Danny Buerkli

The Case Against Government Reform – with Andrew Greenway

Danny Buerkli: My guest is Andrew Greenway. He is a co-founder of Public Digital, a global consultancy specializing in digital and AI transformation, state capacity, and institutional reform. He has also written several books. The most recent is just out, and it’s called Digital Sovereignty: The Power to Decide. Andrew has been doing, I believe, some of the most interesting thinking out there on state capacity, including in the US, together with Jen Pahlka. Andrew, welcome.

Andrew Greenway: Thank you very much. Great to be here.

Danny: Andrew, how did government reform suddenly become a mainstream subject?

Andrew: That’s such a good question to start. I would love to believe it’s a mainstream subject. I feel it’s certainly a lot more than it was, but it remains relatively niche. But how did it become an interesting topic? I think it started to dawn on the political class as a whole, from left to right, that — hang on a minute — a real diversity of our ideas here aren’t working. So perhaps it’s not just about the what we’re trying to do here. Maybe it runs a bit deeper, into the how. And I think that realization has come, a bit like bankruptcy, gradually and then suddenly, and in different ways in different places. In the UK, for example, a couple of fairly major national crises: things like the Post Office scandal, which listeners may be aware of, where a flawed computer system sent quite a lot of people who were innocent to prison; things like the infected blood scandal, where people were given infected blood products by the NHS and then the state was unable to account for that, or apologize, for many years.

So these crises of state capacity, I think, brought this to life for some people. And then I think another part of it becoming mainstream was that general sense of political promises not translating into actions and effects that people could feel or see — growing in obviousness, and then weaponized by some parts of the political spectrum, who took that sense of frustration, and in some cases anger, among the public and have made electoral success of it.

Danny: What got you interested in the how of government?

Andrew: It was the cause that picked me, really. I had no intention of setting myself up as someone who cared deeply about civil services and how they work. But it captured me, I think, when I went to work for the Government Digital Service in the UK, which was 2011, 2012, so quite a long time ago now. That team was set up as a new institution in the center of government, and broadly its job was to make government better at being on the internet.

Before that, I was a very conventional civil servant, really — a classic dangerous generalist, which is another thing I’ve written about quite a lot. I did lots of classic young, generalist, ambitious civil-servanty things, like working in a strategy unit and working in a private office. And then I went to join this team, the Government Digital Service, and I met a bunch of people who were not just excited about the internet, but excited about what government could be if the tools and techniques and attitudes, if you like, of modern organizations were applied to delivering things that made life better for citizens or businesses.

And suddenly I saw stuff happen. They delivered things. I loved things that I could show my parents and my friends — new services, websites and stuff — where people were like, “Oh, that’s surprising. That works,” which is as good as it gets in government, as you know, Danny. You don’t get people who are wildly excited. People go, “Oh, okay. That wasn’t terrible.” That’s great. And I suppose there was the combination of being close to that experience of actually delivering tangible things and seeing that it was possible.

But then I suppose that combined with the frustration I remember feeling as someone who had been in the private sector briefly, in consultancy, before I joined government. I remember feeling in the civil service that the quality of the individuals I got to work with, in all sorts of departments, was exceptional — often higher than what I experienced in corporate life. But they were so much less than the sum of their parts, whereas the private sector often makes more than the sum of its parts, I think. I just thought: this shouldn’t be, this doesn’t have to be. And although quite a lot of the ways one could improve government and its effectiveness are quite difficult and sometimes quite uncomfortable, they’re not necessarily complicated or hard. And since then, I’ve found myself banging the drum for doing more of them.

Danny: Let’s establish quickly what’s at stake here, because clearly, as you’re saying, this is not just about prettier websites — though we all love pretty websites that work. You had a good line in a recent Substack post where you wrote: “If we remain attached to a bureaucratic method and culture that is unable to respond, and in which power and faith continues to ebb away to unelected actors” — and this is the key bit — “I do not foresee an especially happy future for democratic public life in the UK.” Can you unpack that?

Andrew: The British tend not to be a particularly revolutionary people in the violent sense, but we do like a good moan, and that can sometimes coagulate into something quite unpleasant. One of the things that gets me quite exercised about this decline in state capacity is the loss of hope, and the loss of belief and agency to some extent. It’s quite interesting, I think, that our new Prime Minister, Andy Burnham, has been talking up hope quite a lot in his public remarks since he began. But I do think there’s a quite deep danger in people giving up on government, because that leaves space for slightly darker, less democratic forces to move in. Already, in both the UK and other parts of the world, that’s leading to scenes that I think are quite ugly, particularly around tricky topics like immigration, but also on a broader spectrum too.

I’m quite intrigued, actually, that in the modern era we have not yet — and let’s hope that we won’t — descended into the degree of violence that you saw around state action in the ’60s and ’70s, even. It’s interesting now: there are riots and protests taking place in France and Spain. So there’s a bubbling up of more of that kind of direct action. But I feel the competence, and the perceived competence, of the state to make a positive impact on people’s lives is one of the root causes behind a lot of those kinds of frustrations. And to some extent — again, speaking from the British perspective — we like moaning. I’m not proposing that there’ll ever be a future where we’re delighted by our government and its all-seeing competence. But I would like us to retain that belief in the state to do positive things, because without it, I think we risk taking certain things, and certain democratic norms and processes, for granted, and that potentially takes us to quite a dangerous place.

Danny: One of the better-known recent government reform attempts was DOGE, the Department of Government Efficiency. What did they get right?

Andrew: I think one of the things they got right, which is semi-controversial at least, is that they came in with a very clear sense that in order to make government work better, you needed to clear some space. One of the challenges I’ve always faced when talking to friends and colleagues in this space, particularly those who lean towards the innovation end of the spectrum, is that a lot of the thinking about how to improve government is quite additive. It’s like: if we add on this team, or this task force, or this initiative, or this new shiny thing on top of the layers of institutions that we’ve currently got, that will make things better. And to some extent they do, but I always think their impact is quite limited, because government is already a hugely complex system, across which the cost of collaboration and coordination is extraordinarily high, and adding new things, however good and well-intentioned they are, only increases that cost. So one of the things I thought was interesting about DOGE’s theory of change was that it came from the perspective of: in order to make government better, we need to clear some space, we need to close things down, we need to get rid of people. And while I think that was generally done in what appeared from the outside to be a relatively indiscriminate way, and perhaps quite a partisan or ideologically driven way, I still fundamentally think that building state capacity will end up being a process that is creatively destructive. It will require taking things away from what the state currently does — be that processes or institutions or possibly people — and putting things in their place. And I think it’s interesting: DOGE clearly had that theory of destruction woven into it from the get-go.

What I was never quite clear about is whether it had a theory of creation as well. That was certainly not obvious to me from the outset. And I think if you’re just destroying things, and doing that in a relatively indiscriminate way, that’s unlikely to make stuff better. I’m fairly uninterested in questions about the size of the state. I don’t think it’s a particularly interesting way of looking at the problem. I’m much more interested in whether the state is effective, from which I think efficiency follows. An efficient state is not necessarily an effective one. So I do think it got that element right, but it was very much only one side of the coin.

Danny: One of the elements that I find complicated about the debates on government reform is that many people disliked the DOGE approach, for reasons one can understand. But clearly, the very gradual, very-high-in-supposed-thoughtfulness, include-every-stakeholder-imaginable type of reform manifestly did not work. And then the question becomes: what is it then? If it’s neither of these things, what is it? I understand the rejection of the DOGE approach, but I think, as you say, there is a reason, presumably, why that theory of change got some traction. And I think it’s hard to understand what’s happening unless you recognize that.

Andrew: Yeah, I completely agree with that. And I think there were two other things about DOGE which are quite interesting and also have echoes in other successful initiatives to improve state capacity and the state’s ability to deliver. One was the skill sets that they brought into government. They brought quite a lot of technologists in. It was not your typical policy-wonk type of character. Now, DOGE clearly did that to an almost self-parodying level. There were stories of very young, very arrogant tech bros going around saying, “Oh, we can vibe code a tax system in three months. Don’t know what the problem is here.” So there are obviously extremes to which that’s not sensible. But it seems inherent to me that increasing state capacity in 2026 and beyond is clearly a technology problem as much as a policy problem, as much as a design problem. That shouldn’t be a radical or particularly interesting thought, but I think it still is in government circles, where policy tends to rule the roost. So I think that multidisciplinarity of DOGE was quite interesting and, again, has echoes in things that have worked in the past.

The other thing as well — and I’m really torn about this point — is that there’s something about reforming government where it’s quite hard to do if you’ve been in it before. You have to not know how difficult it’s going to be. I wrestle with this myself quite a lot. Would I ever go back into the civil service as a civil servant and come and do this again? It’s really hard, because I know exactly how awful that would be personally and professionally, because they’re extremely difficult jobs. You’re often pushing boulders uphill, inevitably, in a lot of those change roles. It’s a very difficult balance to strike, that mix of insider and outsider, where you know enough to navigate the system so you’re not stepping on all the little landmines that outsiders coming into government often will. But equally, you come in with a degree of — call it innocence, naivety, freshness, whatever — where you go, “Why on earth are you doing that, actually? That doesn’t make any sense at all.” Managing to get that mix right is incredibly difficult. But the idea that DOGE brought in some outside thinking, and the idea that state capacity requires some of that turbulence from outside — I suspect there’s some truth in that.

Danny: This is the old and fun saying: we’re doing this not because it’s easy, but because we thought it would be easy.

Andrew: Yes.

Danny: Speaking of making space for new things, you had this fun idea of a Thanos snap: the idea that if you snapped your fingers — you specifically, Andrew Greenway — and a third of the current top 200 UK civil servants disappeared at random, that would make things better. Explain that.

Andrew: Yeah. I do believe that quite strongly. That wasn’t just a rhetorical device. And I said “at random” in part because what I don’t want to suggest — as I think you get a sense of at the more populist end of this discourse — is that senior officials in any country are these corrupt, blocking individuals who are deliberately being obstructive to either democratic will or the reform of the bureaucracy. I don’t fundamentally believe that at all. The vast majority of senior officials that I meet and talk to about things like this are, if anything, more frustrated by the context and the system they find themselves in. But they all equally find themselves with a very clear professional calculus and set of incentives that strongly encourage them to believe and behave in a similar way, which is to maintain the status quo.

So the point of getting rid of a third or a half or whatever with a Thanos snap is more about that previous point about complexity: government spends an inordinate amount of time making sure it agrees with itself before it can get anything done. And to some extent it has to, because it has to broadly cohere, make sure it’s not doing things that are flat-out illegal, and be able to tell a clear story about what it’s trying to do. That just gets harder and harder the more leaders you have around. I was looking just at the top 200, which is possibly slightly cruel — you could pull the span a bit wider. But the challenge, I think, is that for a lot of those very senior people, their job is to be accountable. And by that I mean to deliver the performance of accountability within the system.

So these senior officials — permanent secretaries and director generals, in UK terminology, the top two layers, if you like — what they are brilliant at is sitting in front of parliamentary committees and explaining why things are actually okay. “It may look like that’s gone wrong, or that we’ve done something unwise here, or there’s no clarity from the department on this, but it’s fine. It really is fine. We promise you.” And you watch some of them do that and it’s flabbergastingly good. You feel better. It’s not better, but you feel better. It sounds like I’m being flippant and unfair. I absolutely think there’s an immense skill in being able to do that, and it actually is quite important, because that’s the game they’ve been set. Those are the rules. That’s the thing they’re expected to do, and have to do, to be competent in their job. None of that has anything to do with delivering stuff and improving outcomes at the level of public services or infrastructure projects.

So I think we have this overpreponderance of skills in that particular direction. And to that point that it’s not just about destruction, it’s about replacement: if we did have the Thanos snap, what are the senior leadership types and shapes that we would replace those people with? I think they are probably people who would be terrible in front of a select committee, and that’ll be fine. But they’re really good at running a big infrastructure project, or they’re really good at developing a public service. Finding some of that balance, I think, is the key.

Danny: It reads to me as essentially a way of inducing an exogenous shock, which does tend to help. There’s probably a reason why the Vaccine Taskforce, or Operation Warp Speed in the US, worked so incredibly well. There are several, but one is just the urgency and the exogenous shock and the obvious, overwhelming need to do something useful. And the other element is probably just that it was quite clear what success looked like. Right? That also tends to help.

Andrew: I think there might be an implication people take from what I’ve said, which I don’t intend, which is that it implies a degree of politicization — in the sense that you should give ministers more power to swap senior officials in and out. On the one hand, I definitely think there needs to be a degree of mutual trust between those two sides, which I don’t think there always is, and which has probably diminished in recent times. But whenever I talk to friends who are in the States, or in more politicized official systems, they immediately blanch when you suggest that politicization could be more of the right answer.

And I’m pretty skeptical that it is. As I say, what I mean is much more that sense of not holding quite as tightly as the UK system certainly does to the idea of permanence, where if you’re an official, particularly a senior official, you rise by virtue of the fact that you’ve spent a long time in the system, and therefore you understand it — its small-p politics, its characters, and its rules, which are often implicit, not written down. You understand those things implicitly, and you professionally succeed off the strength of them. And again, those skills, while important and valuable to some extent, are not the same as those that deliver better outcomes.

Danny: If you could, which government institution in the UK would you most personally enjoy replacing?

Andrew: The Treasury. Absolutely the Treasury. And I’m not unique in that respect. People have been arguing and moaning about the Treasury for well over a century now. I suppose this is the economist’s view of the Treasury: it’s quite an unusual department when you compare it globally, in that it has this role as the finance ministry, but it’s also immensely powerful in terms of economic policy and in terms of sheer political weight. It’s not a CFO figure. It’s much more than that. I’m really interested in the Treasury in part because of its disproportionate influence, but not so much that. And I certainly don’t object to the fact that there’s someone in the system that has to say no a lot, because you need that in the system. You need that ability to be the voice of fiscal rectitude. So it’s nothing to do with that. It’s the fact that the Treasury is this cultural wellspring for the whole of the rest of public service. Many of the senior leaders — or indeed, not necessarily the very top, but the influential people at the level that’s actually doing the work — have done a stint in the Treasury. And it exerts this enormous cultural effect on the character and personality of government.

Any monoculture is quite dangerous, I think. But certainly this one — as you know, Danny, I tend to go quite deep geek on this — there’s a fantastic book called, I’m going to say, The Private Government of Public Money, I believe it’s called. It was written by two academics in the 1970s, and it’s basically a piece of anthropology on the Treasury. It’s based on a bunch of interviews, I presume, and they pick out these very human, psychological characteristics of the Treasury, including things like: it’s elemental to the character of that department to not know too much. When you’re making decisions about whether to say yes or no to spending, God forbid that you actually deeply understand the area that you’re making the decision on. Because if you do, you might go native — to use the slightly inappropriate term — or basically be captured by the interests of that department, which means you can’t take the essential dispassionate view of whether this thing is wise to invest in or not. And I think that’s mad. There’s a quite intentional, deliberate gap in knowledge and insight and understanding when you’re making decisions. And as I say, this was written in the ’70s, and it’s just as true today. So figuring out how you reshape and unravel that particular locus of power would make a huge difference.

And one final thing I’ll say on this: there was a counterfactual I’ve always enjoyed playing with in my mind, and it was never a realistic possibility — whether the Government Digital Service, in its relatively early era, could have been put into the Treasury. It was based in the Cabinet Office. But what would have happened if you’d put that team, which in and of itself bore quite a lot of traits that were a bit Treasury-ish, in there? We had a very strong sense of culture. We had a perhaps excessively strong sense of ourselves and our importance. If you’d managed to harness those two — and they were quite different in the way they viewed the world and what they saw as important — if you’d brought those things together, I think you would have had a profoundly different center. I’m not saying it would have been better, but it certainly would have been different.

Danny: I asked about replacing rather than reforming because one thing you’ve written about, which I think is incredibly important, is the sheer impossibility of reforming certain organizations. It’s something I deeply believe, and I think it’s an underappreciated point. Many government organizations cannot be salvaged. It is not possible. Some can, but many cannot. And so therefore you have to have another way of going about it. The strangler fig pattern is one way of naming that. Say more about that. If you can’t reform it, what is the way to get around it?

Andrew: Yeah. As you say, there are several patterns to this, and it depends, I think, to some extent on the problem you’re tackling. In some cases you get lucky, and the state has to tackle a problem that’s genuinely novel and new, and you can set up a relatively greenfield institution to go at it. The paradigm example, I guess, would be something like NASA, which had a relatively clear field, a clear problem, and the space to do it. Then there’s the strangler fig pattern, which is a bit akin to what GDS was doing with the government’s digital estate: set up a new thing, slowly absorb the things from the old entities that it wanted, and then close out the old thing. I’m quite a big fan of that pattern, and it happens quite a lot. Interestingly, it’s happening more and more in the more sclerotic bits of the private sector, in places like energy, for example — you’re seeing it show up more and more. And I suspect it works quite well for government because the difficult thing about institutional renewal and displacement is that government can’t turn itself off. It has to parallel run to some extent.

And then there are the fight-from-within models. For example, look at what the UK did with Universal Credit, which was the process of reforming a large chunk of working-age benefits. This was done ten to fifteen years ago. That was driven by a team that lived within a legacy department — to use a shorthand — the Department for Work and Pensions, but was effectively quite strongly partitioned off, in a quite literal sense, from colleagues in the rest of the department. They worked in a different building; they often used different computers. There was some connective tissue to some of the underpinning data, for example, and other bits and pieces like that, but fundamentally they were fairly well walled off from most of the cultural and process affordances of the department.

And then there’s another model, which Jen dug into, really — Jen Pahlka, that is, sorry, my American colleague — which is the Colorado Digital Service model, where the Colorado Digital Service have effectively set up teams in different ministries and are using them to infect their colleagues with their methods and ways of working. So there’s no single pattern, nor, I think, is there necessarily a definite one-to-one match between a particular scenario and a particular pattern. But there are certain common qualities that they all share, and the ones that have endured most successfully are the ones where their moment of launch — their “hello world” moment, if you like — was not when they started something; it was when they stopped the old thing.

If I was looking across government organizations now, and the typical suite of ministries that a government may have, I would be looking pretty hard at benefits, where I think agentic flooding and AI are going to fundamentally unbundle and disrupt how that paradigm works. In the UK, I would be looking at things like the Home Office, which has been accused of being not fit for purpose on an almost five-yearly basis. But it’s possibly unfair to pick on particular areas. This is not a case of “this department is failing, therefore we need to start again.” The argument I’m trying to make — and that Jen and I tried to make in our piece in Foreign Affairs recently — is that this is an inherent quality of bureaucratic institutions. Whereas in the private sector you can usually rely on at least a functioning market to put the worst out of their misery and to create the new, you need to midwife that process more actively in government.

Danny: It’s a great piece. Everyone should read it. I’d love to play a game, if you’re down. The game works like this: I give you two names and you tell me who you pick and why. So the first pairing is Stafford Beer or Herbert Simon. Give me a sentence on which elements of those you’re picking them for. Stafford Beer: the idea of cybernetic organization. Herbert Simon: incredible foresight on what AI might do to government.

Andrew: I think I probably lean more towards Herbert Simon, but I struggle to justify why.

Danny: There are no rules to this game. You don’t have to…

Andrew: Okay. Well, I’ll stick with that.

Danny: James C. Scott or Francis Fukuyama?

Andrew: Francis has been very kind and said some very nice words about our book, so I sort of have to speak to Francis, really. Francis seems to be quite a strong advocate for the abundance movement in the States, and makes the arguments around it, which I think are compelling but possibly not the whole story, in that they seem to be relatively focused. They’re quite US-centric, I think, and they’re in that quite legalistic frame of sludge and getting it out of the way. I love James C. Scott’s book so much, though. It’s quite a hard game. Let me step back. In terms of who I’ve actually found most usefully provocative, I think it was James C. Scott. He talked about this concept of mētis in officials, which is essentially this implicit, almost folk wisdom that exists in layers of the state, which is actually what allows it to work. If it’s very rigidly structured, the structures tend to be very brittle — I’m doing quite a poor job explaining this — but it’s the people with that slightly more ephemeral, hard-to-pin-down wisdom who allow the wheels to keep turning. I think he uses the example of Soviet collectivization in the ’20s and ’30s, and the people — the kulaks — in the villages who actually made it work by not following all the rules. So in terms of my experience of government, the James C. Scott stuff was very compelling.

Danny: Dominic Cummings or Geoff Mulgan?

Andrew: I have to say I lean towards Dom. And the reason for that is because Geoff’s a very reasonable man. And I think, to some extent, we need a degree of unreasonableness in this debate. Coming back to what you were saying earlier: the stakeholder-management approach to progress and reform doesn’t seem to be getting very far or very fast — certainly not fast enough. One of the things I find quite frustrating about the innovation discourse, here and elsewhere, is that we’re all just far too polite and far too reasonable. No one could ever accuse Dom of that. And I do think there’s something to be said for that sense of clarity and choice and trade-off, actually. I don’t subscribe to a lot of Dom’s politics, but I do subscribe to quite a lot of his problem diagnosis of the bureaucracy. And I do subscribe to the fact that there’s an element of trade-off in the decisions you make around state capacity. I don’t go as full zero-sum as I think he’s inclined to, but I definitely think that for state capacity to be genuinely, significantly improved, to the point where I think it needs to be, there will be an element of winners and losers to that. And I’m not sure that comes through as strongly from this kind of new public management tradition.

Danny: Can you spell out the winners and losers point? What exactly is the trade-off?

Andrew: Well, we’ve touched on that, I suppose, in various flavors in the conversation. It’s back to the Thanos snap, really. At the personnel and leadership level, there is an overpreponderance of certain skills, intelligences, and attitudes within bureaucratic systems that needs to be rebalanced. And similarly, on the institutional level, some perfectly competent and well-meaning institutions should probably be swept away for newer ones. So that’s what I mean: there will be a sense of loss to this, and that needs to be done in an intentional fashion. One of the things that drives me wild about government is that for it to function, it needs to not work. I’m particularly exercised about this in the context of AI at the moment. A lot of public services basically rely on friction to be affordable. They rely on us…

Danny: We’re about to see some interesting surprises there.

Andrew: We certainly are. You’ll know all of this, but just watching what is happening to certain benefit claims, and watching what is happening to any appeals process that is free — I think we’re very much in the foothills of this. Those things will not be sustainable. They simply won’t. And in order to escape that trap, we’re going to need to make some fairly radical decisions, and probably some quite uncomfortable ones too. But everything, at an individual, personal and indeed, to some extent, institutional level, is very much guiding you towards a head-in-the-sand approach to those things. That is not going to help us.

Danny: Back to the game. Benjamin Bratton or Max Weber?

Andrew: Probably Max. Probably Max. I haven’t thought about those two as a pair for long enough to give you a deep answer on that. But I often return to Weber, I think, because it’s such an elemental text around lots of this. There’s a slight flavor — I always get it, possibly wrongly — of optimism and love for bureaucracy in Weber, which, against my better instincts, I always subscribe to. One of the reasons I do get quite worked up about these things is that it comes from a fundamentally quite conservative place, which is: I quite like democracy. I believe in institutions. I want these things to work. I am not as comfortable with the much more bleeding-edge thinking about radically different ways of organizing human beings to make things happen. I will confess to being, to some extent, a bit of a sceptic about citizens’ assemblies and things like that. Not because I don’t think they’re a good and honorable thing. Maybe I’m just too jaundiced about humans. I just don’t think they lead to as much as they could or should. So Weber, I think, because he likes boxes and lines too.

Danny: Max Weber or Joseph Schumpeter?

Andrew: Well, it has to be Schumpeter, because — I only realized this after the article came out — Jen and I have essentially ripped off Joseph Schumpeter, really, with the creative destruction thesis, which a friend of mine texted me about immediately afterwards. Schumpeter’s books were what I quoted in my personal statement when I was applying for university, so I read him a long, long time ago. But I’ve discovered about myself that I’ve always found cycles quite compelling — much more compelling than straight lines — in terms of the development of organizations and people. This is bordering on a theological point, which I think we’ll steer away from, because I don’t even know enough to be dangerous. But I really find that creative destruction mechanism compelling.

And also, as someone who’s been involved, both in government and in consultancy contexts, in those small, fast-moving teams: it is an absolutely exhilarating feeling, on both sides of that public-private divide, to be part of something where you suddenly feel you can change the world with really not very much around you. And you can. That’s an extremely intoxicating thing to experience. I’m very lucky to have done that on a couple of occasions. And it also makes me want more people to have that feeling. I remember, back in my GDS era, talking to civil service colleagues and friends working in big departments who couldn’t understand what that felt like — simply couldn’t, because there was no comparable sensation accessible to them within that organizational construct. I feel everyone should get a chance at that.

Danny: Stafford Northcote or Charles Trevelyan?

Andrew: Charles Trevelyan was an awful man by all accounts, and certainly remains, even now, a pretty controversial figure in Ireland — rightly so. But there’s no question that he was the one who actually made Northcote–Trevelyan happen. Northcote, I think, was the political frontman of that review of the UK civil service, and was a non-trivial actor in it, although Gladstone probably pushed it more than he did, certainly in the first instance. But Trevelyan had that intellectual rigor and total relentlessness, and also stayed the course. It’s quite interesting: the UK, as you’ll know, is now going through what it’s describing as a fundamental review of the civil service, led by the relatively new Cabinet Secretary, Antonia Romeo, which I admire them for doing. I hope they don’t hold back, although I suspect it’ll be quite difficult for them, as it’s an insider’s team — but I reserve judgment. Trevelyan basically had to bang that drum for twenty years, and also stayed there for that long. I think he was permanent secretary of the Treasury for nineteen years. That is, again, one of the interesting contradictions of state capacity: this mix of very fast-moving, quite ephemeral teams with the incredible longevity you need to see these things through over political cycles and the like. So yeah, Trevelyan, on the basis that he got the thing done — but he was not a good man.

Danny: Matt Clifford or Audrey Tang?

Andrew: I just think the work that Audrey has done is absolutely remarkable. And again, she’s stayed the course for an awfully long time and made Taiwan a beacon and exemplar, up there with the Estonians and others, of what is possible. Matt’s done some great stuff in the UK and been an incredibly passionate, effective advocate for AI and tech within the system. And that is not trivial, because there’s no question that in the last two, three years the UK has positively moved its position on many, many aspects of that agenda, and he deserves a lot of credit for that. But the UK’s circumstances are one set; Taiwan’s are quite another. Taiwan has always struck me as somewhere that’s managed to do — it’s not just government. It feels like quite a whole-of-society play, in a way I can’t really think of anywhere else. Again, possibly Estonia, but even more so, perhaps. So yeah, Audrey.

Danny: John Kay or Michael Barber?

Andrew: I’m going to have to go with Michael on that basis.

Danny: Fair enough. I think the interesting tension between the two is that John Kay wrote a great book called Obliquity, which says that some things we can only achieve indirectly. You’ve talked about this in relation to GDS: had the Government Digital Service been given a straight-up savings target, probably none of it would have actually been achieved. The way to go about it was an indirect one. Michael Barber has done some phenomenal work, but I think intellectually he in many ways stands for the exact opposite idea, which is: there’s a thing you want to achieve, and the best way of going about it is just being brutally clear about what that thing is and then relentlessly pursuing it until you have it.

Andrew: And that’s not wrong either, right? That’s always what I admired about the PMDU push, and that sense of a prime directive. And although GDS had a bit of obliqueness in how it told the story, exactly as you say, saying “we are going to ship the nation’s website on this date” was the point at which GDS was probably most effective. Again, the whole thing is riven with contradictions: it needs agility and flexibility, but actually some fixed posts are quite helpful. Which speaks to a broader point here, and one that I think is relatively silent in a lot of the stories you’ll hear about state capacity, be they the Vaccine Taskforce or GDS or others. They’re often described as the story of an individual, and often an individual leader — which is why this pairs thing is a really good game. Because the thing that I really learned from GDS was watching a gang turn up. There were obviously people like Mike Bracken, who ran it; Tom Loosemore, who was the person who started it; Ben Terrett, who was the head of design — those three were all my colleagues — and a bunch of others. But they all came as one. It was not “this person is going to come in and fix it.” It’s these people.

That very rarely happens in government. It’s almost invariably one person who’s brought in as the czar or the guru, and they’re going to come and save the day. And sometimes you have to make one person that for the public story, because it’s easier to tell. But actually, the ability to do both the prime directive and the oblique thing is quite a good example of where that works. Mike’s very good at saying, “This is what we’re going to do. We’re not doing anything else. And we’re going to say no a lot.” And there were others in the team who complemented that with a different set of tones. You need that, and government’s instinct to find one throat to choke tends not to allow for that very well.

Danny: I’ve read your new book on digital sovereignty. I think it’s excellent. Why are so many takes on digital sovereignty so bad?

Andrew: Well, it’s hard, and it’s fast-moving as a topic. So I’m being generous and forgiving, because in eighteen months’ time someone will say about the book that I’ve written with my colleagues, “What on earth were these people thinking? The world’s completely changed.” So I want to be generous in that respect. I also think it’s quite a hard topic to stand truly dispassionately apart from. In the book, we have generally been pretty silent about the geopolitical choices that are involved. We’ve stretched our opinions slightly further when it comes to how you work with some of the big technology companies, and how you should be quite careful about getting into bed with them in those commercial relationships. But I think quite a lot of the takes I have struggled to get on board with have come from a place of partiality, which doesn’t mean they’re bad. It just means they’re doing something different. It’s a debating-society point, I guess, rather than a willingness to genuinely engage in what the right answer may or may not be.

So there’s a huge amount of selfish interest going around sovereignty, I suppose. And it’s really interesting — I found this really fascinating as part of the process of putting the book together — how different countries and different powers approach it. I’d say the UK in general was quite a long way behind in terms of public discourse on sovereignty compared to some others, particularly some in Europe, but also in parts of Asia and Africa. I would speculate that maybe that’s because the UK psychologically still has a little inkling of great-powerness in the back of its head, and hasn’t really felt itself politically or psychologically able to engage with this topic as a middle power. And if you come into sovereignty and don’t really know who you are and where you stand, you’re going to come up with some pretty bad answers, I think. So it’s the clash of a topic that’s complex, technical, fast-moving and political with the fact that people are often in a state of denial about where they are.

Danny: I think a really important point you make in the book — and there are many important points you make — is that the big evil character, as it were, is inertia, much more than anything else. Much more than some clever gigacorp running circles around government, though no doubt that happens too. It is just the accreted mess of past decisions that makes it very, very hard to change course. That is, in a sense, the first place to look. And that feels to me like a very useful heuristic.

Andrew: This possibly comes from spending a chunk of time in government and around politicians, but I find conspiracies very hard to absorb. I’m always in favor of cock-up, and human foibles and human frailty. That is not to say that there aren’t some very clever people working in technology companies thinking about artful ways in which they can get their claws into a relatively rent-seeking relationship with states. I’m sure there are. It would be weird if there weren’t. But equally, a lot of stuff just happens. It happens by degrees. It happens by increments. It happens by compromises. And it happens because there are other things to think about.

One of my great colleagues in GDS, a guy called Richard Pope, was saying, when I was working with him a very long time ago, that software is politics. I remember him saying that in 2011, 2012, but he may have been saying it long before that. Back then it felt like quite a new idea: “No, it’s not. It’s just what the IT people do.” It’s indisputable, and now I think increasingly mainstream, the idea that software is politics, which means it’s now being drawn into the spheres of the policy, think-tank and diplomatic worlds, which have never, truth be told — other than at the very edges, and a handful of noble individuals — engaged with those topics really seriously at all. So I think you’re seeing a lot of people really struggling to catch up. When I think about the number of people in the diplomatic world who are deeply engaged in climate, for example, and the science and the discussions around that, versus those in tech, I think that’s a really surprising disparity. As I say, there are many honorable exceptions. You’ve got people like Henri Verdier in France, who’s been doing a digital ambassadorial role there for a long time. Audrey herself has effectively done similar things from Taiwan, and there are others.

But I do think a lot of the trouble — and this is what we get at in the book — a lot of the trouble that organizations get into, private organizations as well as governments, is just looking the wrong way quite a lot of the time, and doing that very human thing of not really wanting to confront the quite ugly mess that they’re sitting on. One of the first things we say in the book is: just start by working out what you’ve got. I would be reasonably confident you could ask almost any government in the world how much it is currently spending on technology, and who it’s spending it with across its estate, and it will not know.

Danny: Which is remarkable given that — you’ve made this point, and it’s straightforward, but I think incredibly important — to the degree that government is essentially concerned with the manipulation of information — there are some exceptions, but it’s mostly that — given that it’s mostly symbolic manipulation: congratulations, you are running an IT company.

Andrew: Yeah, absolutely.

Danny: To a first approximation, that is what it is. As you say, it could not be more important or fundamental to the enterprise.

Andrew: Absolutely. Or, possibly even more germanely, maybe a data company. And it fascinates me, the extent of the gap between the public perception and the officials’ perception. You talk to any official in government and they’ll go, “Obviously all the data is rubbish. We can’t really share it between departments. It’s all a bit of a mess.” And the public assumes, “Well, they’ve got all my information. Why can’t they do this?” It’s a bit like power. One of the things that’s been fascinating about the rise of populism across parts of the world in the last ten, fifteen years is that part of it comes from the fact that people look at their presidents and prime ministers and politicians and go, “You’re in power. Why can’t you do something?” This idea that it’s all quite difficult, actually — it is, but that’s not what people believe. “We elect you to do the things.” I don’t know whether it’s a confidence thing. I’m not sure quite what it is.

But on the point of government as a technology company, or as a series of data layers: that underpinning, foundational infrastructure of government across the world is remarkably neglected; it’s in a real state. And it’s because it’s expensive, difficult, complicated. And where’s the political announcement there? It’s just quite a tough ask to fix. But I think it’s an enormous prize. And this is where we’re back at our creative destruction and new institutions point. Would it be best to start again from a blank sheet of paper with that data? It feels like we must have loads of good stuff in those legacy institutions. Surely we can make the most of it. I genuinely don’t know. Maybe we do. Or would it just be quicker to get the clipboards and surveys out and begin again? I really don’t know. But that feels to me like probably one of the most elemental questions of the age to come.

Danny: One thing you advocate for in the book is this idea of situational awareness. Unless you have any clue what is going on, you will struggle mightily to make good decisions — which feels like a fairly straightforward point, and yet… One of the most useful ways I have come across is Simon Wardley’s work on Wardley mapping, which I think doesn’t appear in the book. But I feel like it’s one of the… I mean, it’s also an unfair criticism. Of course lots of things don’t appear in the book, right? And that’s fine. So this is not a criticism of the book. But I think his is some of the most useful work in terms of not just understanding where you are, but also making some reasonable assumptions about where the puck is going — which is where you want to be skating — rather than being fifteen years behind the paradigm and then spending lots of money on that while everyone else has moved on.

Andrew: Simon’s stuff’s great. We referenced it in our last book, so maybe we just thought we’d do it again. But his work is excellent, and you’re absolutely right: his maps and his techniques are real strategy — or at least they’re part of that toolbox. And situational awareness is obviously, since time immemorial, an elemental ingredient of doing proper strategy. We’re probably getting sidetracked here, but proper strategy is a really, really rare thing in any organization, frankly. Certainly in a governmental one, where “strategy” tends to be used as a shorthand for “how do we make sure this particular document agrees with all the other documents that we’ve published, and how do we make sure that everybody who needs to comment on it has commented on it” — none of which is strategy.

I think around sovereignty, and how governments or organizations respond to the challenge ahead, it’s going properly back to basics in a strategic sense. As you say, these are some very straightforward points: where do we want to play? On what territory can we realistically compete, hold the position, and have a point of advantage? And how do we avoid getting pulled onto a terrain where we’re not going to win? It’s going properly right back to the Greek and Roman stuff. So there’s that. And then there’s the more modern take on strategy — I’m quite influenced by Lawrence Freedman’s compendium on the history of strategy, which is a spectacularly good book. He almost finishes it with strategy as storytelling, strategy as narrative effect. And one of the facets, coming right back around to state capacity, is that state capacity is about substance, but it’s also very much about style. It doesn’t matter if you’re actually improving things on the ground if people don’t know and feel and recognize that you’re doing it. And it’s the same with sovereignty. What is the story you’re telling that is compelling and believable and authentic, so that people do think you have a degree of control and agency over the picture that you’re operating in?

And government goes out of its way — or certainly its bureaucracies go out of their way — to not tell stories that are resonant, because they require humanity and antagonists and tension and conflict. I’ve never seen a government press release with any of those in it. Except, interestingly, when you have a crisis — viruses are quite good antagonists, actually, and COVID did see some good stories being told. So that combination of the really elemental — where do we want to play, and why? — plus the more modern “what is the storytelling narrative that sits on top of this?”: that’s strategy, and the country needs more of that.

Danny: Amen. One big element that I felt the book leaves unanswered — which, again, could be said about many books, because you cannot possibly cover everything — is this: if you assume that the capability development of AI systems is going to continue anything like what we have seen over the past years, and if you assume that, as we’ve said, government is to a large degree concerned with information processing, you might conclude that it would be very surprising to see an effective government that doesn’t use AI. And I mean not in the sense of whatever it is we’re doing now, but slightly faster with AI, but in a much more fundamental way. And that seems to start to touch on fundamental questions of sovereignty, because it’s not clear that the UK, or any country other than China and the US, can in fact compete on the frontier. Now, maybe we won’t need the frontier models to do this. There are lots of ways in which this could go differently, but it’s not obvious that it will. So how do you think about that?

Andrew: I agree with you on drawing the trend line. I can’t see how — I don’t think opting out of that is going to be an option. The dangerously glib analogy that I hold in my head for the frontier stuff, where I agree it seems hard to imagine anyone other than the US and China really playing, is that it may act in a similar way to a nuclear deterrent, where there are umbrella-type effects that may play out. And I certainly don’t think the UK is going to do what Britain did in the ’50s — the “it needs our flag on it” type of thing — with a frontier model. I don’t think that was barely an option for us in the ’50s; it’s certainly not an option for us now. One of the things the book does try to get at a bit is decoupling ownership, in a raw sense, from sovereignty. And I do think middle powers like Britain, France, Canada have the ability to shape — not so much the frontier, but certainly markets, and other aspects of the technology and the services that flow from it, and indeed the talent that fuels it. I think there is absolutely plenty of space in which the UK and others can play.

Will it mean quite an interesting new frame on diplomatic interactions with those great powers? Sure. But it was ever thus, I think, to some extent. And I think some of the UK’s belated waking up to the sovereignty questions is a product of a Trump administration which I’ll describe as erratic. Before, we probably haven’t had to contend with those questions in a particularly thoughtful way, because it was just taken as read. So we’re in a more dynamic space. But as I say, I think you can retain and improve your degree of sovereignty if you’re building the agency and capacity to play within the areas that you can, and to shape markets where you’re able — and potentially, as well, to develop new multilateral models, which are very unfashionable in the current age, really. But there is certainly a version of the AI future where multilateralism and those kinds of collaborations resurface, re-emerge, possibly in different shapes — probably in different shapes — and the conversations around the EU and Canada and others are quite fascinating as a precursor to that, I think. So I don’t have any great answers, but I don’t necessarily see the US and China acting as a duopoly on frontier models as a substantial frame change from the great power dynamics we’ve seen in the past. And as ever, those who aren’t those two will have to adapt accordingly.

Danny: In a sense, this is also a commentary on the digital sovereignty debate writ large. Dependence on critical inputs from outside your own borders is not exactly a new problem. Most countries don’t produce oil. For most countries, oil is quite important. And we have evolved ways of managing that dependency. So some of it strikes me as weirdly ahistorical in how we talk about these things. That said, to the degree that you need AI, as a state, to make sense of the world and make decisions, it does feel like maybe a qualitative change from before, one that starts to touch on the fundamental nature of what your government actually is. And to the degree that you have to import that, I could see why some anxieties would be justified.

Andrew: The thing I also find slightly ahistorical — or possibly it’s slightly more recent history — is that most of the NHS runs on American systems. We have already outsourced large parts of the state machinery to other parts of the world, without any particularly sophisticated process. I think it’s a triumph of the likes of the Microsofts and Oracles and others, versus someone like Palantir, who have been very aggressive in their approach to getting into government. That has lit the touch paper on a fairly febrile debate. But Oracle and SAP and Microsoft have been doing this stuff for decades. It’s quite fascinating how those connections are not necessarily being drawn.

One other quick aspect I’ll just touch on very briefly about AI, which I think is also fascinating: it’s quite easy to just think of it as “what can the state do with this to change how the state interacts with citizens?” It’s absolutely going to go the other way too, and already is. And I think the politics of that are fascinating. We’ve already touched on this a bit, but there’s obviously a version of agentic flooding and a frictionless state, which gets to the point where we can’t afford the current models that we have of delivering services. So from the right’s perspective, there are some interesting questions there about economic viability. But also, from the left, you’ve got the fact that there’s every chance that the people who make the most use of agents to do their bidding with the state — to flood the system for school places or entitlements or whatever it happens to be — will be… It’s the sharp-elbowed middle-class thing. So you have this quite deep inequity creeping into interactions with the state, from the other side of the curve. I think we’re not far out from that getting much spikier. We’re not there yet, but both sides of that equation are going to get much spikier.

Danny: On the right, I wouldn’t be shocked to see some accelerationist moves, right? If you want to see the system break down… And to be clear, what we’re talking about is that currently the frictions that prop up large bits of the state are important. Once they go away — once everyone can submit an appeal about wherever their child gets sent to school, or apply for social security, frictionlessly — that is what you’re talking about.

Andrew: Yeah.

Danny: I wouldn’t be surprised if someone provided a couple of tokens for people to do that, just to speed up the inevitable, as it were. Final question: what will AI do to the bluffocrat?

Andrew: Now, James and I — James Ball, who I wrote Bluffocracy with — very nearly managed to write an article about ChatPPE.

Danny: Wait, explain that for everyone.

Andrew: Well, to be honest, we didn’t get much further than the title, but we like the title. Again, there are just two versions, right? There’s one where it’s brilliant at replicating what that generalist adviser class does. It’s even got the sycophancy down pat. It’s so good at that. And in terms of its ability to critique and respond — not so much create; I always find the creation stuff quite weak — but certainly in terms of its response and summation of stuff, the bluffers are quite under threat from the chatbot-type stuff. On the other hand, in terms of the modernization of the state through digital and internet means, the bottleneck has tended to be the number of practitioners you can get in — developers and designers, people who actually build these things. That’s not going to be true for very much longer. And so where’s the bottleneck then? Well, I suspect the bottleneck will probably end up becoming judgment, and the human relatability to make it — the point of accountability, basically. Which assumes, possibly wrongly, that the models themselves don’t become those nodes of accountability. But if it comes down to the good chap who says, “We’ll take this answer and not that one,” that could be a space for the bluffers in the future. I think they’re going to be pretty hard to get rid of. Being a bluffer will have to evolve, and ultimately they’ll have to do what I do, which is be able to bluff about technology as well as everything else.

Danny: With that, Andrew, this was a ton of fun. Thank you so much.

Andrew: Thank you, Danny.

Danny: Thanks for listening to High Variance. You can subscribe to this podcast on Apple Podcasts, Spotify, or wherever you get your podcasts. If you like this podcast, please give us a rating and leave a review. This makes a big difference, particularly for newer podcasts like this one.