AI, and how to make sense of the UK Government’s Budget announcement, without resorting to reading it

The Budget landed this week with rather more drama than usual. Not in the statement itself, but in the forty-odd minutes beforehand, when the Office for Budget Responsibility (OBR) accidentally published its full analysis online — indirectly exposing the economic heart of the Budget before the Chancellor had even stood up. A classic “never event”: the sort of thing that simply shouldn’t happen. But it did.

Markets didn’t politely wait for the Chancellor. As soon as the leaked forecasts dropped, sterling ticked up and UK government bonds rallied as investors reassessed the UK’s near-term fiscal outlook and recalibrated the value of the pound. According to Reuters, the pound gained roughly 0.3%, and yields on ten-year gilts fell a few basis points. Hardly Truss-level turbulence — although we are still managing the fallout from that era — but the impact was immediate and real.

My full faith in cock-up over conspiracy is rooted in twenty years of civil service experience, so I mainly pictured a young official sobbing over a keyboard after accidentally publishing something to a live webpage. Fingers crossed the Treasury is a beacon of psychological safety where these kinds of errors are understood and forgiven.

But the leak shouldn’t be the story.* The real question is: what did this Budget actually tell us about the UK’s prospects, productivity, and the role AI might play in the years ahead? And, more importantly for most people, why does any of it matter when all we really want to know is whether their bills are going up?

Trying to make sense of the Budget when you don’t want to read Treasury prose, you’re drowning in charts, and you don’t entirely trust the headlines

The Budget always lands with the same paradox: it is simultaneously one of the most significant economic moments of the year and something most people immediately tune out of. Understandably so. The documentation is dense. The speeches are mainly rhetoric (and, this year, dragged out over several weeks). The numbers feel disconnected from real life. Unless your staff costs, personal tax liability, mortgage, rent, food shop or wage packet changes sharply, it all feels pretty abstract.

Some corners of the press — and government itself — do their best to make it explainable, but they’re trying to make it explainable to everyone. And everyone isn’t you. You’re unique. (We all are.)

What struck me yesterday, scrolling through the coverage and commentary, is how reliant we’ve become on others to tell us what a Budget “means”. Journalists, think tanks, economists — all doing the translation work. It’s a reminder that our entire fiscal system communicates in a dialect most people don’t speak.

And this is where AI, used mindfully, helps. Not as a novelty add-on or a chatbot plastered on a website, but as a way of turning sprawling documents and macroeconomic forecasts into language you can actually use.

Imagine being able to ask, “How does this affect someone earning £38k?” or, “Where, precisely, is the investment in digital and AI?”

Actually, don’t imagine – just ask Perplexity. Like I did. Then check the sources and analysis. Or make yourself a little RAG assistant** and go to town extracting the details you need – try dropping the documents into Notebook LM and interrogating them directly.

Some prompts worth stealing:

• “Identify every part of the 2025 Budget that relates to AI, digital infrastructure or productivity. Provide quotes and page references.”
• “Give me the top five Budget changes a senior leader should understand, with a short explanation of why each matters for organisational planning, workforce strategy or service delivery.”

The bigger question: are we serious about productivity or still tinkering at the edges?

There are some parts of the Budget and Spending Review that go beyond silicon-hype, signalling real ambition around public-sector digital transformation. For example:

• The government has committed an additional £1.2 billion over the Spending Review period for cross-cutting digital and AI priorities — part of a broader package that brings total DSIT (digital/AI) Departmental Expenditure Limit funding to £1.9 billion.
• There’s a major push to scale up national computing infrastructure — often the missing piece in public-service transformation. The plan reportedly includes funding for a new exascale supercomputer, and an expansion of the national “AI Research Resource / compute power,” designed to massively increase publicly controlled compute capacity.
• More generally, the Budget projects an increase of over £120 billion in departmental capital spending (compared with previous plans) — which at least preserves some headroom for long-term investment in each sector.
• On R&D, public funding remains fairly strong: the 2025–26 allocation for UK Research and Innovation (UKRI) is confirmed at about £8.8 billion, safeguarding the core capacity for national research and innovation.

So, yes, the 2025 Budget does more on tech and AI than hype. But few of those line items translate, yet, into a robust skills or transformation programme. We couldn’t see a major, headline-pledge for a national initiative to really transform how services can operate, enabled by AI. Most of the funding is either for infrastructure (hardware, compute) or for research bodies that might only seed pockets of innovation — not retraining, process redesign or transformation at scale. Observers from organisations like techUK pointed out that while the investment is welcome, the “skills package” has been delayed.

In other words: the building blocks are there, but the plumbing — the staff, skills, change-management, governance — remains optional until we see a follow-through or, realistically, we each reprioritise our budgets to sort ourselves out. Set against the reality of how the public sector actually works — the processes, the legacy systems, the human workflows that can accidentally send markets spinning — it still feels like we’re treating AI as a shiny growth lever rather than a foundational part of how the state itself operates.

If we’re serious about unlocking productivity at scale, AI can’t just be a line in the innovation chapter or a few projects in your IT budget. It has to reach down into everything: how information flows, how decisions are prepared, how policies are modelled, how risks are managed, and how citizens get information that genuinely empowers them.

This week’s leak was awkward, but it was also revealing. It showed that even the institutions tasked with forecasting the country’s economic future can be undone by something as basic as a publishing fail. And if that’s the state of our “plumbing”, no amount of high-level ambition will deliver the productivity transformation the UK needs.

I am optimistic — genuinely. We’ve seen in local government, including with our own Policy Buddy work, how AI can reduce friction, increase transparency, speed up decision-making and free people to focus on work that matters. But it only works when organisations are willing to modernise the underlying systems too. AI on top of broken processes won’t deliver the kind of leap we keep promising ourselves.

Full disclosure: AI was the only way I got through the Budget

I should also admit something. I’ve been working through a mild Covid fog all week and the only reason I was able to digest the Budget, OBR forecasts, market reaction and analysis in anything like a reasonable time was because I used AI. Not to replace judgement, but to sift, speed-read, group, summarise and highlight the parts that mattered.

I do not have the time to read every line of the Red Book on a Wednesday afternoon. AI let me get there faster, without drowning in the noise or the ringing in my ears.

If you do want to follow people who actually read the Budget paperwork — the tables, footnotes, annexes and all — I’d recommend some of my favourite grown-ups: Sam Freedman, Jonathan Portes, Paul Johnson and the IFS team, Torsten Bell and the Resolution Foundation, Jill Rutter at the Institute for Government, and Chris Giles. They do the heavy lifting so the rest of us don’t have to and we should thank them.

*If this all feels familiar, you might be thinking of the time in May 2020 when the official @UKCivilService Twitter account briefly posted: “Arrogant and offensive. Can you imagine having to work with these truth twisters?” following Boris Johnson’s defence of Dominic Cummings. I have it on a tea towel. The tweet was deleted within minutes, but not before making headlines and prompting a Cabinet Office investigation. One wonders how the recommendations translated into approval processes for posting things online.

**Still our favourite thing: when you hook your LLM and GPT into a defined data repository, so it can’t just make stuff up. Enquire within.