Civil Service: Artificial Intelligence Productivity Gains

Question

Asked by

To ask His Majesty’s Government what assessment they have made of productivity gains across the Civil Service resulting from the deployment of artificial intelligence; and how they are measuring and evaluating these gains.

I beg leave to ask the Question standing in my name on the Order Paper and, in so doing, draw noble Lords’ attention to my technology interests as set out in the register.

The Government have assessed the potential productivity gains from AI across the Civil Service, identifying which solutions are most effective and that will scale. We conducted the world’s largest trial of general-purpose AI tools, such as Copilot, to measure their impact and benefit. These evaluations inform which technologies deliver which outcomes and will enable us to prioritise adoption. Our assessment indicates significant economic benefits from widespread, well-targeted AI deployment across His Majesty’s Government.

My Lords, in May, the Government announced a rightly ambitious plan to drive tens of billions of pounds per year in Civil Service productivity savings through AI. In June, they estimated that use of AI tools was saving civil servants 26 minutes a day. Even if that is true, the Minister will agree that time savings, however welcome, do not equate to productivity. Now that the Government have spent £573 million on AI tools and consulting for the Civil Service, how will they measure productivity gains to rapidly build on successes and shut down failures?

The figure of £573 million is, of course, forward spend, so that is not what has been spent; it is a commitment over the next few years. It is important that we measure this. Guidance on how to measure the impact of AI tools was issued last year; there is a rigorous process for doing that. On the 26 minutes that has been picked up by the Copilot study, it is of course a general AI tool. Much greater savings come with specific uses in specific areas, which will not be general across the Civil Service.

The noble Lord is quite right to point out that time saving is not productivity, but what we do know from studies elsewhere and across business is that, when you get those time savings, about half of it goes on core tasks, about a quarter is on other strategic and creative work and about 25% goes on enhanced well-being. That is what we might expect as a result. There is a lot to do to make sure that we implement this properly across the Civil Service.

My Lords, would my noble friend the Minister also consider assessing productivity gains from basing civil servants in the regions rather than in London? In my view, as an ex-regional Minister for Yorkshire and the Humber, that would be an effective way of ensuring that the Government were focused on reducing regional inequalities and encouraging regional growth.

I thank my noble friend for the question. I am not going to try to add the regional Civil Service to my brief, but I will say that this is exactly the sort of area where AI is rather good at working out whether there is a benefit, because it can spot patterns that are difficult to spot individually or by human endeavour. This is a place where you could begin to see what the real impacts are and what drives success rates in the regions.

My Lords, I declare an interest as an adviser to DLA Piper on AI policy and regulation. Given that the Government have just set up a sovereign AI unit with a budget of £500 million to promote domestic AI leadership, what steps are being taken to ensure that British companies and start-ups have fair access to Civil Service AI contracts rather than defaulting to large US tech corporations? Why is the UK Civil Service so heavily reliant on procurement from major US technology firms such as Microsoft—the Minister mentioned Copilot—OpenAI and Google, rather than focusing on developing its own domestic AI tools and platforms?

I thank the noble Lord for his question. I have a full Question on sovereign AI on Wednesday, when I will answer that question in more detail, but in the meantime let me say that there is not a some inbuilt bias against that; it is just that many of the large language models are, of course, from US companies, and those are the ones that are available at the moment. However, the sovereign AI unit will use that £500 million specifically to stimulate UK companies as well.

My Lords—

My Lords—

My Lords, we have time to hear from the Cross Benches.

My Lords, while the results from the landmark Civil Service AI trial are clearly encouraging, does the Minister agree that it highlights the urgent need to train up public sector workers across all departments on the effective and appropriate use of generative AI? I suggest that such training and guidance apply in particular to us—by which I mean noble Lords on all sides of this Chamber.

It was interesting to see the report from MIT last week on the use of AI across companies, which noted that 95% of companies got very little benefit and 5% got massively disproportionate benefit. One of the reasons why you get much greater benefit is training people properly and allowing there to be proper disruption of existing workflows—so I completely agree with the question. What the noble Lord is talking about is an important part of this, which is why there is a series of schemes across the Civil Service, including the senior Civil Service, both to recruit people with AI skills and to train staff.

My Lords, I declare an interest as a former Minister for the Civil Service—although it was an awfully long time ago. The record shows that the biggest increase in output and productivity in the Civil Service and the biggest fall in numbers in modern times occurred between about 1972 and 1982. Of course, the driving force for that was not so much energy efficiency, although there were attempts to improve that, as removing whole industrial functions from the public sector. The Civil Service numbers fell from about 815,000 down to about 510,000 in 1982. That was an enormous cost. The lesson of that is that, if one really wants to increase productivity and slim the Civil Service, as I believe the Government do, there should be the removal of whole functions from the Civil Service—in this case, the industrial Civil Service—into the private sector.