AI futures: agency for humans not agents

AI may or may not be out to kill us but it will make things profoundly worse and not better if we don’t make sure we are moving thoughtfully and intentionally into embedding it even deeper into our everyday lives. This is even more important for public services and for public service reform.

This is not a technology problem; it’s a social one which requires everyone to pay attention. This means we need to think systemically and not in silos. There is rich learning from how we failed to do that as social media started to drive so much social change, let’s not make the same mistakes around skills, governance and purpose.

Technology is now through the fabric of so much of our lives and is core to how we deliver public services. This makes the future for AI a system change opportunity and challenge.

My north star is an exploration as to how we use technology to enable public services to be more accessible, relational and sustainable. I have come to realise that I think about AI from three different perspectives:

• As a technologist who’s deeply interested, in terms of the actual technology, in how you use it as a tool of transformation and how we might better shape the market and improve the choices we have when bringing technology into our services.
• I’m interested as a chief executive of an organisation of getting on for 5,000 people with huge demand pressures and a need to ensure that we stay within the limits of our budget while improving our services.
• I’m also thinking about the wider impacts of AI on our local economy and, in particular, on those people who are currently struggling to access work. Young people, people who’ve been excluded from work because of long-term health conditions. We want and need people to be economically active for longer, but they don’t necessarily have the skills to thrive in a very different context.

There are trade offs in these perspectives. As a CEX I need the productivity gains that can’t be ignored but question whether we can adapt the skills of our workforce and future workforce fast enough in order to avoid making the challenges for young people entering the workplace worse. The way that I’m unpacking this set of problems is starting with the stuff that I’ve got greater control over and then moving outwards. It’s helping me to think about the problems that I want AI to solve rather than just being dazzled by its possibilities.

We experienced the social media revolution in waves and I think we can apply the same framing to AI and how we apply it. But be aware, all of these are actually in play at the same time, the waves are more about where you are focusing.

1. In this first wave, we want AI to eat the work that bad technology has created. My people are overwhelmed by bad processes, by double keying, by workflows which don’t make sense, by misaligned handoffs between different systems, often different legacy systems. We are all drowning in our inbox, and we are generating noise in our system with our technology platforms rather than them allowing us to focus and prioritise. The first job of AI is to clear up the mess that the other technology has made.
2. Once we’ve done that, the second wave provides the ability to continue to optimise and get AI to do the heavy lifting to allow people to have time to focus on deeply relational things. What do I mean by that? I want my customer service operators to be unconstrained by a time limit on a call. By the time somebody’s worked through an elegant agentic experience and come to the conclusion that they’re an edge case that needs an actual conversation, that customer service person, is able to spend as much time as the individual on the phone needs.
3. At the point at which this is happening, we’re really talking about AI creating an interface layer. That means that we’re suppressing, and compressing the legacy technology, which is causing a lot of problems in the first place. We’re creating a more connected set of interfaces with simpler user experiences that allow us to maximise time and create decision-making support while keeping a human in the loop. We have created a more elegant environment for people to work in, in a way that actually allows them to focus on what’s important.

I worry about our desire to continue to optimise. I have for a long time been concerned about a race for efficiency being something which really undermines resilience. I’m also concerned that by optimising for the transactional, we miss the opportunity to be relational. There are two axis at play here:

1. The tension between a transactional state versus the ambition for a relational state which focused on human interactions and we see people as people?
2. The balance between deeply participatory state and one where we cede power to representatives.

We really need the system to be in balance across those four axes, with people able to move between those quadrants. I think we need to acknowledge that those are the axis that exist. I hope we’re trying to design for a future where we can optimise transactional but recognise where we need to be human, where we can enable people to participate and solve people’s problems together.

At this point, what we’ve really done is use AI to modernise and optimise the environment in which we’re currently operating. We could then start to think about what comes next. This next section is a sketch of the kinds of questions I see in there.

It’s not hard to imagine that we’re only a few years from people actually mediating their online experience via a large language model rather than directly via the web. There’s a real question about what the futures of browsers are in all of this, and that actually means that you have to consider what the future of your website is.

I think there’s a fork in the road here, that public services need to consider before that comes upon us.

There’s a version of the future which is a deeply transactional one, where those AI interfaces are interacting directly with our processes, our systems. What we’re doing is we’re creating a perfect customer and supporting individuals with the transactional services. We may want to do that, but the fork in the road is really about the role of public services in all this and what we lose by removing the human in the loop by optimising for agent to agent transactions.

The alternative future draws a different line between the things we automate and the things that need to be deeply human interactions. Its a future where we make sure that we design a continuum of relationship rather than just moving the boundary between state and citizen to a different place.

But that is still trapping us in a transactional paradigm. In thinking about the future role of public service, do we actually want people still coming back to us to solve their problems, or do we need to open up the possibility of designing in the fact that communities are great at solving problems for themselves? How do we want to deal with people finding their help in the community, not necessarily people coming to the state? How do we make sure that they’re visible when they do need something? How do we make sure we manage the boundary between who does what? How do we make sure that people interact safely?

It’s really hard for me to imagine this future without us having a layer of civic technology which allows communities to safely collaborate, to safely interact with each other. This will blur the boundary between the state and the individual and create that space where communities can solve problems themselves in a space which is not transactional, but relational and participatory. This is exactly where my PHD research took me and I still can’t quite get my head around what this looks like, but it’s a thought experiment which I want to continue.

In all of these futures we need to understand where decision-making lies and how to ensure that our representative democracy continues to function. All of these futures create a very different set of relationships between citizen, state, and technology, which need a lot more mindful design.

I think there’s a couple of years at least of work to be done to eat the work that technology has already caused and to create that much more relational state. I also think over a few years of the replacements of browsers and apps with a generic experience, but I think it’s probably time that we started the conversation. I think it’s definitely time that we start thinking about the civic technology that we need in place if we’re going to avoid codifying a deeply impersonal transactional state.

The Rewiring the State statement is hugely exciting and is an opportunity to think about all of this stuff, but its needs us to be active shapers of technology and not passive consumers. If we don’t think about the technology as it could be when we get there we’re not going to shape the market to have the tools available that we need.

More than that, we’re not going to solve the thorny issue of who actually pays for civic technology, something which has long been an issue. We’re not going to be able to impose ourselves on the marketplace to make sure that those agentic interfaces that will be the tool that people use to work with government are actually going to be in their best interests. They are going to be accountable and will be able to be held up to scrutiny.

No one future is inevitable, but without serious and concentrated effort we will drift into a transactional future for our public services because thats where we are stuck now and we haven’t spent time imagining the other possible futures. Lets take that time.