Civil Service: Artificial Intelligence Productivity Gains

Lord Vallance of Balham Excerpts

Monday 1st September 2025 Lords Chamber

[The Minister of State, Department for Science, Innovation and Technology (Lord Vallance of Balham) (Lab)]

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.

[Viscount Camrose (Con)]

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?

[Lord Vallance of Balham (Lab)]

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.

[Baroness Winterton of Doncaster (Lab)]

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.

[Lord Vallance of Balham (Lab)]

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.

[Lord Clement-Jones (LD)]

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?

[Lord Vallance of Balham (Lab)]

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.

[Lord Londesborough (CB)]

My Lords—

--- Later in debate ---

[Lord Londesborough (CB)]

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.

[Lord Vallance of Balham (Lab)]

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.

[Lord Howell of Guildford (Con)]

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.

[Lord Howell of Guildford (Con)]

The question is: what areas are the Government going to look at now to remove functions from the state sector, which will be the sure way to increase productivity and reduce numbers?

[Lord Vallance of Balham (Lab)]

It is an area where AI is important, because AI does just that. There are certain things AI does to improve the efficiency of what is already done, and certain areas in which it does things that cannot currently be done. Both of those areas will lead to disruption of current workflows. This goes back to my previous answer: the disruption of workflow around AI is the big change management challenge.

[Lord Fox (LD)]

My Lords, my experience of introducing technology—not AI but other technologies—to large organisations is that there is a huge cultural aspect. I agree that training is very important, but does the Minister agree that, for AI to succeed, the entire workforce has to want it to succeed? How are the Government and Civil Service going to embrace the huge cultural change required to take full advantage of this technology?

[Lord Vallance of Balham (Lab)]

I agree, and I have alluded to that in previous answers. The challenge is a cultural one around workflow. I go back to the MIT report, which shows that, especially in big companies, there are high levels of adoption and low levels of disruption. The challenge is to get high levels of adoption with appropriate disruption taking place. That is a cultural challenge. That is why not only training but leadership are needed to make this happen.

[Lord Markham (Con)]

My Lords, we are all aware that probably the biggest challenge in government is productivity in the NHS, which has declined by about 20% since the pandemic. However, in none of the conversations and negotiations on the resident doctors’ dispute have we heard about productivity being a key part of the solution. Can the Minister reassure us that we really are trying to drive productivity gains, so that any wage increases can be fully justified?

[Lord Vallance of Balham (Lab)]

I thank the noble Lord. I can certainly reassure him that AI in the health service is going to be one of the big areas where changes occur. It is going to introduce changes in a number of ways. The first is in workflow organisation, which is key for the NHS to look at. The second is in new treatments and ways of diagnosing. It is already being used to reduce the need for humans to look at X-rays and so on to get them through more quickly, so there is an efficiency gain. There are also new approaches, such as using AI to make sure that remote monitoring—for example, taking photographs of suspicious moles and seeing if they are malignant—is potentially massively enhanced. There are many opportunities for productivity improvement through AI.