A friend of mine once worked for consulting firm that won a government contract. For the next nine months he was deployed with a Whitehall department to support digital transformation projects.
The work started off well, but after a while, he and colleagues ran out of things to do, and began to be ignored. For the last several months, he would show up to the office and would sit, twiddling his thumbs, until his shift was over.
I didn’t believe government waste was that bad, the friend said, until I became the subject of it. He was staggered that the state would spend so much money on consulting fees and get so little out of it.
There will be countless more experiences, like those of my friend, of the government squandering taxpayer money. Which was why it was welcome when PM Keir Starmer said he would use cutting-edge AI to remove waste and make the government more efficient. But can he really find £45bn, the savings figure promised?
Let’s look at the government’s own tech department, DSIT. The department’s total expenditure in the 2023-24 year was £12.3bn. That’s around 1% of total government expenditure for that year – so DSIT’s share of the savings would be £450m.
Spending on staff costs was £1.1bn, or around a tenth of the total budget. This is presumably the biggest cost centre the government wants to find savings in – and Rachel Reeves has spoken about cutting over 10,000 civil servant jobs. In DSIT’s case, that would mean scrapping two in five staff by 2029 – a tall order.
Various predictions abound about quite how many jobs will be replaced by AI – and how fast. Generative AI has been mainstream for more than two years now, but so far the number of jobs that have fallen victim to it is modest, save for in roles which involve discrete, repetitive communication tasks, such as in call centres, in copywriting and in programming.
But in more complex, less predictable multitasking, AI is still some way away from being as good as humans – anywhere from two to ten years, depending on who you ask.
Why is why it seems unlikely that the government can deliver so much in savings so quickly. And that’s before you consider the up-front costs of licencing this technology and training staff to use it. So I’m prepared to predict that the government will not save anything close to £45bn – at least not with AI alone.
The much bigger opportunity for cost-saving, which AI cannot solve, is in the duplication of work that takes place between government departments, who each have their own datasets and systems of digital organisation. Centralising this and improving interoperability would be a monumental feat but the long-term cost savings would be enormous. And it would also allow AI tools to be trained on much larger datasets, making them more valuable. But like any attempt at government reform, things will never quite go as well as intended.
The post Can AI really save £45bn from the state? appeared first on UKTN.
