I’m getting increasingly frustrated with how lots of people are talking about Artifical Intelligence (i.e. AI) as if it can fix all our problems. I understand the excitement around AI, it’s a powerful technology with the potential to drive significant change.
In particular, AI is often being positioned as a key part of unlocking the UK’s economic growth. That ambition is shared by many organisations, not just government, and it’s easy to see why: when used well, technology can be a powerful lever for productivity and impact.
Economic growth fundamentally depends on one of two things: either an expansion of available resources, or an increase in the productive efficiency of how existing resources are used. In a finite system like the UK economy—where most resources are already in use or under defined ownership—growth can really only come from doing more with what we already have.
That’s why there’s so much enthusiasm about technology as a lever for productivity. In practice, when people talk about AI, they’re generally using it as a kind of shorthand for applied computer science in everyday life.
But here’s the problem: many of the people pushing AI as a solution don’t understand how it works.
They treat it as a kind of magic that will somehow transform government, boost efficiency, and unlock growth. What they’re missing is that the hard work required to make AI effective—especially in a complex system like government—has nothing to do with AI itself. It’s the same structural, process, and data challenges we’ve always faced.
It’s worth thinking about the “front end” and the “back end”.
When we talk about artificial intelligence, it’s easy to get distracted by the shiny front end—those slick user interfaces that look futuristic and promise instant answers. That’s the part everyone gets excited about, and rightly so: it’s impressive, it’s accessible, and it makes AI feel tangible.
Behind the scenes lies the “back end”—the foundational processes, data, and systems that do the real work. For AI to deliver value, that back end needs to be robust, consistent, and well integrated.
This is where some of the current optimism around AI—especially in large, complex environments like government—needs to be accompanied by a recognition of the underlying work required.
AI, particularly the kind built on large language models, essentially acts as a clever front door. It can interpret language, generate convincing responses, and create a sense of seamless interaction. But underneath that sleek surface, you still need a functioning back end. You still need the processes, systems, and data that do the actual heavy lifting.
Where does the efficiency actually come from?
Government services are often delivered through systems that have evolved over decades. That brings a wealth of institutional knowledge and deep public service expertise—but it also means that some processes may be undocumented, data may sit in silos, and different teams may use different standards or systems. These are common challenges in large organisations, and they’re not unique to the public sector.
Making AI work in this context requires clarity, consistency, and collaboration. It means investing in data quality and governance, aligning processes across teams, and building digital confidence and capability across the organisation.
This is not an AI problem. It’s a change problem.
Process change. Systems change. Cultural change. Strategic and Cultural alignment. Skills development.
And change in government is hard. These things take time, care, and coordinated effort. Much of what makes government work, relies on the knowledge and experience of dedicated public servants, who often navigate complexity using expertise built over time. People who understand how to navigate the bureaucracy, not because it’s written down, but because they’ve learned it over time.
Tacit knowledge and the ‘culture’ fill in the gaps where formal processes don’t exist. How culture gets developed deserves its own article, but basically: culture isn’t something you plug into an algorithm. It’s an organic thing that grows around informal practices and institutional memory. Doing again what was successful the last time you had to do something.
I’m not a Luddite I promise
Don’t get me wrong, None of this is to downplay the potential of AI in government. I use AI all the time. Part of my day job is using computer science to solve real world problems. Improvements in technology are good.
Imagine being able to have a quick chat with a digital assistant that can book appointments, explain your tax code, or follow up on a missed bin collection. Or being able to optimise ambulance and paramedic rotas with a click. (I’m sure we could come up with thousands of use cases).
These are real, tangible benefits that AI could help unlock.
But to get there, we have to lay the right foundations. That means doing the hard work of building the right systems, standards, and skills. With those in place, the potential is enormous.
Thanks for reading
If you enjoyed this article- the best compliment you could give me would be to share it with someone else who you think would enjoy it.
And if you liked or disliked this article, please let me know why in the comments below.
If you’d like to know when I’ve posted you can sign up to my blog here.