. You can do specific means-testing and targeting. That puts a lot of burden on the public. People spend a long time filling in very long forms. With “Tell us once”, data exchange has the opportunity to remove that. You are turning it on its head: “These are the things we know about you; tell us the things that are missing.”
As Laura said, there is definitely a risk of unintended consequences, which is why you need to design understanding into the system so that both the individual understands what is happening, as best they can, and civil society organisations have the opportunity to understand how things work. That sort of thing is not prioritised today. It is very hard for an organisation like Child Poverty Action Group to understand how universal credit works, for example, so transparency has to be part of the mix, but that is not a reason not to do it. We can save the public and public servants a lot of time. We can remove a lot of extra form-filling, and we can start to make services that are much more proactive. There are challenges to Government. Are the Government going to choose to tell people that they are entitled to this or that benefit? They are political choices ultimately. Technology enables it; it is a political choice. It would also allow Government and policymakers to target new ways. For example, in health, if you want to get people into prevention schemes you have the ability to target them; you have the ability proactively to offer people access to services in a way that is really difficult at the moment.
Laura Gilbert: It is fully transformative. Everything you said is exactly right. If you were to say that we are doing that now as a Government, every Department would radically have to change the way they do things. You would have to digitise everything. You need APIs now. You would have to agree to share things that currently Departments refuse to share with other Departments. There is no way to do it without that.
Q23 Steve Race: What sort of data would be shared, and why?
Laura Gilbert: There are secure ways to do it. You can do anything from a checking system to the form-filling version. For the form-filling version you have to have a way to share things. Any Department could access somebody’s name and address in order to pre-fill forms. There is also the kind of service where you go back to the initial Department and check something. If somebody says, “I declare I don’t have a criminal record,” of course you can go and check and get a yes or no answer. Those are very similar technically, in that you have to have automated data streams, but the implementation is different. Did that answer the question?
Chair: Steve, do you have another question?
Steve Race: That’s fine.
Q24 Chair: We have talked about transparency and openness and some of the challenges in the civil service in applying that technically and culturally. I hope you would agree that there are two key stakeholders in digital transformation: the workforce and the users. In terms of engaging the civil service workforce, we have been told that the Cabinet Office routinely discusses issues of digitisation with the unions. Was that your experience? Were there other forms of effective workforce engagement, and what lessons could the Government learn from the past?
Laura Gilbert: I wouldn’t have any idea whether the Cabinet Office is discussing digitisation with the unions. I presume they must be; it is just not something that was necessarily communicated. I don’t remember any communications like that when I was there.
I think real efforts are made to communicate with civil servants. There is the one big thing initiative—I don’t know whether you have heard of it. It started two years ago. One big thing is data. Everybody has three months to do some data upskilling. It is a really interesting idea. A lot of effort went into it. Again, the implementation can be a bit off. People do not learn from watching videos. I do not think that spending quite a lot of money on a consultancy to come in and develop videos on what data is, with a little arrow which you then follow, is an effective way to do it, but the will is there. The one big thing this year will be AI, as I understand it. I am strongly encouraging that what that should look like is giving people a device and saying, “Here is the AI. Here’s how you use it and here’s how you generate a policy paper,” and then you make sure that it is right, rather than just talking.
Richard Pope: I cannot talk to conversations with unions or anything like that, but one thing that works in the same way as designing services for the public is involving them, understanding what works for them and testing it. You have to do the same with public officials as well. Designing systems that work for civil servants and other public-facing officials is how you get efficiencies.
To point to one example I saw recently, there is obviously a massive backlog in SEND provision and EHCPs in local authorities at the moment. If you look at the software that teachers, educational psychologists and local council officers are made to use, it is awful; it is just dreadful. They lose hours and hours of time using software that does not work for them at a time when we have a huge backlog and a shortage of educational psychologists. If you design software that works for them, which we know how to do, if only you apply the right people and the right teams and start in the right way, you can save loads of time, but you have to design it with people.
Laura Gilbert: Exactly right. The hours we spend just on HR software in government to try to manage staff and pay people are unbelievable.
Q25 Chair: While agreeing completely that it is the workforce who will drive the success of this transformation, it is interesting that we do not seem to have an understanding of the official ways of engaging with the representatives of that workforce—that is, the unions. We have talked about savings, but we have not talked so much about funding. There is a thing called digital funding, and we are trying to find out exactly what that covers, but in terms of DSIT controlling, driving or engaging with other Departments for the funding of digital transformation and investment in new systems, do you have recommendations on that? Was it an issue during your time?
Laura Gilbert: It is definitely an issue, because of course there is not enough money and people choose to prioritise differently. In much the same way that they are looking at new ways to procure, new ways to fund this might be quite useful if we genuinely believe we are driving this much benefit over this period of time. The conversation with Treasury looks like, “If this much is saved, can we put this much investment back into the pot?”, separating the pots for digital. If we are not driving those sorts of savings, we are doing it wrong, so there might be something in that space that could be useful.
Richard Pope: The new digital centre needs the ability to stop/add projects. A lot of the savings generated from the original GDS were from what is called spend control. Just to be able to review upcoming digital projects and say yes or no to them is a capacity that needs to exist.
It is right that the Treasury is experimenting with different approaches to funding projects, especially when it comes to digital infrastructure. The way you fund that is different; you need to be able to start small but fund for a very long period of time. I know that everyone points to Estonia as an example of digital government, but the thing people do not point to is that it took a really long time. X-Road, which is their data exchange, didn’t create any value until they had 50 services connected to it. Then, suddenly, you get the network effects. You have to fund these things for the long term, but hopefully you make the teams so that they are set up in the right way and are working in the open, so it is easy to understand when they are likely to deliver value.
Laura Gilbert: And, crucially, to stop things that are not working. Government do not do that. Once it is funded it carries on and carries on. Some of the things we are doing now have not been working for five years and are still being funded.
Chair: Perhaps you can write to us and tell us what those things are.
Q26 Emily Darlington: As a final question, this is a chance for you to influence the future inquiries of this Committee and, hopefully, Government. What should we be focusing on? In particular, are there things that you proposed to the previous Government that you still truly believe are transformative but were never taken forward and started? This is your attempt to pitch those projects again.
Richard Pope: That is a good question. It is repositioning transparency of digital services not as a “nice to have” but as governance, and the way in which we understand whether or not services are working. That is both for society and for individuals. It is really important that the public have an opportunity to understand how digital services work and how they can change them.
Q27 Emily Darlington: How would that look to the public?
Richard Pope: We have ended up adopting a west coast American approach to design where everything is seamless and it just works. That does not work in public services. Public services get better only through public understanding of the system. Digital sometimes masks that, so practically it means helping people to understand the institutions they are interacting with; helping them understand their rights and the underlying legislation; and helping them understand when data will be moved from one Department to another, not so that it gets in the way but just to help people understand how their Government works in a way that private sector services don’t have to worry about.
The other part of the question I have now lost. Would you repeat it?
Q28 Emily Darlington: The other question is about projects you pitched under the previous Government that didn’t get taken up but you think this Government should.
Richard Pope: That was the thing I was always banging the drum for— greater transparency.
Laura Gilbert: I really agree on the transparency. I was very interested in digitising the spending review. That is something that you should really do. When you submit a request for funding in the Budget, there should be a bit of a scan and it comes back and says, “Well, you don’t have an evaluation plan, so we’re not even going to read it.” There has been a lot of work on trying to evaluate major Government spend. The evaluation taskforce was set up, and about 8% of major spend was evaluated in 2019, and now it is roughly around the 50% level. I would have to check the numbers. There is still a long way to go on that. If we are not evaluating whether the spending is driving the outcome we want, we are wasting a lot of money. I would like to see a lot more focus on the evaluation of the spend, and, when it is not working, stopping and spending on something else.
I would like to see Government, basically, run like a hedge fund where you invest in certain things that you kind of know will work slowly over time—digitisation, and so on—and you have a few big pitches, but if there are things that you invest in and they go down, you stop funding them. There is a variety of risk that goes into a real portfolio. Let’s invest in some risky things, but not unless we know we are going to start doing them. Digitising the spend process, with much more transparency around it inside and outside the system, could be transformative. If you are looking into things going forward, trying to get some numbers on where the digital spend has been and correlating them with what happened as a result of that digital spend would be very informative.
Chair: Those are very good ideas; thank you so much, Richard Pope and Laura Gilbert. We have run a little bit over time, but that is testimony to the fullness of your answers and the interest of the Committee. We will now break for the next panel.
Witnesses: Rachel Coldicutt and Joe Hill.
Chair: Welcome to the second panel of the Science, Innovation and Technology Committee’s first session on the digital centre of government. On this panel, we welcome two new witnesses. I am going to ask Allison to start the questioning.
Q29 Dr Gardner: Welcome. My first question is to you, Joe. Originally, GDS was housed in the Cabinet Office, which on the face of it made sense because it had the convening power to look cross-departmentally. However, there was the challenge that it had 40 other things to do, as was reported, so it was moved to DSIT, which may be a bit more specialist but it has its own challenges to drive the reforms without that central power base. You have said that it is not clear whether DSIT is the right Department to host the new digital centre. What needs to happen for you to be convinced that it is?
Joe Hill: Thank you. To start, the challenges are exactly as you have put them. At the moment, we are relatively agnostic on what the right long-term home is. Neither seems quite perfect. The thing that would persuade us that DSIT was the right home, the Cabinet Office was the right home or some other construction of the centre of government was right would be evidence that one of those is the best model for doing the things that the digital centre needs to do, the things that only it can do. Those are a combination of doing some of the setting of core standards, things that you want to do once, and you want to do for as much of government as possible, some of the things that previous witnesses talked about, like data standards, transparency and interoperability, and where the best place is to have the driving force for the Government’s key priorities. Sometimes, that driving force will involve stepping on a few people’s toes—it is impossible to avoid that—in which case you want to set it up in the strongest possible place to get the right outcome.
Q30 Dr Gardner: Thank you very much. Rachel, my next question is to you. I am going to slightly load my question by saying that it is a core belief that whenever you are doing any sort of transformation, particularly within technology, you don’t do stuff to people; you do stuff with people. Peter Kyle mentioned reassuring the public that privacy and safety have to be baked in from the outset, so we have to develop that trust with the public. How can we give the social licence to operate with the public and develop that trust? What do you think the public really want from a digital transformation of public services, or are we doing what they actually want? Is there a balance to be struck between digital transformation of public services and retaining the public trust?
Rachel Coldicutt: There is lots in there.
Dr Gardner: Yes, sorry.
Rachel Coldicutt: There are two difficult things for the Government to think about, and one of them is that most people are not thinking about government most of the time. It is sort of like your difficult ex who you are not thinking about. What we see in the blueprint is an idea that government wants people to be thinking about government a lot more than they are.
The way that you win trust is by things working. Very often, the problem is that when something works we do not really notice it. We only notice things that do not work. Every time you are able to get a call-back from the doctor that day, that is great, and you don’t really think about it and don’t notice it. Every time you are sent down a pathway of signing into things, agonising and waiting, it becomes problematic. The important thing is to be delivering things for people that work when they are most vulnerable. Unfortunately, that brings with it a lot of complexity.
To give you an example of something that I saw at the end of last year, I was at an airport in Europe, and there was someone ahead of me trying to get their QR code scanned and it was not working. We were in a place where there was a flight out of the country every other day, and if it didn’t work it didn’t work, and there was no failover. There was no, “If this hasn’t worked, here’s another thing and another thing.” If we want people to trust things, we need to develop things with them.
Equally, we need to understand that people’s lives are hard. We do not want to be constantly creating new work for people to do to understand how government works. To Richard’s point earlier, we need to design services in ways that help people to understand how they work, so that, for instance, if you have to wait for two weeks to hear back, there is messaging. I worry that a lot of the political messaging at the moment is that AI can just sort it all out, make everything magical and all the problems go away. The likely gap between the ease that is being talked about and the reality of juggling things is not very glamorous. That may be why it is hard to address.
Dr Gardner: You make some very good points. I will pick up on one thing you said. It is great to have these technical solutions but we need to be careful, so that if they fail we build in quick, easy and effective back-ups.
Q31 Chair: Thank you very much. Rachel Coldicutt, I am struck by your statement. It is true that the Government expect people to think about government more than people want to think about government or think about government. The fact is, though, that people have—I know this from my constituents—a sense of a lack of agency when it comes to government and technology. Technology and government together are a combination where people feel particularly powerless. For the effective digital transformation of government, we need that to change. The new GDS says it wants people to feel that they have a voice, that they are listened to, and that they can play a part in shaping the services they rely on. You said how important it is that services are made with people. How can we get people’s voice as part of service design so that the services that are delivered are what people want?
Rachel Coldicutt: One of the things that is very underrated is that, when you make a service of any sort, you collect lots of data about what people are doing and not doing. We have definitely seen that the GDS was able previously to optimise based on how things were used. The first thing to do is to understand the data that government already has about things that are not working. It can feel like a very good thing automatically to default to public consultation, but actually lots of the answers are there, which means that when you work with the public you are asking them the things that you really need to know and not asking people to go on a journey of zero to 100 in making it work.
The week before last, I was helping to run a citizens’ assembly in Liverpool. To your point, one of the things that we saw there is that the granularity of running a service and the jargon and the technicalities become very overwhelming to the people running those things and delivering them. When you get someone who is working in the weeds on a programme coming and talking to members of the public about how that could be better, they are likely to start talking about things that people have never heard of or never thought about. Actually, people have much more developed ideas of the sorts of experiences that they want at quite a high level. Most people want accountability and those sorts of things. The trick is not to waste everyone’s time by asking general things that very often never get delivered on, but to ask specific things and delivering against those and working with people to evidence them.
Q32 Chair: Right: specific use cases and indeed specific pilots of specific services that the public can engage with—not to put words into your mouth, but I think that is part of what you are saying. What do you think of DSIT’s recent digital inclusion action plan?
Rachel Coldicutt: It doesn’t seem tremendously different from where we were in 2014, and it is not very imaginative. There are a couple of things in there. On the one hand, if you are not a very confident user of technology, I imagine the last thing that you would want is an old Government laptop that someone had to operate with the help of IT. That is the kind of thing that sounds nice, but probably doesn’t help anyone at the heart of it.
The thrust of it overall is making the people work for the service, not making the service work for the people. I would like to see a much more contemporary engaging with understanding of what technologies look like in people’s lives. If you are a person who for whatever reason feels alienated and under-confident, and if you do not want to do pension credit online, the thing that will incentivise you to use technology might be talking to your grandchildren, not filling out a form.
We have to understand that sometimes digital will not be the best route for everything. I would like to see something that is more about extending the fun parts of what the digital world offers. If we want to include everyone, a really good place to start is to look at the minimum digital living standard, which looks at what a good digital life looks like and how more people are able to have one.
Q33 Adam Thompson: Joe, I have a couple of questions for you. First, how do you see the role of the Cabinet Office and HM Treasury in delivering the new GDS?
Joe Hill: There are plenty of ways that both of those organisations have to be involved, because the kind of transformation we are talking about involves big cross-government policy levers that the Cabinet Office tends to secretariat and control, and hopefully will involve good decision making about the public finances to make investments and to make savings like those we just talked about.
There are dangers for both of those organisations at both ends of the spectrum, and hopefully we will find a new middle ground that delivers what is best. To expand on what I mean, at different times both the Treasury and the Cabinet Office can be quite overbearing for Departments; and there is a huge amount of process checking. I know from personal experience as a former civil servant on both sides of this that you can get to the point where there are far more people across Government who are trying to mark your homework on a digital project than there are trying to help you deliver the digital project, and at some point you think, “Where is the help and support?”
At the other end of the spectrum, you can find that those bodies are conspicuously absent on some of the biggest decisions. If we take investment as a key example, the Treasury is about to do a spending review. I was involved in several of them when I was at the Treasury. The public finances budget for technology is run in almost the opposite of the way that most private businesses would do it. Most private businesses try to control their long-run run costs, and they are a bit more flexible about change and transformation costs, because you can always reduce your transformation spending if you need to, whereas Government tend to take the opposite approach.
Departments doing a spending review submission will work out what their running costs are, how many staff they want to have and how much they think they need to pay them, and then they bid for a whole load of other stuff. When they get the Treasury response in, they tend to say, “How many staff can I actually afford, and at the margin what little bit of transformation do I need?” It is the complete inverse. I don’t think the Treasury spends enough time saying, “Hang on, we’re talking about a £10 million pilot here or there, or a bit of investment. Where is the real cost baked into the system and how can digital help us transform that?”—the big handfuls of expense in public services.
Q34 Adam Thompson: Thank you, Joe; much appreciated. As the Treasury and DSIT go through the process of reforming the Government’s digital funding approach, what do you think that the new GDS should be pushing for within that? Very specifically, what is your understanding of the concept of digital funding?
Joe Hill: I might take that part of the question first, if that is all right. Digital funding should mean what it says on the tin, which is how much we are spending on digital technology in both the existing infrastructure that we have and the kind of transformation that we want. There is a bit more of a movement at the moment to formalise mapping knowing what we spend digital money on. Some people have recommended creating a whole new spending category to capture digital spending in particular.
Q35 Chair: What do you understand by “digital spending”?
Joe Hill: The spending that the Government have on the core technology infrastructure that they use to deliver their day-to-day business, such as cloud compute, some old on-premises compute, the subscriptions that they are paying for applications that they run on that, common components like database infrastructure and, crucially, spending on quite a lot of staff who are involved in building, maintaining and running those services. That will comprise the whole gamut of different kinds of staff. Previous witnesses talked about a few of those people. It is a big category depending on how you define some of those people and particularly very large swathes of public spending. It is not particularly comprehensively mapped. Exercises to do that to the nth degree have been done over the years. They are always very stochastic. Every few years, someone says, “Oh, we don’t really know. Let’s do that again.” It takes them a very long time to do it, and they think, “Oh, God, we can’t keep doing that every year,” and they probably should. We should probably have a common sense of how much is being spent on it.
You asked about the move to change it and transform it, and what DSIT and the Treasury should be looking at. We have researched and written a bit on this topic. The initial direction from the Treasury in its review of digital spend that it published two weeks ago is really promising, but we would like to see it go further. The key thing it needs to think about is a genuinely risk-based approach to technology. There are very large amounts of Government technology spending locked into very long-term change programmes. It takes a year to write the business case, another year to get through all the technology approvals, and a year to procure from the company that is going to come in, and buy it. They come in and find that there are no in-house staff to help them do the work, and suddenly you are three and a bit years in. The technology has moved on in that time. All those things bake in a huge amount of spending.
The main reflection that I have when I talk to former colleagues of mine at the Treasury, having now worked outside it on technology projects, is that how expensive the Government think technology is will be much higher than it should be. There are lots of ways that the Government can do it more cheaply and more effectively, and often that is getting out of their own way on some of these processes. The reason we have all these really overbearing checks and balances is that we have created something through the checks and balances that is very expensive, very time-consuming and then inevitably goes wrong. That risk-based approach is core.
The only thing other thing I would add—it speaks to something that Richard talked about a bit earlier—is on taking a different view of the transformation of digital services and the ongoing run cost of the current digital services that we have built and maintained. Often, you find that both of those things are lumped into a programme with a whole bunch of other stuff in a way that does not make it easy to interrogate; people in departmental finance teams and the Treasury are routinely getting business cases to sign off where they ask for it to be explained, and they are told, “You couldn’t not sign this off because it includes a load of the ongoing run costs of a big bit of government,” in which case you think, “What’s the point of me signing this off?” If we change both of those things, it would be a really good and positive sum for everyone, but it takes the Government getting out of their own way a bit.
Q36 Dr Gardner: I am smiling because your description of what it is like to do these things rang a few bells with me. Rachel, I want to talk about the Government kick-starter initiatives. We have some quite interesting ones: the Government app and wallet, the LLM chatbot, the upskilling programme, the vulnerability scanning service, and one that is a bit vaguer, which is piloting improvements in how to better manage long-term health conditions or disability in collaborations across Government. That was flagged as being a bit of a vague goal. From your experience, when developing or rolling out these kick-starters, what lessons should the new GDS learn from the roll-out of new digital processes and tools elsewhere, particularly within the collaboration across Government Departments on issues such as health?
Rachel Coldicutt: There is something really interesting, in that there is a tendency in many of the proposals to think very big. It is important to remember that cutting across lots of things is not the same as being big. Making specific end-to-end journeys work for people who need services to join up across DWP, Health and Education may or may not be extensible to everything else. There is an idea that, with the time and the money that we have, government can be turned into a sort of galaxy-brain kit of parts and that everything will be really easy. To my point earlier, if we are specific about making specific journeys that are currently problematic for vulnerable people, and create lots of latency in the system and those work, we can build up from them. It is much harder to say we are going to make everything better for everyone and then deliver those edge case needs.
Both Richard and Laura previously spoke to the fact that these things are all okay, but they are quite vague. If the point is to feel that progress is being made, we could have a much more granular and specific list of things. In terms of understanding how they are working, there is a huge role for qualitative research. The thing about digital change is that it can sometimes take years to show in aggregate. If we are looking for metrics to be everything going up on the graph, it might take years and years for something to be rolled out to deliver that sort of scale, whereas being specific and being led with qualitative understanding and then evidence to roll out more could be a very powerful approach.
Q37 Dr Gardner: Thank you very much. It is quite nice to look at some examples of the difficult-to-solve journeys and experiences of people. That is really interesting. Broadening that back again in a way, Joe, do you think the kick-starters are ambitious enough in their remit?
Joe Hill: I agree with Rachel. They are very large in their remit in many ways, but I am not sure they are ambitious enough. The common component to most of them, except for the health conditions and disability one that you mentioned, which we can come on to, is that they are broadly all levers that DSIT itself owns and largely controls. The gov.uk app and wallet need some data and support from other Departments, but broadly they are platform components of what DSIT does that it can spin up and control.
I would like to see more ambition and more focus on digitisation, automation and getting efficiency savings and productivity benefits in the big handfuls where the Government cost a lot at the moment, and those are not often things that DSIT can directly control. They are in big frontline services, whether those are run by central Government Departments, public services or local government. They tend to be the big caseworking functions and the big parts of frontline services. With the exception of that one, which I agree is quite vague and quite high level, we do not really see that come through. One of the reasons you probably do not see it come through is that those tend to be seen as the responsibility of the DWP, HMRC, the Home Office—the Departments in charge of those—but that is really where the biggest bang for your buck is going to be, and that is where DSIT should be exerting more of a role if it is to be the digital centre and trying to drive change. It can do that in collaboration with Departments and with frontline public services. To add up the big numbers and the big ambition in that document, that is where you need to look.
Dr Gardner: In a way, it is not to be so guided by the shiny, new digital products, but to look at the more fundamental, slightly more boring but really transformational issues. Thank you.
Q38 Chair: Thank you very much for that, Allison. Reform has suggested six specific use cases. Is that what you are referring to, Joe Hill? They include updating basic tasks on gov.uk and providing AI chatbots in call centres.
Joe Hill: It is a combination. It is those particular use cases, although you would find many of those use cases in most Government Departments, with a particular laser focus on where you find them in the big, very expensive frontline public services in Departments. There is often a temptation in government to announce or to pilot, often because you have a deadline to meet when you need to do the announcement or do the pilot, in something that feels easy—often a new area or a very small area like some of the use of AI in particular bits of policymaking. While those are often the easier things to do, they are not necessarily the biggest, hardest and most important. I would like to see a little more focus there.
Chair: Great, thanks very much. Martin.
Q39 Martin Wrigley: Thank you very much. Sorry, I was getting carried away in the answers. Looking at how GDS is bringing all these different areas together and the challenge of implementing common standards across local government, central Government and other bodies and bits and pieces, where should they be going, Joe?
Joe Hill: They should be going where there is a combination of the biggest wins to be had and where they think that in the short to medium term they can make the most progress most quickly. The Government are often very tempted to boil the ocean. It feels like in the press release they need to say something about everything or they are going to do a little bit for every Department. Actually, that is a really common failure mode of big Government strategies because you try to do everything everywhere all at once, and none of it really happens. Some of those things will be big caseworking functions in places like the Home Office, DWP or HMRC, where it is going to be uphill and it is going to be hard, but if you can make casework or productivity a bit higher and if you can clear backlogs faster, you stand to save the taxpayer a lot of money.
We also see some of those functions mirrored in NHS trusts, police forces and local government. Central Government do not have the same kinds of levers over those organisations to compel them to change tack, and they need to think a bit creatively about how they do that with investment, deal-making and bargaining. The digital centre should set up its approach so that it can find quick short to medium-term wins, action those very rapidly, and move on to the next one.
Q40 Martin Wrigley: Do you think they are doing enough in terms of identifying commonality or differences in data? We heard not long ago how hospitals measure missed appointments using the same notation for multiple different activities, so you cannot compare data from across hospitals, let alone from across departments and across other systems—the data language, the data lexicon, or the way things talk to each other and understand what they mean.
Joe Hill: Standardising data and the availability of it to access through common APIs across Government was the biggest win to be had seven or eight years ago and still today is probably the single biggest win, although the state of that landscape has improved a bit. What is tricky sometimes is that leaders in public services want to be using AI and they want to be seen to be operating at the absolute frontier, and then their technology teams tell them, “You’ve got to fix these things first.” Sometimes you can do both and you can work quite iteratively, particularly if you have good technical capability in-house and you are not entirely reliant on procuring it with the long timelines that we just talked about, but you need to do some of the fundamentals. A bit like Richard said earlier about legacy technology, you cannot see it as a “Done, finished, we’re never going to have to do that again.” You will always have something that feels like legacy technology. You should always be trying to update it. Very few other companies—certainly not technology companies—budget for, procure and think about their tech stack and their data in the way that central Government do. They have all moved on to a much more agile way of doing it.
Q41 Martin Wrigley: Where do you think GDS should start with procurement reform?
Joe Hill: There are probably two sides to that. The first incarnation of GDS was the starting point for government building much more in-house digital capability and not always outsourcing things to the private sector. We have come a long way since then. I still think there is a way to go on that. That may feel like a tension of interest between those two parties, but I don’t think it is. The most frustrating thing for private companies that come in and want to work with Government, want to build good services, want to get paid and get a reputation for being good is when they come in and find there are no in-house people who can answer any of their questions about how to plug this in and there is no knowledgeable sponsor to work for. The first part of being better at procurement is continuing to build all those digital skills internally to help work with private companies.
The second bit is just shifting the dial and bringing Government procurement practices for tech up to date with the rest of the world. There are basically two components of that, one of which is being more willing to use software as a service, to buy software, to buy a licence fee, rather than to buy a technology company to come in and build you one bit of tech now and you are never going to have to replace it. Unfortunately, although I signed off lots of those business cases once upon a time, you are always going to have to replace it. It is kind of a myth. Most of the world has shifted to buying either whole applications or, more commonly, parts of its digital infrastructure from private companies, using its own in-house expertise and putting those together. Government are not set up to do that at all. Government buy very little software as a service. There are procurement frameworks that are specifically designed to do that and give them much more opportunity to try things before they buy and to have periods when they are testing and innovating and can ultimately say, “Actually, this doesn’t work for us.”
The final part of it that is crucial is incentivising a bit more competition in these markets. Much like the rest of public procurement, in large parts of Government tech procurement there is only one bidder for the contract, or no one bids for the contract. In no sense are these things competitive. We focus on the ends of the market—really big technology companies, Government strategic suppliers and really small businesses. Often, a lot of those end up being sole contractors doing bits of consultancy. The Government need to think about how they create markets and contracts for and grow soon-to-be ex-SMEs. We do not have enough focus on people who are at the top of that and who could one day be competitive and useful on the big bits of Government work. There is a real hard cut-off where it feels like you are trying to move on to much bigger frameworks. You cannot demonstrate the size and stuff. It is that step change that you need to overcome.
Q42 Emily Darlington: Rachel, we have talked a lot about the agenda in terms of the efficiency for Government or the public experience, but actually it is a huge culture change for civil servants. What, in your view, do we need to do? We keep hearing that the Government are talking to the unions. What role should the unions be playing?
Rachel Coldicutt: Something that I have been really struck by over the last few years is how normalised it has become for people in technology roles to say that a service is not fit for purpose. That seems to have become a new normal. There is something in that about there being two fundamentally different models of doing things. Very often, in all the digital transformation work I have done over 30 years, it is not about technology; it is about people. It is about working together and building trust. That never happens if we begin from a place where some people over here are saying that the people over there aren’t able to do the job. There is something about how the culture of digital transformation within government might be more sympathetic. I know that that is hard when innovation is deemed to be important and large amounts of money have to be saved.
In all the digital transformation work that I have done, I generally find the people who know how to do a job will be the ones who say immediately, “These are the ways it could be faster and easier. These are the things that drive us up the wall.” Rather than not asking those people and working around them, which can often take longer, can there be a more collaborative way of understanding that? I have for 15 years been in loads of meetings and workshops and other things where everyone is talking about the number of jobs that a technology is likely to displace or change. These things always happen much more slowly than whatever the marketing deck that you are being shown appears to indicate. Rather than thinking about work that I have done previously in healthcare as always being about efficiency, it is better to think about it in terms of outputs. If we think about what a good service looks like, ultimately, there is nobody who wants to stand in the way of that. When you make it about efficiency, it becomes about people protecting their roles, which is not terribly helpful.
Q43 Steve Race: I have a question for both of you. DSIT has “A blueprint for modern digital government”. They are planning to develop that into a road map for digital and AI to be published this summer. What would you both want to see coming out of that road map? Secondly, what do you think DSIT needs more of in terms of skills and resources to deliver the vision that is coming out?
Rachel Coldicutt: Looking at the blueprint, what I am really struck by is that it feels a little bit like something that has been produced in response to the question, “Can we have a vision?” It doesn’t necessarily feel like a list of all the jobs that need to be done. I worry that, particularly around it being published, we have equally had a lot of political speeches that have indicated places where AI might be rolled out, and other things have happened that begin to make the task list of the people in the digital centre of government become very long. The possibility of doing all those things becomes harder. To repeat myself, it is about specificity. The idea that government might wrap itself around individuals with more personalisation is very vague, and it could mean almost anything. I would like to see specific examples of use cases.
AI might not be the answer. That was talked about earlier. I might even recommend that innovation be something that sits slightly separate from delivery. We have heard lots about the fact that maintenance is hard. It is very difficult to get money moving into things that are already working. Possibly, the danger of the plan that is about to come is that we might see more resource going into newer things rather than keeping everything moving.
I hesitate to speak to the resource that is needed, but I do not think that Government need to build a behemoth. It says in the blueprint that we could be looking at 1,000 people working at the digital centre. We know that agile, iterative working is the best way. Rather than making it about consolidating everything, it would be useful to see scaling up where other activity is happening as well.
Joe Hill: Absolutely. I completely agree with all of that. On the question of what to add, it is so tempting to keep adding to these things. The thing that I would like most is something that is simple, straightforward, articulates their vision and doesn’t feel like hundreds and thousands of civil servants’ time went into it. I would much rather what they were doing was demonstrating how far they have come since February. Lots of this stuff is very quick to build and quick to do. The amount of time that gets spent on big Government documents for a press release often detracts from the importance of what they are trying to talk about. That would be my first point.
Having said that, there probably are a couple of things it would be great to see. I would like more focus on Government doing things quickly in technology, more focus on speed, and particularly more focus on getting things to scale quickly in government. We use the word “pilotitis” in our research—an interviewee originally said it to us—to describe the phenomenon in government where people release strategies that mention technology projects or come to Committees like this and talk to parliamentarians about technology projects. If you mine under the surface, it is a three-month bit of software that someone cobbled together and there is a slide deck, and it never goes anywhere because the budget never continues, whether it was good or bad. These things come and go, but the majority of Government technology does not really change very quickly. I would like to see the Government try to buck pilotitis and get much more quickly from, “Is something working?”, to, “Can we scale it up?”, or, “If it’s not working, can we just cut it off and end it?” That frees up money, resources and time for the things that matter.
The other one that we have highlighted a bit in our research, and which is controversial but I would like to see the Government taking on board, is a much more risk-based attitude to the adoption of some of these technologies. There is a thing that people say in the civil service and Government, which is that Government are very risk-averse about any transformational innovation. I actually do not think that that is true, because the Government are clearly willing to tolerate very high levels of failure in their current public services most days of the week. Open any newspaper. When we talk about technology and automated protocols that I have worked on, or bits of projects, where we are constantly shooting for the 99th or 100th percentile of performance of a technology in a task that probably human public servants do today, I wonder whether that is a fair like-for-like comparison sometimes, particularly if the technology is much cheaper and much easier to roll out in public services where we have huge recruitment challenges.
That is a difficult issue for the Government to reckon with because it is to do with public trust and confidence, but in some of those areas, the difference is not whether the public would like to see a public servant or have their query processed in an automated way. The difference will sometimes be whether it is processed at all or whether they get no service from the state. I would like to see the Government updating some of their guidance and attitudes on how they think about the risk and performance particularly of automated technologies versus how public services work today.
Q44 Chair: Thank you very much. Rachel Coldicutt, I was struck by what you said about the critiques of the civil service as not being fit for purpose when it comes to technology. Following on from Steve’s question, what would you like to see in six months’ time in the road map that you felt demonstrated that the civil service had been brought on board and was valued as an agent for technological change?
Rachel Coldicutt: There is lots of space to celebrate the things that work. It is very easy not to remember that UK Government digital services are still widely regarded to be among the best in the world. We are probably accustomed to that now. We have become used to it. As I spoke about earlier, when things work they go unremarked. There is possibly an interesting comms challenge here about how DSIT can talk more about the infrastructure and the things that work and that no one has to think about. Bringing those a little bit more into consciousness will help people to get more confidence in the ability of digital government to deliver.
Chair: Thank you. Emily has a question. We will probably finish with that.
Q45 Emily Darlington: It is a good note to end on. One of the roles of this Select Committee is to hold the Government to account. What do you think are the measures that we should be looking at to evaluate the first six months of the new GDS? What kind of metrics should we be looking at and using?
Joe Hill: I can absolutely have a go. The key metrics are, first of all, whether it is helping to hire really talented people into government to do some of this who otherwise probably would not choose to come and work in the state. It is a bit of an unusual point. Salary, as we talked about earlier, is important as well as the branding of the mission and who you are going to get to come and work for you and how. Although it feels like a bad metric to be chasing given where we are at the moment, one of the best services GDS can provide is trying to hire some of the best people from the private sector to come and do some of those tours of duty in government. The other things would be the measures they are starting to put in place for frontline public services, the technology there around the speed with which things are going to be built, the scale at which they are going to be deployed, and the relative performance I talked about earlier. What performance are we willing to tolerate compared to the current performance of that public service rather than absolute standards?
Rachel Coldicutt: I wonder as well. Six months is not very long. As we talked about, there is a very ambitious slate and things keep being added to it. I have two conflicting thoughts about it. On the one hand, what is measured becomes the thing that is important, and people focus on that at the expense of everything else—certainly, some of the more complex and visible things that might be a bit boring. I would caution against prioritising big metrics and asking people to come evidencing huge savings and reforms quickly, and perhaps use the six-month benchmark as a way to understand what feels useful to measure inside the team. Very often, when you start making things, what you expect them to do and how you expect them to work are not quite how they end up working. Proper digital agile accountability might take account of that in a slightly different way from what we are accustomed to.
Emily Darlington: Thank you. Those are very considered responses.
Chair: Yes, and a very good note on which to end, with a considered response and a considered assessment of where we might be in six months’ time. Thank you so much, Joe Hill and Rachel Coldicutt. It has been very informative to have you before the Committee today, and we thank you very much for your contribution.