Public Accounts Commission
Oral evidence: National Audit Office Main Estimates 2026-27, HC 1615
Tuesday 21 April 2026
Ordered by the House of Commons to be published on 21 April 2026.
Members present: Clive Efford (Chair); Sir Geoffrey Clifton-Brown; Sarah Olney; Chris Vince.
Questions 1-36
Witnesses
I: Gareth Davies, Comptroller and Auditor General, National Audit Office; Dame Fiona Reynolds DBE, Chair, National Audit Office; and Rebecca Sheeran, Chief Operating Officer, National Audit Office.
Witnesses: Gareth Davies, Dame Fiona Reynolds and Rebecca Sheeran.
Chair: This is the public session of the Public Accounts Commission. I remind members of the Committee that parliamentary privilege does not apply to our proceedings. I welcome the witnesses to this session on the NAO’s main estimates for 2026-27. Could the witnesses introduce themselves for the record, please?
Gareth Davies: I am Gareth Davies, Comptroller and Auditor General.
Dame Fiona Reynolds: I am Fiona Reynolds, chair of the board at the NAO.
Rebecca Sheeran: I am Rebecca Sheeran, chief operating officer at the NAO.
Q1 Chair: I am going to start off. The NAO is seeking a small increase in funding next year, which, after accounting for inflation, actually represents a real-terms reduction. Given the expansion of its responsibilities and increasing audit complexity, is now the time to be squeezing the budget?
Gareth Davies: Let me take that first. Afterwards, I can hand over to Fiona, who can explain how the board has scrutinised the budget as well, which would be helpful.
My priority in setting the budget for next year is to ensure we can do our job to the highest standard that you would expect of us. We have had strong recent improvements in our audit quality results, as you saw at previous sessions, and we are determined to maintain that.
The big growth that you mentioned in our work has largely come through in previous years. Certainly, our audit methodology has now adapted to the increased demands of the auditing standards, so we are not expecting a big increase again in ’26-27. The volume increase that meant we had to come for a supplementary estimate last year—you will remember it was on the train operating companies—has largely worked its way through as well. The last few are coming through next year, but they are in line with our assumptions and we are geared up to deal with them.
It is also true to say that some audits are disappearing from our regime as separate audits, largely because bodies that were previously arm’s length bodies are being re-absorbed into Departments. There is not a big increase in our workload for next year; we intend to maintain our output of more than 60 reports for the PAC on the value for money side.
We are strengthening in a few important areas as well. We are strengthening our financial and risk management team because we are finding that to be a very fruitful area of work for us at the moment, and we need more expertise there. This is all in pursuit of our strategy, of course, which sets out to make a bigger impact on productivity, resilience and financial management in Government.
I can assure you that there is no let-up in our quality and impact here—on the contrary. There is a reason that we have been able to contain the increase in the way that you have seen. Figure 1, which is on page 6 of our estimate document, shows pretty clearly a £5.2 million increase in our costs—so we are properly accounting for inflation, pay award and the investment that I mentioned—but our income is going up by £4.5 million. That is partly because of the train operating companies and the full-year effect of those, as they are fee-paying audits, but crucially, the spare space in our building is now fully let, which is a straight £1 million of income without additional costs.
You can see how, by increasing our efficiency, we have been able to contain the overall increase for next year. Fiona, you might want to say something about the board’s work.
Dame Fiona Reynolds: Obviously, the board is very concerned with making sure that the NAO has enough money to do the job that we are asked to do, and particularly to deliver our new strategy, which, despite now being a year old, is still very much at the stage of ensuring that we align ourselves around impact and the twin outcomes that we set. It is demanding, but it is not all about money. Some of it is about the way that we spend our time and the focus of, in particular, our senior people on impact and engagement with Government Departments, Parliament and our external audiences and stakeholders.
We challenged the budget assumptions very hard. We are looking for efficiencies, just as every part of Government needs to. We are keen to show that we are aware of that and are responsible for making sure that we spend our money as well and efficiently as we can. But we do believe that we have the right level of resources to do the job, and that you and others will be proud of what we do.
Q2 Chair: You are confident then that you will not need to request a supplementary estimate during the year, as you did last year.
Gareth Davies: Yes. Now of course there has been a big change: the PAC meeting that considered this estimate was on 12 February; two weeks later, the war in Iran began. Of course, all the speculation is about a spike in inflation coming down the pipe from that crisis. To state the obvious, all this was prepared well before that, but at the moment, we do not have any better information to give you a different budget assumption for the coming year.
Built into this is an inflation and pay increase of 2.8%, most of which—2.5%—is pay increase. Only 0.3% of our budget is exposed to contractual variations affected by inflation. The risk of an inflation spike affecting our input costs from suppliers is very low. The bigger risk is that our staff face financial hardship in the event of a serious increase in prices, but we would then be in the same boat as the rest of Government—and, well, the rest of the economy.
It is much better for us to tackle that in extremis. If, once we have reprioritised our budget and driven out any remaining inefficiencies that we can find, we have to come back to you with a supplementary estimate because of some really unprecedented spike in inflation, we of course will—that would mainly focus on the pay level—but at the moment, we do not think it would be responsible for us to speculate about that.
Q3 Chair: How confident are you that you will meet your targets on the timeliness of your audit certifications?
Gareth Davies: There is a big push. We are right at the start of the year-end audit process now, and we have three months of intense work ahead of us. The teams are geared up and fully resourced for that, and they have reported that they feel better prepared than they have done for many years.
We are determined to hit our target of 70% of audits pre recess, which would be a big step forward from 64% last year. That is our key operational target. Obviously, that is while maintaining quality—I always have to say that, because we never put pressure on teams to say that the audit is complete when it is not, and we will always have the evidence necessary to give a sound opinion.
Q4 Chair: To move on to staff and pay, is the award of 2.5% across all grades, and is it linked to performance in any way?
Gareth Davies: It is a flat award across our grades. The reason for that is that two or three years ago, we restructured our pay scales, which built in automatic progression for those at the bottom of the grade until they get to the mid-point. We did not have that in the past, so we were often forced to skew our pay awards toward people paid at the lower end of the grades getting a larger percentage. Our mechanism now means that that is not necessary, so a flat award is clearer, simpler and effective.
Rebecca Sheeran: In terms of performance, the one exception to colleagues receiving the 2.5% is the very small number of colleagues who were assessed as not meeting our expected performance standards in the last year.
Q5 Chair: How do you incentivise high performance?
Gareth Davies: We set challenging targets. It is very clear in the NAO that those who are successful in delivering for Parliament against our objectives and making an impact are the ones who progress in the organisation. It is largely harnessing that desire to make a bigger impact, to take on greater responsibilities and to get promoted that is the driving mechanism for us.
Actually, my experience in both the public and private sector is that things like bonus schemes for high performance work best in the private sector, where you can demonstrate an increase in the income for the business because of somebody’s performance. That does not apply to us: however well somebody does, they are not going to increase the income for the NAO, so you end up squabbling over a fixed budget. I have always found the demotivating impact of that to be higher than the motivating impact of it.
In the context of a public body like the NAO, I do not think bonuses are very effective. What really does motivate and retain good people is good management, good leadership and a high performance culture, and I think we have that.
Q6 Chair: What has been the response of the trade unions to the proposal?
Gareth Davies: I cannot speak for them, but I can tell you that they have decided not to put it to a vote of the membership—in other words, to accept that we are going to be imposing it, in the language of that. It is not positive agreement, but it is not, “We’re going to ballot on whether we reject this or not.”
Q7 Chair: How would you characterise the current recruitment market of qualified accountants? Is retention an issue?
Gareth Davies: As you will remember from two or three years ago, high turnover has been a problem in the past for us, particularly in the newly qualified grade. It has been much less so in the last year and a bit, and it continues to be the case now that our turnover is lower than we expect in our assumptions built into the budget.
That definitely comes under the heading of a nice problem to have. Our experienced staff are staying for longer, which is good for audit quality and good for continuity and Departments working with our teams, but the reason it is a problem, although a nice one, is that it puts real pressure into our budget. We have the same model as the audit firms and we take on a substantial graduate recruitment every year. The places for those people are freed up by people moving on higher up the grade structure.
If people are not leaving, it puts pressure on the number of people we are able to recruit in a particular year. We do not want to artificially depress that because we need those people qualifying in three years’ time when the market has changed. Long experience says it is a very bad idea to cut back on graduate recruitment because you are seeing lower turnover higher up; you will regret that in three years’ time. That has given us a budget management challenge for the last 12 months, which we have managed successfully, largely by cutting back on our use of temporary audit resource, which is not a bad thing for quality and efficiency.
Q8 Chair: Your staff numbers are set to peak at 1,122 in this coming financial year and then start to drop in ’28-29. In what areas of the business are you looking to expand your number of employees?
Gareth Davies: The reason why the headcount has continued to go up next year despite the workload flattening off is that we are in the main year of our insourcing strategy, which we have discussed with you in previous years. For two or three years we have been steadily bringing audits back in-house and not contracting out to the firms.
There are a couple of reasons for that. The main driver for it is value for money. We have found that the fee rates charged by the firms have gone up faster than our own costs. Where there is no compelling specialisation reason for outsourcing the work—for example, the fact that it is a financial services-type operation where we have less of that specialisation than the firm—it is not good value for money for us to outsource straightforward audits that we can do ourselves. In fact, some of the savings that we are delivering are coming out of that process. But of course it does mean our headcount has gone up a bit because we are doing the work in-house rather than outsourced.
Q9 Chair: So the main driver behind increasing staff numbers is not the increased workload. You have already factored that in.
Gareth Davies: Yes.
Q10 Chair: It is the fact that you are bringing more work in-house.
Gareth Davies: We are doing a higher proportion of the work ourselves. We are going down from something like 93 outsourced audits. When this programme finishes next year, we will be in the 50s for outsourced audits.
Q11 Chair: Looking ahead to ’28-29, when staff numbers will start to decline, how do you intend to achieve that? Will that be through natural attrition of staff or will that be a redundancy programme?
Rebecca Sheeran: At this stage, it is important to put in context what we are talking about in the three years of the estimates that we have modelled. The numbers are quite low; they are roughly seven to 10 next year and then another 10 to 20 the following year. They are based on what we know about our work right now and getting productivity benefits from the investments that we have already made in improving our audit methodology and the technology that supports that. It is not around speculative estimates about future technologies or future changes we might make.
We will of course stay very open-minded and we will be looking for further opportunities, but we would expect to achieve the numbers we are talking about through natural turnover over the three-year period. Of course, if we find that we are looking at something a bit different in terms of the pace of change in how we work, we will re-evaluate that, but this is our best picture at this time.
Gareth Davies: The other caveat to add is that if you looked back and challenged us on the accuracy of our three-years’ time forecast for staff numbers, it would not be brilliant, because things change. This is based on everything we know now, but if there is a reorganisation of Government Departments and there are fewer or a greater number of audits, of course those assumptions will move on. That is just a bit of a health warning around the accuracy of those two-year-out figures.
Q12 Chair: How do you balance the workforce between Newcastle and London? Are there different challenges in each location in terms of retention and recruitment?
Gareth Davies: It is a good recruitment market for us in the north-east. We are also trying to increase the proportion based there—not dramatically, but as a percentage of the total—because it is an additional recruitment market, it is very good value for money because we are not paying London weighting equivalents in Newcastle, and we get brilliant people who are looking for this kind of work in that region. We have a bit of expansion space in our new office in the north-east as well. For those reasons, we are trying to increase the proportion gradually. It is currently under 200 people based there out of the more than 1,000 we have, so it is still by far the smaller office, but there is room for us to grow there.
Q13 Chair: With new technology moving on, is location becoming less of an issue?
Gareth Davies: It is tricky. There is obviously a big debate about this. You can do some of this work remotely, but good auditing does require you to actually be with your audited bodies. There is nothing like seeing the body language of the finance director to really judge, essentially, “How much can I trust the judgments that have been presented to us?”, or, “What is the atmosphere like in the office?” It is amazing how these things matter. Of course, it is very hard to put your finger on exactly how you do that, but the best auditors I know believe strongly that you have to actually spend time in the environment you are auditing if you are going to really understand what is going on there and get a sense of where the pressure points are and so on.
Q14 Sir Geoffrey Clifton-Brown: Can I come back to you, Rebecca, with one or two staffing questions? You implied that every member of staff was performance managed because otherwise you would not know which ones were not up to spec. Is that correct?
Rebecca Sheeran: Yes. As most organisations do, we run a formal annual appraisal process. We have a simple framework in the way that Gareth described, where staff are awarded either a “highly valued” rating or a “performance concerns” rating.
Q15 Sir Geoffrey Clifton-Brown: Just one or the other.
Rebecca Sheeran: Yes.
Q16 Sir Geoffrey Clifton-Brown: So some of them will soon get the message.
Rebecca Sheeran: Yes, and fortunately that is a very small number of staff. We have a highly skilled and engaged workforce—
Q17 Sir Geoffrey Clifton-Brown: Gareth mentioned that the main inflationary pressure is likely to be pay awards. Do you have an annual pay negotiation round? Have you done that already this year, and would you reopen it before next year?
Gareth Davies: We do have an annual round. As I explained earlier, it is quite common in our set-up for the union to neither accept nor reject the pay offer but to accept the imposition of it, if you like. I think that is the position we are in this year. Last year, actually, the offer was formally accepted for the first time in a long time. In the past, we have had ballots and rejections of the first stage of those things, but at the moment, we are expecting the 2.5% to be implemented in May salaries backdated to 1 April.
Q18 Sir Geoffrey Clifton-Brown: So if they wanted to go for an inflationary increase, they could not do so really for another year.
Gareth Davies: No. If there is a really exceptional spike in costs, there are precedents for an in-year increase on top of the one that we are talking about here, but that would be very unusual and it would have to be an extreme set of circumstances. I would envisage that only if something were happening more broadly, not just in the NAO.
Sir Geoffrey Clifton-Brown: Fair enough, thank you.
Q19 Chris Vince: I have been mulling over how to ask this question without trying to get a political view from you. We have spoken about the increase in the NAO’s audit responsibilities before, and a reason for that was bringing the railways back in-house. You have also touched on the fact that, previously, there were 93 outsourced audits, which you are trying to bring down. Without going into what we think the Government might do in the future, do you anticipate being asked to take on new audit responsibilities? Do you feel prepared for the potential risks that that might bring?
Gareth Davies: We have listed at, I think, figure 3 in the estimate document all the new audits that we know about for 2026-27, the year we are discussing here. There is also a possible list—there are only three on the possible list. That is where we are not sure about timing. We do not expect any of those to add to the number we have for 2026-27, because it is already April, so there is limited time to add any new work in for this year. It is worth pointing out that, at the end of figure 3, the last table shows bodies that are being re-absorbed into their host Departments and will cease to become separate audits. While we are still auditing the expenditure, it is now part of the overall Department. That brings down the workload for us, because we do not have to treat them as separate with lower materiality levels and so on.
There is quite a lot going on this year, as you can see from bodies coming in and going out. By far the biggest group is the train operating companies, as we have discussed before. Those audits are going well so far. This is our first year of them. We will be auditing several of them before the summer recess. We are confident that we can take on the rest, because we now know how those operations work. Other than that, there is nothing in the pipeline that we are aware of that could affect the year we are talking about.
Q20 Chris Vince: Equally, there has been some discussion—I do not expect you to comment on it—about the future of the Independent Commission for Aid Impact, the organisation that examines value for money of aid spending. Have you had any conversations about the potential of the NAO taking on more responsibility for examining value for money for aid spending?
Gareth Davies: From conversations with the FCDO, I understand that that suggestion has gone away—of involving us directly in some way. Of course, our remit covers aid expenditure, as it does all public expenditure, and we have some work under way at the moment on the humanitarian activity of the FCDO. The question was, “Could the NAO increase its coverage of that area?” if there was going to be less from ICAI. As I understand it, that suggestion has been withdrawn now and we are not being approached to do that, but we will continue to treat aid expenditure as important. Even though it has reduced, it is still a significant amount of public money and it will need proper coverage from the NAO.
Q21 Chris Vince: If that were to come back on the table, and I appreciate that you have said that you feel confident that the NAO would be able to deal with the additional work and have the special—
Gareth Davies: I am always reluctant to have our work hypothecated. We have scarce resources and a huge range of public spending risks to cover for the PAC, and in my view, the more discretion over that we have, the better. I am quite challenging to anyone who wants to hypothecate a proportion of our resources for a particular purpose, just because that would restrict our room for manoeuvre when a higher risk might emerge somewhere, in our view. We have a general insistence on our independent ability to choose where we audit. Of course, Parliament is responsible for setting our legal framework and our objectives, so if Parliament took a strong view, of course we would respond.
Q22 Sarah Olney: Gareth, I want to talk a bit about productivity, efficiency, resilience and, particularly, AI and tech, which you mention in the report. There is quite a lot of detail there about how you will exploit digital technologies, data and AI. Can you tell us how are you finding efficiencies in your VFM programme through utilising more of those new technologies?
Gareth Davies: Yes. This is obviously developing very fast—that is the first thing to say—so our aim here is to be what we call a fast follower, rather than a trendsetter. We want to keep up with developments in the wider audit profession, the consultancy world and Government.
The first thing to say is that we are making sure we understand how Government are deploying all of that, because part of the assurance job that we do for Parliament is to give a view on how well that is being done and what value for money it is achieving. That is the first thing.
On our value for money work itself, our first step into that is to acquire Copilot for every member of staff, having done some successful pilots ourselves. Everyone will be trained in using Copilot, for financial audit and value for money work. In value for money, it is essentially like having an extremely productive team member assisting the team. That is how the teams are finding it.
We would never rely on Copilot for the human judgments on value for money that our experts have to make, but it can speed up the analysis process quite radically. It can give us greater analytical power, and it can point us at anomalies in data that we would not have been aware of otherwise. We are seeing it as essentially augmenting the ability of our VFM teams to do high-quality work, to come up with new insights for the PAC and to accelerate the analysis that we did previously. It is an augmentation of our capability, rather than a replacement for any of the work that we did before.
That is slightly different from the financial audit side, where there is much more scope for automation of standard audit procedures and so on. We are very lucky that our relationship with the audit firms means that they are very happy to share with us their use of the technology, which again is developing very quickly. We have had some excellent sessions with some of the big firms, who have explained how far they have got, what is proving to be useful and so on. We are learning very quickly from that and working out how to scale that to an operation the size of the NAO.
One of the really interesting things about AI is that I think it is more scalable than previous technological advancements. It does not actually rely on paying a fortune to a software developer for a bespoke bit of software; it relies on developing in-house expertise to use these actually quite simple-to-use tools like Copilot. That is even so with—where the big money is at the moment—what is called agentic AI, using AI agents to carry out defined bits of a process, and then linking them up to speed up the overall audit process in this case. Some of the firms are close to deploying AI agents as part of their audit approach, so that is clearly where we are going; the question is how quickly we get there.
Again, as with everything, we will never take a quality risk on that. We will do this at a pace that we are confident we can maintain quality. The best way of envisaging this for the future—we are probably talking three years into the future now—is that, rather than managing an audit team who are using some software, we will be managing a combination of human auditors and AI agents to deliver the audit. That will apply to VFM as well as to financial audit.
Once we have got your head around that, it illustrates that it is just as much a people challenge as a technical challenge, because we will all have to be able to do it. We are being trained ourselves. There is no avoiding it. You cannot be an enthusiast capable of doing it while everyone else avoids it; we will all need to be capable of doing this. It will also need serious technical leadership, so we are about to advertise for an executive director to join the top team and lead on digital and AI, which will be a first for the NAO. I know the Chair of the PAC will be keen on that, because we have been exhorting Government Departments to do exactly the same.
You can see this is moving quickly. I cannot tell you exactly what it will look like, but we are now seeing the shape of it. Because it does not require very large capital investments and is essentially a capability thing for the organisation, I don’t think it is a question of, “Can we afford this? Can we come to you with an affordable budget?” I think we can. It is a question of, “Can we increase the capability of the organisation quickly enough?” That is the challenge that we are facing here, I think.
Q23 Sarah Olney: I hear what you say about AI augmenting rather than replacing, but what does the increased use of AI imply about the number of VFM analysts you are going to need?
Gareth Davies: We will see. Clearly, we will not employ anybody we do not need to employ. As I say, I would rather show the PAC and Parliament more generally what this does to our capability with the same number of people first and then give Parliament the choice: “Are you content with maintaining the budget at this level but with greater impact or, if there is an opportunity to reduce the volume of staff but not get the breadth of impact that we would otherwise have been able to have, do you want to take it?” There is a choice looming, and we would definitely want to engage Parliament in that choice.
Dame Fiona Reynolds: We had a discussion today in the board about digital, and I think the risk is that everybody sees AI as simply about efficiency and saving money. That was at odds with our sense that, as the NAO, we are here to achieve impact and to deliver for the country through our work. Using AI to enhance what we do is as important to us as gaining the obvious efficiencies that you can gain, particularly on the more prosaic and repetitive audit tasks. So I think in the value for money area, there are real and exciting opportunities about delivering better and more with the same resource or using AI creatively to deliver even better outcomes for the country.
Gareth Davies: I think the bigger impact on the shape of our team is on the financial audit side. I think the traditional pyramid with lots of trainees doing the detailed work, supervised by managers and directors, will become less like a pyramid and more of a flat structure. It will definitely reshape the trainee experience. We will still take on a significant number of trainees—maybe not quite as many as in the past, but still a significant number—but their experience will be very different from mine in the 1980s. For somebody fresh out of university, I spent an inordinate amount of time photocopying and chasing bits of paper—invoices. They will not be doing that, because it will be automated. They will be learning how to make audit judgments much sooner than I did and probably taking on greater responsibility for managing the AI agents I was talking about after just a year or two of working for us. You start to see how the jobs will change with this, but at the moment we are obviously in the foothills of all that.
Rebecca Sheeran: In the near term, therefore, our focus is very much going to be on equipping the people that we have now with the skills that they will need. Of course, part of that will be looking at our programme of training for new entrants, but equally important will be to equip all our other colleagues to be confident, responsible users of AI within the quality standards and frameworks you expect from us.
Q24 Sarah Olney: How heavily at the moment does the NAO rely on AI? Clearly, it is a changing picture from everything that you have said and still, I dare say, experimental—
Gareth Davies: Yes.
Sarah Olney: To the extent that auditors can be experimental. But do you see that developing?
Gareth Davies: I do. It is going to be unavoidable for any organisation that uses any software, because it is now appearing in just regular software. Your email system now has AI built into it, whether you like it or not. So has Excel, and so on. There is a level of that that is—some people don’t even realise they have it, but they have.
Our governance arrangements are having to catch up very quickly with this and make sure we know exactly what we are doing. We have good controls in place, we think. I keep mentioning Copilot, because that is the solution we have latched on to, but everything else is locked down so that you cannot use this on NAO systems. You cannot use ChatGPT, for example, on NAO systems, because we cannot be confident that it stays within our secure boundary. So all that is in place, and we will obviously keep it up to date as we go along.
It is even more important that our cyber-security is strong, so we are continuing to strengthen that. At the moment, we do not have fundamental reliance on tools other than Copilot and the bits that are built into the standard software. We are essentially doing two things at once. One is deploying these productivity tools quickly, such as Copilot, but we do have a team that is at the more advanced end of how we might automate our audit methodologies. I think we are a couple of years away from seeing that deployed across all our audits. We are trying to keep the two streams running together: improvements now, bigger steps forward in a couple of years.
Q25 Sarah Olney: You touched there on the risks of AI, which are clearly considerable, and you mentioned your governance arrangements. Can you talk a little more about what you perceive the risks to be and how you have adapted your governance accordingly?
Gareth Davies: Maybe Rebecca could explain the governance arrangements because she has managed the process for bringing them in. The risks are that we do not actually know what is going on in our own work, which is a fundamental problem. We are emphasising the point: this is humans using enhanced tools; it is not abdicating responsibility.
You will hold me to account for all of our reports, whatever the methodology we have used to get to the findings. It is no good me coming to you and saying, “I know there was an error in that Report, but it was because we used AI.” We have seen examples of that quite publicly, and that is not where we are going to be. The jargon is “human in the loop” but it actually means, “These are your audit judgments, as the auditor in charge of this job.” You must be confident that they have all been arrived at properly. Of course, we use tools to do that, but we need to know what they are doing.
That is the first thing; the other is the data security point I mentioned. The way a lot of the AI agents work is that you upload reports to them and they analyse them for you. We now know how to do that so that it is all secure within our boundary. That is the only way you can do it now in the NAO. That is another big point. Crucially, as I said at the start, we have got to understand how Government is using it. I am sure the PAC is going to want assurance from the NAO, as you have already been looking at DWP’s tools for tackling benefit fraud, where they have started to experiment with AI algorithms. Is that working and delivering the results intended? Is there any bias in these systems? Can we trust the reports it is generating?
I think that is really important assurance work for us to do. We are conscious that we are going to need to increase the confidence of our audit teams to be able to give you that assurance consistently. It is another area where we need to keep up with technical developments, so we can do that.
Rebecca Sheeran: In terms of governance arrangements, we are ensuring we are putting in place arrangements that are consistent with Government’s AI playbook. It is also really important for us that we are following the excellent FRC’s recently published guidance for the use of AI in audit. As Gareth said, what we are using mainly at the moment is Microsoft Copilot. We ensure we understand how that works. In terms of the risks, we are ensuring we have a human in the loop to manage risks around accuracy, quality and ethical use. That again is where our focus is going to be on training and upskilling people, and ensuring that every user understands how things work.
We have put in place an AI governance board to help us ensure that we have a structured approach for considering risk and responding. That also involves ensuring that we understand where generative AI is increasingly built into some of the third-party supplier offer, whether that is things we rely on for managing our digital estate or learning and development tools, which increasingly have an embedded offer.
There we make sure that anything we bring into the NAO estate complies with very stringent data security standards, for example. We have a series of other tests within our AI framework that third-party products must comply with. As Gareth has said, we are very much in the foothills, but we are looking to have in place robust governance arrangements to support us as we get into more exciting and ambitious territory.
Q26 Sarah Olney: I want to ask briefly about the insourcing that you previously mentioned. You are doing more in-house now that you are not outsourcing as many of the less complex audits as you were. How effective has that been at reducing costs?
Gareth Davies: Very effective, actually. The first year of that was last year, and we were able to demonstrate that it had achieved the cost savings required. Our work with the firms is done through a framework contract. We now know what the price of those contracts is for ’26-27, and we can see that it has delivered the reduction in spend on the firms that we were seeking, so we can be confident that that will be delivered.
Q27 Sarah Olney: Obviously, you are bringing more of these in-house. Does that mean you are building up a greater body of specialists to do your audit?
Gareth Davies: There are two things here. The audits I was talking about first were those where we outsourced the whole audit, and the ones we have brought in do not require particular specialism—our staff can generally take on those audits. We have a separate contract with some firms for specialist advice, typically in areas that are not directly audit—the three largest are valuation advice from surveyors, actuarial advice for our pensions work, and valuation of financial instruments.
As I have discussed both here and at the PAC, we are starting to challenge ourselves and say, “Well, there is so much of this work now, because of changes in auditing standards. Does it make more sense to do this in-house as well?” We are just about to appoint our first actuary, who is joining us on a part-time basis. That suits us as a proof of concept, because they will be able to work full time in our busier period, and part time when we do not need that kind of intensive work. We are taking on one very experienced person not only to do some of the straightforward work directly, which will save us those costs, but to help us be a better client for the bits that we still need to outsource, and essentially to get better value for money for what we spend on that.
Having an expert commissioner on that is going to be a big advantage to us. The plan over time is to see how big the team needs to be, so that we can be increasingly self-sufficient in that kind of advice. We do not know the answer to that yet, but the business case for that has us saving about £100,000 in the first year, so it feels like a good start.
Q28 Sarah Olney: Rebecca, you have already talked about the London office fully renting out the bits that you are not using, but at the same time you are planning to increase staff numbers. What does that mean for your staff, particularly in relation to their working conditions?
Rebecca Sheeran: You are quite right that we have increased the amount that we are letting out. We are currently letting out 10 floors, which is the target level for where we are right now, and that earns us the income that is set out in our financial plans. However, we have just brought an 11th floor to market, because we have now come to the end of the significant refurbishment of the NAO-occupied spaces.
That project was all about making better use of the space. While we are now letting out more of the building, we still have the same number of workstations in the NAO-occupied space, and we have also created opportunities for more flexible use of that space, such as teamworking stations and more small meeting rooms. It has not reduced what is available for NAO staff.
Q29 Sarah Olney: My question is more about the fact that you are increasing headcount, and whether you have the extra space for them.
Rebecca Sheeran: Over the last three months, across the two buildings in London and Newcastle, we have been at about 80% occupancy. There are no signals at the moment that we do not have the space we need for our headcount in the year ahead. Actually, we are seeing more people wanting to be in the office, because they can make better use of the space.
Q30 Sir Geoffrey Clifton-Brown: Gareth, I want to ask about small bodies, and the excessive burdens placed on them. You and I corresponded about a particular one that had a budget under £2 million, but it still had quite an excessive audit fee, as well as an excessive amount of time spent by the auditing director on dealing with the audit. You recommended in your letter back to me, as noted in your email, that “x has been considering making a transition from financial reporting under IFRS to FRS 102. This would enable x to apply the small companies reporting regime, which allows for a simplified presentation of financial statements.” At the time when we were discussing this, we also discussed with the Treasury that it might consider a relaxation of some of the requirements on very small public bodies that would not have this sort of reporting regime if they were in the private sector. How far have we got with all that?
Gareth Davies: As you know, the Public Accounts Committee is about to report its inquiry on that.
Sir Geoffrey Clifton-Brown: Yes, exactly.
Gareth Davies: I know that you are planning to recommend that the Treasury takes this as a serious opportunity. The core argument is that central Government is the only sector of the economy in the UK that does not have a small body financial reporting regime. Companies certainly have it, charities have it and even local councils have it—the smallest level of local authority at parish council level has a completely different regime that does not require an opinion on their financial statements in the way that larger councils do. Central Government is the anomaly here.
The exhibit that we included in our report on this issue showed that there are bodies with 16 or 20 staff who have essentially the same financial reporting requirement as a Government Department with tens of thousands. The argument for a more proportionate regime seems very strong to me. As you would expect, we try to adjust our audit approach in sensible ways to avoid overburdening small bodies, but they are all governed by the same auditing standards. There are no auditing standards for small bodies, because the auditing standards assume that the thing that flexes is the financial reporting regime and the rules set by that. That explains why small public bodies overseen by central Government are particularly problematic.
I mentioned earlier that one of the things that might help a bit is the trend to bring some of the ALBs into the host Departments and stop them being separate sets of accounts. However, obviously some of them need to remain independent ALBs, so the challenge still exists. We are working closely with the Treasury on the audit response to this as well as the financial reporting response. However, in the end, it would definitely make sense to say that although we want a set of accounts that properly accounts for the money, we can reduce the front half of the annual report which requires a lot of detailed disclosures and feels disproportionate for a very small organisation.
Q31 Sir Geoffrey Clifton-Brown: Going back to some of Sarah’s questions on value for money, I gather that you plan to look at a more modern approach to the publication of the traditional 10,000-word value for money report laid in Parliament and that that is going to become a thing of the past. What does all that mean?
Gareth Davies: I am not sure that I did say that. Are you quoting me?
Sir Geoffrey Clifton-Brown: I am not sure whether you did say that.
Gareth Davies: Given a second chance to express that, I would say that we get a lot of feedback saying that what causes problems for our reports’ users is being able to read them on mobile phones. They are not designed for that; they are PDFs that are easy to print off from our website. However, if you want to read the report on your phone, it is not configured. The technology exists to do that and that is our key aim. The feedback we are getting is that the people who read our reports expect to be able to do it in that way. However, that does not mean compromising it for existing users like the PAC, which will continue to have hard copies and whatever format is useful to its members. The key aim of making this more digital is that it works in whatever format you are accessing the report from the website.
Dame Fiona Reynolds: I would add that the board has been very interested in different kinds of outputs from the VFM process. Obviously, the main audience is the PAC, which, as Gareth has said, will continue to get reports in the traditional mould. However, creative summaries and more infographics and snapshot examples, which can be shared more widely to help raise public awareness of the issues and show that the NAO has solutions and contributions to make could be extremely valuable. It is an add on rather than a substitute in any form, but there are some exciting opportunities to convey the messages more clearly and in a way that is more appealing to a wider audience.
Q32 Sir Geoffrey Clifton-Brown: I would be careful on the infographics. I find that when you use them, I spend more time looking at the infographic on the page than I do reading the entire page because it distracts me—so be careful.
Dame Fiona Reynolds: I will take that.
Q33 Sir Geoffrey Clifton-Brown: I will say something else about Sarah’s questions on AI. The other day, when we looked at your Access to Work report, there were 150 pages of dense evidence that went with it, so I put it through Copilot. Within 15 seconds, I got a summary. I read it, which took me over an hour, but the summary was pretty good. That is the shape of things to come.
Gareth Davies: It definitely is. The trick is having confidence in the way it has been summarised. This is a real live issue for us. Everybody who publishes reports on their website is now finding that AI has replaced Google search and the traditional search engine approach. Instead of the search engine taking the reader to the report itself, the first thing they see now on the search page is an AI summary of what is in that report. That is a big deal for us, because we did not write that. We are not comfortable, because these are very careful, evidenced-based reports.
As you say, the summary will be okay. It might even be pretty good, but it will not be what we actually wrote. We are working with the people who understand how AI search technology works to work out how we can deal with that, because it is important that somebody has access to what we have actually written as well as an AI summary. I do not think that is a forlorn hope. More to come on that.
Q34 Sir Geoffrey Clifton-Brown: We on the PAC hugely appreciate your value for money work; a very high percentage of the reports are superb. I am not sure the Departments always think so. That is another matter, but they are agreed with the Departments. Obviously, your work is focused on the Public Accounts Committee—we make use of these things twice a week—but to what extent do you see it as part of the NAO strategy to support Select Committees more widely?
Gareth Davies: There is good demand out there for that. There is a well-established order of precedence, though: the PAC is the audience for our work. Sometimes I know you agree that another Committee might be better placed to take a particular one, either because your programme is full or for whatever reason, and that works well.
We have to be careful, of course. If we stimulate too much demand, we will struggle to meet that and deliver to the PAC what you are expecting, but we maintain very good relationships. I write to the other Select Committee Chairs once a year to offer a meeting to discuss what their programme is looking like and to share what we are doing, just so they are aware of it. Quite often, they will take our reports that you use as useful supplementary evidence for an inquiry that they are doing and, as you know, there is regular guesting in both directions between the PAC and the other Select Committees.
I would characterise it as a pretty healthy relationship. Some Committees make more use of our work than others, usually because of past experience and so on. The other thing that helps is our programme of secondments to Parliament. At any one time, around 10 NAO staff are on secondment to Parliament. One or two go to your Committee, but they go to other Select Committees as well. Earlier this week I met Ruth Cadbury, the Chair of the Transport Committee, and one of our staff is on secondment to her team. It is a pretty healthy set-up at the moment. There is obviously a limit on the total amount that we can offer to do directly in support of those Committees, but we are in good contact.
Q35 Sir Geoffrey Clifton-Brown: Given that there is this demand from other Committees, and demand on your time—you have been to the Labour party, and you are coming to see us tomorrow—using your excellent example of the infographic you produced on asylum seekers, which was absolutely first rate, would you consider devoting more resources to being able to do more for other Committees?
Gareth Davies: Of course. We are there to support Parliament; if Parliament decided that an additional 10% of value for money resource would allow us to better support Parliament, of course we would do that. That is a call for Parliament and you as the focal point for approving our budget. If you think there is a clear signal from Parliament for a bit more VFM work—particularly if we are right that AI will allow us to increase our impact for a given level of input—that is a conversation we should continue to have. This budget for next year is constructed on us maintaining the current level of support for you and, to a limited extent, other Committees.
Sir Geoffrey Clifton-Brown: Can I just thank you for what you do for the PAC? NAO people are always unfeignedly helpful; they really are.
Sarah Olney: Hear, hear.
Q36 Chair: Can I just chase you up about technology and the way it is developing? Have you factored in the increased costs of replacing equipment? Things must become obsolete much more quickly than they did in the past.
Gareth Davies: Actually, all our systems are cloud-based so we do not have big server rooms in the old model any more. We pay annual licence fees for access to those servers, so it is a very different model of IT delivery nowadays. The equipment we have is basically just our laptops and our connectivity arrangements: wi-fi in the building and our access to the remote servers. It is not really about an equipment budget any more. It is definitely about a software licence budget, and that is typically a revenue cost every year rather than a capital cost. The cost base is shifting, and we are engaging with Government, who are experiencing this in a big way at the moment, in our VFM work. We should be coming to you to say, “Here’s the licence cost necessary to use this technology in an efficient way.”
It is interesting, talking to the firms; they are describing that the bills they are sending to their clients are changing. The traditional audit bill used to say, “x hours of partner time and x hours of trainee time—here’s the total bill.” Now, they have a separate line for technology deployment on the audit. I was explaining that model earlier, where the audit will be delivered by a mixture of humans and AI. The bill is now starting to reflect that combination in a way that it did not before. Before, it was just an overhead included in the hourly rates for each grade. They are starting to charge their clients explicitly for the use of technology in the work. We haven’t started that yet, but it is interesting to see how the profession is changing.
Chair: Any other questions? No. That concludes our witness session. Thank you very much; we are really grateful for what you do and for your coming along to give us evidence this afternoon.