Business and Trade Committee
Oral evidence: Artificial Intelligence, business and the future of the workforce, HC 125
Tuesday 16 June 2026
Ordered by the House of Commons to be published on 16 June 2026.
Members present: Liam Byrne (Chair); Dan Aldridge; Antonia Bance; Chris Bloore; John Cooper; Sarah Edwards; Alison Griffiths; Leigh Ingham; Justin Madders; Charlie Maynard; Mr Joshua Reynolds.
Questions 127 - 150
Witnesses
III: Dr Alec Price-Forbes, National Chief Clinical Information Officer, Department for Health and Social Care and NHS England; Alyn Jones, Executive ITC Services Director, Somerset Council; Kate Jones, AI Policy Lead, UNISON.
Examination of witnesses
Witnesses: Dr Alec Price-Forbes, Alyn Jones and Kate Jones.
Q127 Chair: Welcome to a reconvened House of Commons Business and Trade Committee, as we finish our third panel on today’s session on diffusion of artificial intelligence. Thank you so much to our witnesses for joining us today. Kate Jones, why don’t you kick us off? Just give us a sense of how you think you see artificial intelligence being diffused through the public sector. We have heard a bit today from witnesses who have said that the public sector has a critical role to play in kick-starting and accelerating AI diffusion. Just give us a sense of what you can see right now.
Kate Jones: Our members tell us that they are experiencing two main forms of AI at work. First, which we have already heard a lot about today, is generative AI. They are experiencing that on a spectrum, where some are having narrow, domain-specific tools rolled out at work, such as EdTech—these tools being used to make lesson plans for teachers—all the way through on a spectrum to general-purpose LLMs, most notably Copilot, but not exclusively so. Tools such as that are being offered in a blanket offer to all staff to support with all admin and knowledge work. We would characterise the roll-out overall as distinctly immature.
The second form of AI we are hearing about, which is normally associated with the private sector, but we are hearing about it more in the public sector, is algorithmic management. That is primarily concentrated among our lower-paid, privatised and outsourced workforces. I am thinking in particular of care workers, energy sector call centre workers, and cleaners and porters in hospitals. We are aware of some others, which hopefully my colleagues here will be able to talk about, but things such as predictive analytics are not as visible to workers in their workplaces.
Q128 Chair: That is super useful. Dr Price-Forbes, tell us about the NHS. Where have we got to with the NHS?
Dr Price-Forbes: Where have we got to? The science fiction writer William Gibson said, “The future is already here, it’s just not very evenly distributed.” Both from non-AI technology but also specifically AI, that is what we are seeing. The challenge is not so much around the technology and AI. Our challenge is really around adoption and how we can now scale proven technologies that have a good evidence base, to make it business as usual from a standardisation point of view.
The positives are that we have one AI solution that has been scaled across the NHS. Brainomix, which is used to detect strokes on scanning and can accelerate how patients get through the service to get treatment, has been rolled out across all 100 stroke units in the country. That is 100%. It has demonstrable benefits. It has doubled the number of patients receiving thrombectomies and is tripling the number of those who have a stroke who can actually live independently afterwards. That is measurably impactful. There were studies done with nearly half a million people. We recognise that challenge with 600,000 strokes a year. That is being used more routinely.
The other big bets we are really focusing on, as Kate has referred to in terms of generative AI, would be ambient voice technologies, which have had quite a lot of press. That is the capability where a large language model and generative AI will listen to a consultation between a patient and healthcare professional in any setting and will create a transcript and then an AI summarisation. The largest global pilot of that was done in 2024 across London, across nine settings. It demonstrated consistent evidence of cognitive reduction in load, reducing administrative burden for clinicians: there was a 51.7% reduction in admin time, with 35% saying they felt less cognitively strained. In some areas, particularly shift-based services such as emergency departments, there was a 13.4% increase in throughput because they are able to go through and see those patients quicker.
The challenge now is how we have set up and leant into that, recognising that there were a multitude of different solutions being piloted across the service, and that remains the case. We have very much set out the standards for what we would expect from a business platform capability as well as a benefits assurance perspective. We set out a national model for AVT with a registry that has 23 suppliers. That has given a clear demand signal to suppliers—in the previous panels you were talking about SMEs and so on—of enabling that level playing field for market entry. We have real positives coming from AVT.
The final one I would mention would be AI triage, where there are some good studies. One particular study in Sussex using the NHS app is demonstrating a 4.5% reduction in GP access because it is diverting patients to the right information or the right service at the right time.
Q129 Chair: I remember when I was a young junior health Minister, I was struck one day when I realised that the NHS is actually an economy the size of Argentina. I do not know what the current stats are, but it is a very big and complicated place—a very big economy in and of itself. How do we think about understanding the extent of the roll-out of AI in the NHS? We could look at this by looking at a range of different specialisms, functions or trusts. How should we conceptualise how AI will roll out through the NHS in some kind of logical structure?
Dr Price-Forbes: For the stats, there are 1.5 million in the NHS workforce, and we have a similar number—if not slightly larger—in social care. It is important that we talk about health and care professionals, given the impact of wider social determinants on the lived experience for the patients we serve. The challenges at the moment have been by and large that the problems we are trying to solve are across the whole service, in terms of administrative burden, documentation challenges and just information challenges. We have been very successful in digitising the NHS, and more than 95% of acute trusts now have electronic patient record systems. We have not been good at actually leveraging the benefits that come from having higher-quality structured data, which we heard from previous panel members that you need for AI to be able to act upon.
The challenge now is how to set the standards from a national perspective. We are clear with the national AI road map that has been developed in order to support the Government’s ambition to become the most AI-enabled healthcare system by 2035. That has been built with health and care organisations, regulators, academia and industry. It very much supports how we not only return constitutional standards, but address the reform strategies of urgent emergency care, elective recovery, neighbourhoods, and wider Government priorities around economic growth, the life sciences plan, and sovereign technological capability.
We have positioned the AI road map around six pillars, recognising that different providers, systems and regions are coming at it from a different level of digital maturity and capability. That is in terms of the systems that are deployed, the infrastructure, the data quality they are getting, the interoperability of those solutions, and the leadership capability, particularly as we are going through the change at the moment with the merging and the diverting of funds and priorities to frontline services. We need strong leadership capability from the digital data technology workforce, which amount to only 3% of our workforce and are diminishing. We need really good workforce capability, recognising that according to studies, the literacy of the NHS workforce is slightly less than that of the rest of the UK industry.
Q130 Chair: You talk there about seeking to build sovereign capability. Does that then take you into difficult questions about the choices of partners? Obviously, there is a real issue in some parts of the public sector about, for example, whether to use Palantir. Is Palantir a business that is being used in the NHS today? Is it the kind of business that could be used in the NHS in the future? How would the NHS go about taking those sorts of decisions?
Dr Price-Forbes: Chair, you may be aware that there has been a separate committee this afternoon looking at Palantir as an organisation. As an official, I would not want to comment on the specifics around Palantir other than to say that AI needs high quality, structured data. The challenge that the NHS has suffered with for many years is fragmented, siloed data. We have digitised organisations, and then we have tried to share that data.
Our 10-year plan now is a shift towards how we build services and data flow around the individual in their community, to prevent them from needing to go into a hospital setting. My favourite phrase in the 10-Year Health Plan is that care should be, “Digital by default, in a patient’s home if possible, in a neighbourhood health centre when needed, and in a hospital only if necessary.” That is a phenomenal shift from the current model, which is that 95% of care is face to face, often within hospital settings. We need to bring the public, patients and workforce for that.
Finally, we have launched the National HealthTech Access Programme. The value-based procurement was published last week. We published an SME playbook back in February, and we are working on an Innovator Passport. Recognising that procurement has historically been a bit of a barrier, particularly for SMEs. SMEs make up 85% of the supplier base for health and care, we recognise the need to level the playing field so that, rather than just going for lowest-cost options at scale, we can look at better demand signalling to improve market access and work with industry, suppliers and SMEs to produce solutions that solve the problems we are trying to fix. We can look at how commercials and procurement can be changed to then leverage that, so we can lead to adoption and scale and the growth that comes from it.
Chair: Let us just round out the picture and hear the story from local Government.
Alyn Jones: Certainly, from a local government position, it is important to understand that it is not a homogeneous group. It is fragmented naturally by virtue of being very different organisations. My own council is significant in size but not necessarily significant in terms of the whole sector. What is critical is we are seeing local priorities come through, which are driving the adoption of AI, particularly in things such as the front door, optimising how we can interact with our citizens. It is also in areas such as predictive analytics, so falls prevention in adult social care, for example, or looking at things such as predictive analytics around highway defects in some areas.
More importantly, it is looking at those regular and repetitive administrative tasks and how they can be taken out, to support the workforce to make those decisions and support their views, so they can focus on the interaction with citizens rather than the bureaucracy of the activity. It is very much a mixed picture, depending on those local priorities. Everyone is going through different stages and governance steps, which is causing quite a lot of delay in the general adoption of some of the opportunities.
Q131 Sarah Edwards: This panel has been excellent so far in giving actual examples of the productivity gains being generated by artificial intelligence. I have been getting quite frustrated by just hearing, “It makes things better,” or, “We get more productive.” I want to know how. I want it to be described more. Thank you for some examples; as you can probably guess, I want to hear a few more, particularly on this measurability of the productivity gain.
I am going to go to Kate because I have already heard elements of the answer. I am going to give Alyn and Dr Alec some time to think about some more examples of how across the public sector we can measure and see the opportunities for really increasing, benefiting, improving, or whatever it is that we are looking to do. It might be decreasing, such as decreasing admissions to hospital. Kate, would you like to kick us off with some positive and hopefully great examples?
Kate Jones: I can definitely kick us off. I am not sure if you were here at the very beginning, Sarah, but my answers will be largely relating to generative AI because there are a lot of uses of AI, such as predictive analytics, that the average worker just does not come across in their day-to-day working life. That said, workers reporting measured and demonstrated productivity improvements is really rare. Members are much more likely to tell us that engaging with AI outputs is eating the gain from using AI tools. We have a bunch of different examples of why and how this is happening.
First, we are hearing about a rise in incidences of something called workslop, which is the work version of AI slop. Increasingly, people are complaining that they are having to handle vast quantities of AI-produced information that is coming to them in their workplace and really slowing them down. There was recently a report by the Harvard Business Review, which said that every incidence of workslop that is sent through to a worker takes about two hours to handle, unpick and make functional.
Q132 Sarah Edwards: Can you give us an example of what that might be? What are they receiving and why? What might that thing be? What does the slop look like?
Kate Jones: You might be someone who works in a local council and you ask for a policy write-up on cameras—there was something about cameras on roads earlier—and what you get is someone has panicked and said, “I don’t have the time to do this,” and typed into Copilot, “Can you please write me a report on cameras?” You get this technically quite good-looking, well-formatted, fluent English response, but it does not actually engage with any of the local issues, often is incorrect in a bunch of different places, and generally has drawn its opinions from across the internet and not from the specific context that is needed for you to do your work.
Beyond AI slop, the other thing we are hearing is that when specific tasks are automated, they are not being eliminated, but just pushed on to other people. We are hearing this a lot in the NHS from our members who predominantly work in administrative and operational roles. Increasingly, their jobs are being automated by AI and generative AI. They are often being redeployed, but they are reporting that their jobs still need to be done because there is a lot of work that goes into them that is not automatable, such as human judgment, chasing things across the NHS’s very fragmented systems, and safeguarding referrals. Instead, what we are really hearing is that clinicians are having to pick up that work, even though they are already overstretched.
That is not to say that there definitely are not productivity gains to be made, but the picture is a lot more complicated than the average study or figure would suggest. Two things are really clear to UNISON. First, if there are to be productivity gains, they will come from targeted use. That is when a tool is rolled out and introduced with a specific, ideally worker-identified use for it in the workplace, and where workers have been consulted, trained, and supported in the use of tools.
It is also critical that any benefits are going to be really context dependent and deployment needs to match. A blanket, one-note roll-out of a tool is not going to produce any productivity gains on a large scale. That is why UNISON would advocate for worker voice to be integrated into any roll-out as soon as possible, so that the people who actually use the tools can make sure they work.
Q133 Sarah Edwards: That is fantastic. That leads me to my follow-up to Alyn. You mentioned earlier that highways defects were one application of AI and the front door was another. I am guessing that means when people come in with an issue around council services or something such as that. Have you incorporated that worker voice into the roll-out? Have they influenced the tool, or have you brought that in-house? Can you speak to how it has measurably improved what it is that you were trying to do or the problem you were trying to solve?
Alyn Jones: Yes, of course. I have a better example in the creation of education, health and care plans. We know there is a significant backlog in the system. We know there are significant issues with that. We found that by using the evidence, guidance and skillsets of those who are administratively compiling that information as well as the professionals who are involved in judging them and interacting with the parents, child and school, by using and developing the tooling with them rather than it just being a technology-first approach, we are actually starting to understand the problem we are trying to solve: what the issue is, what the delay is, and what the concern is. We can then respond to it in that way.
Chair: We have a vote in the House. Are colleagues able to come back after this one vote? If you do not mind, we will just go and perform our democratic duty for the one vote, and if you are able to hang on, we will come back and just finish the questions. That is incredibly kind of you. Thank you.
Sitting suspended for a Division in the House.
On resuming—
Chair: Welcome to a resumed Business and Trade Committee inquiry looking into the diffusion of AI. Sarah Edwards is going to continue her questioning.
Q134 Sarah Edwards: Alec, now that we have had a short breather, perhaps you can give us more examples of how right across the NHS—as you are dealing with it on a slightly larger scale—there are obvious productivity gains. You already gave some really interesting examples around triaging and how you actually can perhaps help those smaller companies by giving that demand signal. Could you give us a little more colour and flavour around what that looks like?
Dr Price-Forbes: There are two good examples, which would be very much point solutions. First, there is one around teledermatology. Skin Analytics has a product called DERM that was developed in this country. It is a class three medical device. It is the first in the world to be autonomous, and it is 99.7% accurate in differentiating between cancer and non-cancer. That has been demonstrably evidenced and piloted. It is now in 20 sites live, and there are 600,000 patients being referred for queries of cancer. That was a 10% increase in demand in 2024. That is addressing the demand challenge.
Q135 Sarah Edwards: Before you go on, let me just check on that one. This is a piece of medtech developed by the NHS with a partnership—or was it developed internally? There used to be an entire department that actually created these things and then sold them globally. There was actually a revenue-generating arm to the NHS, which I think got scaled back quite a lot. Do you think there is an option for this type of thing?
Dr Price-Forbes: The AI Lab was set up in 2019 to look at the whole process around funding. It was funded to the tune of around £140 million. It was closed down last year. That was very supportive in accelerating some of this work looking at capability from SMEs, that demand signalling and so on. That is something we are looking at.
The national AI road map I have laid out is very much centred around six pillars: making sure that we look at governance and standards, data and infrastructure, the workforce and the skills. I would apply that to public and patients as well: how we create the evidence base and evaluate really robustly in parallel as we go along; how we do adoption and scale really well; and how we get partnership and innovation.
That will enable us to do it once, have a do-it-once philosophy from the centre, but also enable different providers, regions and systems, based on their own digital maturity, to flex innovatively to be able to do things where it is appropriate to do so. That is the standard we are setting through the AI road map to create that growth mission as well as improving standards around the quality and safety of care being delivered.
Q136 Sarah Edwards: It would extend to medical device technology development, using that NHS data and piloting options? You obviously have the ability to pilot that.
Dr Price-Forbes: Yes.
Sarah Edwards Great, that sounds excellent.
Dr Price-Forbes: We have also seen that diagnostics is another big field, going beyond dermatology. However, we are really focusing on areas where it is really important around breast cancer, prostate cancer—there were announcements last week by the Secretary of State—and lung and bowel.
The AI Diagnostic Fund looks specifically at lung cancer. That has led to basically a 50% reduction in the time it takes—eight days down to four days—to make a diagnosis, of either confirming somebody has lung cancer or refuting it, which speeds up their pathway. That is really important at a time of uncertainty for patients.
Q137 Sarah Edwards: You mention only lung cancer. Is there a particular reason why that was singled out to apply the AI?
Dr Price-Forbes: It was because it was looking at the imaging. It was very much around the diagnostics and using AI capability within imaging, a bit like the stroke detection that I mentioned earlier with Brainomix. These are real opportunities. It is important to say that we need a system-wide lens. The problem we are trying to solve here is that historically we have had a paternalistic NHS. I am still a practising consultant. I was clinical yesterday, both in person in the hospital and doing a virtual clinic last night. What patients have told us consistently—I was seeing new patients yesterday—is that they do not want to repeat their story.
When we talk about ambient voice technologies, that is great as a point solution to summarise and reduce the administrative burden I have when I am seeing a patient. However, there is nothing particularly innovative about a patient repeating their story 12 or 13 times in different care settings, before they see me at nine months as a specialist and I make the diagnosis. We are looking at how we can almost start creating a computable asset of the patient voice. That is why the NHS app is so central to our tech plan. You will hear more of that in coming weeks.
We want to really enable AI capability and voice capability within the NHS app to start leveraging not just the potential of patients’ journalling but for them to be able to tell their story. For that to be then high quality, structured data upon which AI triage tools and so on can signpost them to the right information or service is a real reimagination of what we are doing at the moment with the health and care system, and it is really exciting.
Q138 Justin Madders: I would like to come back to some evidence you just gave, Kate, about the situation when new technology is rolled out. It seemed to me from what you were saying that not only does this sometimes have quite significant impacts on people’s work, tasks, and how they are utilised—it may have negative or positive effects, and obviously there is a debate about whether there should be some proper guardrails about consultation there—but you also seem to be suggesting that quite often the employer is rolling out the technology without really giving a proper guideline as to how it should be used. They just say, “Here’s some new technology,” and they are expecting people to trial and error. Is that a good characterisation?
Kate Jones: Absolutely. I want to say that is not universal at all. The public sector is huge and there are some really good examples of AI being used. However, generally speaking, we are seeing a lack of consultation and training, which is really quite scary for the future of public services if AI continues to move at the speed it is moving at. I am trying to think of what the best example would be. I actually have a positive—I should not say that in a surprised tone.
Social workers are really beginning to benefit from a roll-out of AI transcription tools in a bunch of ways. First, they are saving time on documentation workloads, which is historically a really big problem for social workers and one of the biggest things making them leave the career. That then means they can spend more time in the case meeting focusing on the person in front of them, reading their body language, and hopefully having some time freed up afterwards to spend thinking more about the case.
We think the reason this technology has been so well received by social workers is that first, it fills an identified need that workers themselves asked for and said, “This is somewhere where we’d really benefit from automation.” Secondly, a lot of the tools being rolled out in social care are built for social care. They are tools that have been built with the workforce as well as the specific tasks that are going to be done in mind. Some have consulted workers even during the development of the tools themselves. Things such as the data they are trained on are specific to social work, so they just end up being a lot more effective and working a lot better. That is really quite a distinct example from what we are seeing across the public service.
Q139 Justin Madders: That is good to hear. It sounds like it should be a template for best practice. In terms of social workers using this software, obviously we know AI sometimes makes mistakes: it hallucinates, it sometimes has inbuilt prejudices. Obviously social workers are making some very, very significant life-changing decisions. Are there any checks and balances to make sure that these software instruments are working in a way that does not cause any unnecessary difficulties?
Kate Jones: In a word, no—not really. Liability for mistakes made is a big concern, not just in social work but in medicine. The Medical Protection Society put out a statement last week saying it is really concerned that doctors could be liable for medical malpractice if they do not spot errors made by AI in diagnostics.
Specifically in terms of social work and other regulated professions, AI is just moving so quickly that the liability structures and regulation are not keeping up. Workers are really being put in the firing line. We are already hearing about workers facing disciplinaries because they have not spotted mistakes in AI tools. While that liability and relationship really need to be worked out, the cases in which we are seeing those mistakes go unspotted are when workers are not being given time back from time savings.
Q140 Justin Madders: Just on that point, presumably if you have a workplace where someone is disciplined for a mistake that AI has generated, you are advising your members to double-check everything, and therefore any productivity gain is lost?
Kate Jones: Yes, absolutely. The checking is a big time sink for time savings.
Chair: I am going to speed up just because we have lost a bit of time this afternoon.
Q141 Dan Aldridge: You have pointed to the significant responsibility of public institutions and public servants in this space. One of the questions that I am really interested in is what the biggest constraints for our public bodies are in adopting AI. The Local Government Association highlighted a number of significant problems around lack of funding, lack of staff capabilities and lack of staff capacity. It went on to say that these pressures limit councils’ ability to move from experimentation to scaled deployment. We still expect a hell of a lot from our local authorities. If you could expand on what you think are some key barriers and what Government can do to try to mitigate that, that would be great.
Alyn Jones: As colleagues have just said really, we are seeing that those colleagues who are really important in the development of these tools that make processes more efficient are exactly those colleagues who are under the same workload pressures and caseload pressures that need to be addressed. It is the balance between those two elements. Funding is often directed at the development of the tool, not necessarily the collaboration that is required in terms of the development of the tool and how it is used.
That narrow focus around the use and development of the technology often misses the wider requirement for that support to come through. That would ensure that councils, in this case, have the benefit of a buffer, an opportunity to invest in the services and staff, create that training, and create those proper policy frameworks that enable decisions to come through. That is one area that can certainly be supported.
Central Government control of this is not something that the local government family would want to see coming forward. We would like to see enablement. We would like to see more sharing of best practice, more sharing of those ideas, and providing that pooled resource whereby we can all pull on the experiences. We are all going through the same experience, just at different times to reflect local priorities. If we had the ability to have that pooled experience and benefit from that and then be able to share that knowledge and experience across the sector, with that Government support, it would mean that we could adopt some issues and address some concerns far more quickly and get to the real crux of the point, which is supporting our residents.
Q142 Chair: You do not have that today, Mr Jones?
Alyn Jones: We do not have it in a wholesale way. We have individual investment and individual support associated with particular technology developments and working through individual areas. We do not have that widespread support of the technology, supporting councils to go through that level of change. That is something that could be relevant, and we see that in a number of examples from the Local Government Association. What has just been quoted is exactly a case in point. One of the key areas is that procurement guidance often focuses on the development and adoption of that particular activity. It is very difficult in a standard traditional procurement model to anticipate what technology will be used.
Chair: We are going to come back to procurement in a minute because that is so important.
Q143 Dan Aldridge: There is a hell of a lot there. There are about 19,000 elected councillors and however many hundreds of thousands of people working on this. Are there enough leaders driving this? We could argue that with the 650 MPs that we have here, we do not have sufficient leadership at the scale we need to drive some changes. Do you see the leadership within the sector? Is there anything that you would recommend that we could do?
Alyn Jones: Absolutely, we can see the leadership across the sector. We can see those areas where they are really utilising the technology to the benefit of the resident. We can see those examples. It is just not widespread. It is just not something that is part of the standard toolkit of a local government officer. We need to make that part of the standard approach, so that we ensure we are using every single tool in the toolbox to address budget pressures, workforce pressures or the residents’ priorities that are coming through. That is the skill we need. There are some incredible case studies, which I have just outlined earlier. It is about using those case studies, rolling those out and showing how they could be beneficial to other areas.
Q144 Dan Aldridge: When it comes to the NHS and the controversy around Palantir, one of the issues that it had been brought in to try to fix was that legacy technology; the legacy data architecture. Is the public sector limited in its capacity to adopt widespread AI by some of these issues? One of the things I really want to unpick is, with Palantir having brought in the Federated Data Platform, what the risks are of removing that. Those are the questions that really need to be fleshed out in public for people to understand.
Dr Price-Forbes: If we just focus on what the Federated Data Platform is already delivering in terms of benefits, it has been rolled out in and signed up by 35 integrated care boards and 170 trusts. Some 139 trusts are live and 137 are now reporting benefits. That means that over 94,000 patients on a cancer journey have had that cancer journey improved by 6.8%. We have seen hundreds of thousands of patients safely removed from waiting lists, where it is appropriate and clinically relevant.
It is already demonstrating benefits. It has a 10-year programme profile of benefits of £2.4 billion over that lifetime, with £646 million of that being cash releasing, and it has already released £217 million of monetary benefits. It is there. The wide point is the barriers to adoption and the scale, with growth being at the intersection of that and being the sweet spot. That is the challenge we are facing.
The barriers to that have historically been skills and workforce. We have not invested enough in people and change management. We have known for decades of the productivity paradox. For 30-plus years, we have talked about how industries take a while to adopt new ways of working when they are being digitised. We have been at fault in some respects for digitising what we were doing on paper. That is not particularly innovative.
We now have a real opportunity to leverage the assets we have already invested in; particularly, as I said, more than 95% of trusts now have electronic patient record systems. We need to really optimise their use, which means investing in people, making sure we do change management, and using behavioural science and human factors to get hard-pressed health and care professionals using the solutions in front of them, which will enable us to surface higher-quality data. The Federated Data Platform is that data orchestration that can start creating those insights, from which you can then act and intervene in different ways.
We need a data platform. That is the data platform upon which we can then start leveraging in the future the utilisation of AI, because AI needs good quality data. The FDP is not an AI platform, just to be clear, it is a data platform that brings together fragmented silos of data but then enables us to utilise that in different use cases.
Q145 Dan Aldridge: Are you saying that that is at risk by removing Palantir from the system?
Dr Price-Forbes: We will need a data platform.
Kate Jones: Our members are raising concerns about Palantir to us day after day, and increasingly in recent months. In terms of what you said, Alec, there was a report published in the FT this morning about how the campaigning group Foxglove did some FOIs on the actual impact of the FDP. A lot of its claims of service improvements are coming from a very limited number of hospitals and being extrapolated across all the UK. That is just one thing I wanted to say: members are telling us that they do not believe the FDP is the right tool for the job.
In terms of Palantir more broadly, as procurement goes on and it is considered for more and more contracts, such as the Single Patient Record, UNISON members are incredibly concerned by Palantir’s actions in Palestine and with US ICE. They are vehemently opposed to further contracts being signed with a company that does not reflect British values. We would say that public spending should be building the kind of society we want to see.
Dan Aldridge: The question we do not have time to go into now is around the values at the heart of it. Whatever either of you say, where is the truth? That is what we need to get to.
Chair: Let us come to this question of procurement though.
Q146 Sarah Edwards: It is a good place to segue on to that because obviously, £1 in every £6 of Government spending is on procurement. It is a huge possibility that we can leverage that spending base to make sure we are getting the right services. How do you view the current procurement processes that the Government have? We heard earlier about how the local Government services and NHS are quite rigid structures potentially—that is how they have been described—but not necessarily.
In theory, they may not work so well for AI procurement because maybe we have not amended the way we articulate the types of things we are trying to procure to the types of problem solving that AI could do. Perhaps I will start with you, Dr Alec, because you described how you had set up some systems and you are enabling that. Maybe you are starting to create an alternative route. Maybe you can talk us through that, and then I will ask the other panel for their thoughts on it.
Dr Price-Forbes: We are recognising that we have had a fairly fragmented ecosystem for procurement, which has made it very difficult for market access. It has made it difficult for commercials and procurement because the skillset for that within individual providers or regions has been quite difficult. There is therefore a real logic to doing that at the centre and really having that capability and skill mix. That has led to the problem, which to a degree Kate has just referred to: whatever the source and provenance of the data platform or any technology solution we are using, adoption and scale have been our challenge. That is a global challenge; it is not just a UK challenge. How do we really fundamentally do change management?
We are really grateful to the Treasury. We have had an uplift. There will be announcements on that. We will be really focusing over the next four years on change management, not just acquiring assets and technology. I mentioned that the National Healthtech Access Programme is very much about levelling up and trying to create a better playing field for market entry, so that we can have really clear demand signals around the problems we are trying to solve for the health and social care system. It is really a shift towards values-based procurement, rather than just going on what the lowest cost base is without any proven definite benefit downstream.
Q147 Sarah Edwards: Can I just check, is the current model from the Government, where they almost give you that pro forma saying, “This is how you procure services or goods and things,” quite restrictive, or are you able to almost create your own? Are they starting to reformulate that approach, or is it something that you have just done separately from that? In other sessions we have heard that the Government procurement system is quite rigid and inflexible. For example, if you are applying, not an AI model, but a digital system that is static, it works. However, if you are applying an AI model that is moving and needs constant management, it does not quite fit in that sense.
Dr Price-Forbes: Yes. This is creating a more fluid situation. Historically, it has been quite restricted. The initial findings from the National Commission into the Regulation of AI in Healthcare were published last week, and there is a 10-point plan around that. Related to the point around shared accountability that was being referenced on liability, the plan says that we need to be looking at manufacturers and providers as well as clinicians and regulators, so that we can have shared accountability across the life cycle. It is a life cycle, because we recognise that what we are deploying today will materially change, because that is the nature of AI.
We are very much looking at how we can level up and create that growth market for market entry, but get commercials and procurement better towards values-based procurement, so that we can get adoption and scale and deliver benefit. The phrase we use is, “Everything we do should be safe, assured, and delivering benefit.” That is very much the model registry we did for ambient voice technologies. It was very much about setting standards for the first time around the question: what are the business capabilities in terms of ensuring clinical safety, data security, cyber-security and compliance against standards, but also what are the benefits? Where is the evidence for the solutions we are deploying, so that we can then scale up and get benefit for wider society?
Q148 Sarah Edwards: Alyn, how is it applying in your part of local government? Do you feel like you have the ability to procure the types of services you need? You certainly mentioned and gave examples of how it is being used currently. Is there anything you would change or improve on that, or do you feel you are being given some freedom to be able to pull in those AI-type tools and deploy them without too much friction?
Alyn Jones: The flexibility is there within the procurement guidance but historically, and to a point today, local government still are buying products. They are not buying platforms. They are not buying the integrated approach that is there. The approach is, “There’s a problem. We’ll buy a solution to that.” We need to see a move away from that into more values-based procurement. Those partnerships and platforms move us into that place. The challenge therein is the measurable benefit. Often local government is judged upon whether the thing that it is replacing is cheaper. Is it going to still deliver the outcome that is really important for the resident, but at a lower cost? Cost is certainly a significant concern for a number of authorities that are very close to being financially unstable.
That is part of the challenge. How do you reconcile an investment in a platform or a partnership, where those measurable benefits will come through much later down the line, with something that is functionally appropriate for the here and now, but does not give you any future-proof approach to that? We have to move to a way of thinking about how we are assuring ourselves of what we are procuring. What are the steps we are going through, rather than the nature of the procurement process itself, and what is the outcome we are trying to get to as part of that?
The current procurement rules really narrow that engagement. They focus purely on, are you buying the right thing? If you are buying the right thing, you are then held to account of making sure you deliver the right thing. With the pace of change, we had a platform that was operating in February that was not functioning. Today it is working brilliantly. That is the difference. The products and platforms are moving so fast that often it is iterative, and they develop so quickly. Those individual products do not have that flexibility. That is the concern we have with the procurement rules at the moment.
Q149 Chair: I am just conscious that we are about an hour over time, so I am keen to wrap up very quickly. I just want to ask two questions really. First, do you share our view that the public sector could be an accelerant for diffusing AI? Secondly, what is the No. 1 recommendation on your wish list for Government? Alyn, do you think that the public sector is a potential accelerant for diffusing AI?
Alyn Jones: Absolutely. When we look at the collective buying power we would have as a local government sector and the challenges we have, we are ripe for that opportunity to be able to rinse it through.
Kate Jones: The public sector could definitely be an accelerant of good and responsible AI. Procurement is one route for that. We would particularly call for statutory bargaining to include AI when it is introduced to a workplace.
Dr Price-Forbes: Yes, I absolutely agree. We need to reframe the relationship around AI towards augmenting and supporting staff as a decision support tool, not replacing them. It is going to be very much more around oversight, judgment and validation of tasks, so that when I have done a consultation and I have said I am going to request an MRI scan, refer somebody for physiotherapy, request bloods and prescribe a drug, that is presented for me to validate.
There is also something about how we start moving to focusing not just on productivity-enhancing AI, which will reduce cost, but on innovation-focused AI, where we can start getting people, processes, partners and platforms really operating in a mutually supportive feedback loop with a flywheel effect, where you start creating value and generating growth. That is going to be super exciting for health and care.
Q150 Chair: What is your No. 1 ask of Government to help with this?
Dr Price-Forbes: We have to focus on people and change management. Trust is hard earned and quickly lost. We have had a challenge with the productivity paradox with technology we have put in already, where we did not train staff or—dare I say—public and patients. We have to come on a journey. We have to make sure we are transparent and accountable, and we bring people on that journey by making it really clear what we are trying to do and the problems we are trying to solve.
Kate Jones: Broadly, worker voice in AI and the public sector.
Alyn Jones: Co-ordination and change. They are the two things that are critical in order to make sure we are getting that done.
Chair: Unpack that a tiny bit for us.
Alyn Jones: We need the co-ordination from central Government and the associations that support us, to make sure that we are understanding how things are developing, that we are learning from best practice wherever possible, and that we are using those examples and can share that information more readily. We have already heard that the critical part of this change is the workers and staff. We need to ensure they are supported from the leadership right the way through the organisation, to be able to adopt these new ways of working.
Chair: Great. I cannot tell you how helpful that has been. Thank you so much indeed, both for your patience and for your expertise this afternoon. That concludes this panel and that concludes this session.