Episode notes
Tom Bartlett was Deputy Director of Data Engineering at NHS England, where his team built the national layer of the Federated Data Platform — the Palantir Foundry deployment now live across 137 trusts. With the contract's break clause in the political spotlight, Tom does something rare for an insider: he steelmans both sides properly.
The case for triggering: critical national data infrastructure concentrated in a single supplier with real geopolitical and reputational risk, where every year of delay deepens the dependency. The case against: two decades and eight failed national data strategies finally produced something that works — and there is nothing to switch to.
I'm not here to tell you your politics. I'm here to figure out what this contract really is, and if the NHS has any credible alternatives.
In this conversation
- Tom's mental model for the Federated Data Platform: it's not a data warehouse, it's an operating system — Foundry is the iOS, trusts and clinicians build the "apps" on top, and that's why comparing it to Databricks or Azure misses the point.
- The £330M price tag looks smaller once you compare it to the alternative: NHS trusts are already burning comparable money on fragmented, duplicated data warehousing and EPR spend — he estimates the true total cost of ownership nobody's ever added up would be "a scary number."
- His clearest technical explainer of Palantir's "ontology" — the mechanism that lets the same data layer handle both surgical front-line edits (like a banking transaction) and big-data analytics simultaneously, which he says no other vendor has replicated.
- A direct, first-hand rebuttal to the "Palantir is selling NHS data" claim made by MPs in the House of Commons — including the GDPR controller/processor distinction and the collateral damage: opt-out rates up 5-10x since January, skewing future research data.
- The rapid-fire close is the sharpest part: his one-sentence pitch to a skeptical clinician, his warning to a skeptical MP not to "tear down the only strategy that's ever worked," and blunt advice to Palantir executives on why their defence-industry language alienates NHS staff.
Transcript
The strongest case for triggering the break clause is that the NHS has concentrated a critical national data infrastructure in a single supplier, with significant geopolitical, reputational, and political risk. The longer the contract runs, the deeper the dependency becomes. If you're going to make a change, the break clause is the least disruptive moment to do it, because waiting longer just increases that switching cost.
And if you think UK sovereignty over health data infrastructure is a strategic priority, then the short-term pain is worth the long-term gain of independence — wherever you've heard that before. That's a coherent argument. I don't agree with it, because there's nothing to switch to, but I understand why people hold that view.
And Tom, what's the biggest argument for keeping with Palantir and not triggering it? If you're really trying to convince someone, how would you convince them?
I think it's important to look back over the last two decades. There have been eight national strategies and we're still in a terrible mess with data. We've finally got something that works after all these years. We've got 137 trusts live on it. The programme is rated green. And these first six products are just the very beginning — you're going to have so many products that clinicians will innovate and produce.
If you trigger this break clause, you go straight back to square one. We've got no place to go, with hundreds of millions already spent, and you've got 137 trusts reverting back to spreadsheets, whiteboards, WhatsApp. Show me an MP who wants to explain that to their constituents.
As deputy director of data engineering at NHS England, you led the build of the Federated Data Platform with Palantir. I'm really curious what kinds of problems it can solve — I think a lot of people don't actually understand what it is. Can you explain it in a very simple way? What does it do?
My team was involved in building the national part of the Federated Data Platform. The FDP consists of various components, one of which is Palantir's Foundry, and Foundry is installed in every organisation — an install for NHS England, one for each trust, one for each ICB. Each area's got their own tenant, and within that, each organisation is the data controller — they've got control over who can access it and what it's there to do.
When I was talking to my engineers about it, a lot of them said, "Why are we using Foundry? We've already got Databricks, we've already got Azure — we've got all these best-in-class systems. Why are we moving to this?" And if you look at it just as a system on its own, that's a natural question. Why use it for data warehousing? If you're just sticking in that analytical space, it doesn't offer that much more.
But when you start looking at what it can do on the front line — I like to think of it like an operating system; someone's called it that before. If you compare it to iPhones, you've got iOS, your operating system, and then people build apps on it. Apple doesn't build all the apps itself — anyone can build apps and sell them on the operating system through the App Store. It's a very similar model. Being on Foundry gives you a standard platform across the whole of the NHS — 220 organisations just counting the trusts, then there's the 25-odd ICBs and NHS England. The standardisation across all those different organisations provides the foundation to build on.
You're asking what it can do — well, if you're building apps for the NHS, what do you want it to do? Your imagination is the limit, really. You've got clinical processes that are very self-contained, and maybe, because it's on that standard platform, an app could pull outcome measures from all the other areas doing the same operation and show you your outcomes compared to everybody else's. So you're giving something back to the clinician, not just a place for them to record information.
Then there's what spans different areas — an ambulance trust versus a mental health trust versus a hospital trust running an A&E. There's a convergence point for those services: the ambulance is handing off to A&E, mental health patients are in A&E. But A&E can't see who's coming through the ambulance, because they haven't got access to the ambulance systems. They can't see that a patient coming in is also being seen by mental health services, with a whole mental health history to consider. If all of those systems are on the same platform, you could build applications that straddle those organisations — help the A&E team see the mental health history of the patient, alert the mental health team that their patient's just been to A&E or is on their way in the ambulance. The ability to respond to that information goes up.
It's not really about analytics — it does do analytics, and people have raised concerns about how it does analytics. If you compare it to Tableau, which looks beautiful, the tools within Foundry aren't as good. Yes, you've got things like R, you've got some native tools, but ask any analyst who's been working in Tableau and they'll say, "Can I just carry on using Tableau, please?"
So Tom, is it fair for me to think of the Federated Data Platform as a kind of operational operating system — almost — that allows all the many parts of the NHS to communicate effectively, but also to build things that then work across the whole system? Really, the sky's the limit in terms of what can be built — it's this foundational layer. Is that the right way to think about it?
Yeah, exactly. The sharing element of it is interesting — I think the Palantir term for the technology is "marketplace," which lets you bundle up an application that can then be reinstalled in any other Foundry instance. The NHS has rebadged that as the Solutions Exchange. That's not just the marketplace technology, it's also a kind of commercial framework where people can build things. Maybe that's a local team, maybe some clinicians working with their local data team — or maybe, with the AI coding support that's within Foundry, someone quite savvy has built their own technology in their own sandbox in a trust somewhere. Then they've been to a conference, told everyone about it, and everyone's like, "I want to try that."
So in future they could put that on the Solutions Exchange, and other trusts — because they're the data controllers — could choose to take those products or not. I expect, like we see with smartphones, there'll be competitors. There'll be a commercial incentive for different developers to make things better fit for the front line, and they'll get more data used out of it as a result.
One of the components of the FDP that really enables this is the canonical data model. You don't just need a standard platform, you need a standard data model underpinning it. Otherwise, when you build something in one area and try to transplant it elsewhere, the data definitions are different and you end up having to remap everything. We already see that cost in lots of places in the NHS — it's a hidden cost people don't factor in or really understand. When you see data projects taking a long time, it's often because people are having to figure out what each data item means and then remap it to wherever it's being moved.
A lot of the discourse has been quite political — about people's views of Palantir as a company, and everyone has their own views — but I didn't find that that interesting. What I wanted to find was more of an objective take on what this partnership unlocks, and whether Palantir are the right partner for the NHS.
And one thing I kept thinking about, and didn't quite get to a satisfying answer on, is: is there a good alternative? I understand the alternatives to Palantir in this world are generally big tech players like Oracle and IBM, then big consulting houses, and finally an in-house build by the NHS itself. But from your vantage point, is there actually a real alternative to Palantir, or are they uniquely positioned to do this?
“And when you look at the NHS, the data people are being paid on a clinician's pay scale. It's a fantastic, valuable profession, but nobody goes into nursing for the money.”
Tom
I've said a few times already that they are uniquely placed. If you work through your list — the competitors in the big tech space, you saw who applied for the bid that was put out in '23, and a lot of people say this company or that company could do it. I've genuinely looked into those, because I would love it if there was another alternative that could put some competition in this space and be a genuine competitor. But whenever I look at them, they're very much in that analytical space I described. They could potentially start moving towards the front-line applications that add the real value, but they're not really focused on that, and it would be hard for them to do. Whereas the architecture of Foundry is such that it's completely enmeshed — no one else has got close to that.
The PwCs, the Deloittes and the Accentures of this world provide staff and the capability to deploy things. They don't really build software. So they can help with it, but they'd never be able to produce a replacement.
And the in-house route, I think, is really unrealistic. Look at the NHS's track record: NHS Digital took a 30% cut in workforce merging into NHS England, and then NHS England merged into the Department of Health and Social Care with a further 50% cut. That applies to data roles as well as every other role. You'd need a very stable workforce to build something that could compete over a long period of time. Palantir have had 20 years in Silicon Valley and billions of pounds invested to get where they've got to now.
And when you look at the NHS, the data people are being paid on a clinician's pay scale. It's a fantastic, valuable profession, but nobody goes into nursing for the money. It's very easy for companies with money in the private sector to come and poach people doing a great job in the NHS. I've seen that quite a lot.
Is there a talent thing there as well? If you're able to pay three to five X what the NHS is willing to pay, you can just get a bunch of really talented people who'd be hard to get in the NHS in the current situation.
I think one area that's perhaps got more hope than the NHS building it in-house would be a government-sponsored company — somewhere in the public sector but not within the NHS — with the freedom to hire more broadly than the NHS does, with some backing and focus behind it, really focused on digital sovereignty and homegrown AI capability. A homegrown rival to Palantir. If that route was taken, it's probably got a lot more legs than doing it within the NHS. But even that would be a decade-long investment, and we'd have to be prepared to divest from American companies, given we've just made a massive investment in — or allowed them to invest in — our capabilities. So there'd have to be quite a bit of policy change to get to that.
Tom, I thought you made a really interesting point on the cost. The figure that's been reported — the amount committed or forecast — is £330 million from the NHS. It sounds like a big number, but I think you had some interesting thoughts on how it might not actually be that big a number. Even while I was practising, in a pretty large trust, we had a whole migration to Epic — I forget how long that took, but I remember it being almost a £200 million deal just to transition to an EHR, in one locality. This is a big national project — is it expensive, from what you understand?
It's all relative, isn't it? You've made the comparison with electronic patient records. I think it's a good deal compared to what's being spent now on data analytics. Go to any particular trust and you'll see they've got a data team and a data warehouse, they've either bought Microsoft licences or subscribed to a cloud provider, and they're spending huge amounts of money there. The FDP can be used as a data warehouse, and that's one of the opportunities for trusts — replacing that data warehouse with the FDP. The £330 million is centrally funded, whereas each NHS organisation is currently paying separately for its own data warehousing and analytics. So they'd make that saving, but also add a lot more value.
If you compare that investment case — say you assumed £100K per trust of existing spend, and it cost an extra £100K to get the FDP, even though the trusts don't actually pay that — look at the value you'd unlock: visibility in the analytics space over all your shadow IT, mitigating all the risks around confidentiality and loss of data that trusts are currently running, perhaps unaware. I think the true comparative figure would be the overall total cost of ownership of data management across the NHS right now. Nobody's quantified that. If you did, I'm sure it'd be a scary number.
We've spent about £2 billion since 2021 on electronic patient records and improving those — handing all this money to big American companies, for much less of a novel upgrade and much less value. So if you're talking value for money, there's a really strong case for everything the FDP can do that I've described.
Can you work through an example of a use case the FDP enables? And then I was wondering if you could help me understand the two or three unique capabilities Palantir bring to that. As an example, from my understanding, the FDP would enable something like: if in one trust someone had coded an AI tool that helps prioritise and get through waiting lists faster, another trust across the country could now also use that tool very easily — they'd have this shared language and interoperability and transferability. Could you help explain what it is that Palantir are doing with the NHS that enables that to work?
Yeah. I think the most unique feature of Palantir's software is their ontology. To me, that's really what bridges the gap between front-line applications and the analytics space. It's quite hard to understand, because as soon as I say "ontology," a lot of your listeners will think they know exactly what I'm talking about, and that's part of the confusion — the word is well established in computer science and data science to mean something specific about the semantics of a data system, but Palantir have co-opted it and applied it to some bespoke functionality that, as far as I know, only their platform has got.
What the ontology does is allow data to be treated as code. You package up the data with the semantics, the ability to live-link it to other pieces of data, and the ability to take actions on it. For example, you'd have a button on a screen that says "remove patient from waiting list," and within that ontology there'd be an edit to that line — that patient's status moves from waiting to no longer waiting. That doesn't happen in the analytics workspace. The orthodoxy is that you pass data between analytics and front-line applications; you don't bring them both together.
Just to get a bit technical for a minute: a front-line system that needs to do surgical edits — like a banking system, where you put in how much you want to pay, press the button, and that one record in the bank's database gets updated with that one payment — that's a very surgical edit. The systems that support that are optimised for it: very highly normalised, so the data structures are efficient enough to handle millions of such edits at any one time. That's how banks can have so many customers using their apps simultaneously.
In the analytics space, they're not trying to do surgical stuff — they're trying to look across all the bank accounts, across billions or trillions of records, and find trends. That's optimisation towards big data: pre-aggregation of figures, cubes, that whole snowflake approach to how the data models are set out. That's had its own evolution, gone through huge amounts of progress, but the two have never come together.
What the ontology does is invert the index and use what's known as Elasticsearch to do both at the same time. I won't get too much into the details of how the tech works, but suffice to say that within the same ontology objects, you can use it for analytics and for those surgical edits. That's what allows you to build an app on top of the ontology — the ontology is the backend for the apps, and you build the front end, whether in native technologies like Workshop or in open-source front-end web technologies like React. But it also serves as an analytical backend — people can query, say, a waiting list product across the whole country. And it performs in each of those two use cases. That's really the secret sauce of what makes Palantir — Foundry — unique to this problem.
God, that makes sense. Is there anything else from Palantir? I hear they have these forward-deployed engineers, a bunch of other things — Gotham, Foundry, et cetera — and I don't quite know what they are. Is there anything else outside of the ontology that makes Palantir uniquely suited to this kind of work?
“If you see a problem, it's yours, you own it, you fix it. If you see an opportunity, it's yours, you own it, you fix it. There's a lot of empowerment.”
Tom
I'm sure they'd give a much better answer than me, but some of the things that stand out: the forward-deployed engineers, I think that's quite a good system. They rotate their engineers into organisations like the NHS to work with people like the engineers I used to work with, then rotate them back to work on developing the platform itself, bringing that front-line knowledge back into development. I saw them making requested changes quite quickly — much more quickly than the EPR vendors I'd worked with in the past. That's a good working style.
And I became aware of the culture around it — if you see a problem, it's yours, you own it, you fix it. If you see an opportunity, it's yours, you own it, you fix it. There's a lot of empowerment. I went to an event once and was talking to somebody who'd only just joined the company, asking what it's like working there, and he described a really high level of autonomy he'd been given. Unfortunately, that doesn't really translate to the NHS — that's not something I've seen a lot of, where people are given the ownership and authority to just go and do something. There's a lot of approvals needed, a lot of disagreement that stops things happening and slows things down. So there's probably a bit of tension between how the NHS works and how Palantir works, and it sometimes creates friction.
The other thing to mention is AI. I don't think it's unique to Palantir, but it definitely helps: they've got the AI-powered forward-deployed engineer, or "AI FDE." That's what allows clinicians to just prompt into the system, "please build me an app that does whatever it is I want to do," and it uses Claude or one of the other large language models to build, code, link up, and deploy a solution.
I saw that working really well. Palantir came to me and said, "Tom, we've built this thing, can we try it out?" It looked at all the pipelines we were running — we were pumping huge amounts of data through the system, consuming compute, consuming storage, generating cost. They'd built an AI that would scan the logs, find what had used up a certain level of compute or storage, assess where the efficiency was, write the code to make the improvement, and present it as a pull request to the engineer, who could then review it and sign it off. The first time we did that, I asked one of my engineers to go and look at it and make use of it. He reviewed it, it was good, he approved it — and it saved tens of thousands, straight away, and that was just the first one. So there's so much opportunity here, not just for supporting the front line, but also for supporting the data processing and analytics side with AI tools.
So Tom, Palantir have this pretty unique capability, based on their ontology, and great teams they forward deploy, and it doesn't cost a lot of money either — pretty decent value for money, though that depends on your take. Are there any things where Palantir falls short against comparable companies? I'm thinking of the other companies that bid for the FDP project — Oracle, IBM would have been in that group, and other companies involved in collaborating, like IQVIA, some of the consulting houses. It's a bit of a broad question, but are there areas where they fall short, or comparable companies that outperform them?
Yeah, I think the big one is analytics. There are a lot of very specialist companies who've done amazing things in that space — whether it's basic things someone would do quickly in Excel, or those beautiful visualisations you see in Tableau. They're not out of the box in Foundry the way other companies do them. When NHS England was moving its analyst groups onto the FDP, a common complaint was, "It doesn't do what my other tool does."
Also — and I don't think this was necessarily a platform issue, more how it was implemented — the information governance around getting data in and back out again created friction. So there were areas like that where it wasn't as good as some of the other technologies in play at NHS England. That's probably the main area I'd spot.
One complaint I've heard about partnering with Palantir is that we're selling — or giving — Palantir NHS data. That's a common thing I see in the press. I think the answer is a bit more nuanced, but I'm curious for your take. I know we can get into the weeds of GDPR, and it might be useful to talk about the difference between a processor and a controller. Is that a valid concern — that in partnering with Palantir on such a big project, you're basically handing over NHS data to them?
Yeah, I think it's quite a serious point, because it's not just campaigners saying it — MPs have said it in the House of Commons. The quotes were — let me look this up — Polanski and Corbyn both said it. Polanski said it gives them access to analyse our medical health records. Corbyn said Palantir are getting their hands on personal health records and everything else, for research purposes. That's an egregious misstatement of what's actually happening. It doesn't give Palantir access to analyse medical health records; it doesn't allow Palantir to get their hands on personal health records for research purposes. That's a really specific claim and it's totally wrong. It's a shame people are talking about this, because it's baseless. The way the data is governed — the controllership, the legal framework — is GDPR controllership, and it sits with the NHS.
If Palantir were to do anything like that, they'd be breaking the law, and it would completely destroy their reputation as somebody anyone could work with. So it's just factually incorrect, and it creates fear that need not be there. It depresses and frustrates anybody actually trying to work on the FDP. And it distracts from the true debates — about how we set up data sovereignty, what the vision is for a UK version of Silicon Valley, if that's the direction we've got to go in. That's what the Westminster Hall debate should be talking about, not trying to tear down things that are actually working and have been rated green by the government's watchdog on projects — where I think only about 15% of programmes get rated green.
It's also had an impact on things like the patient opt-out. You've got campaigners like Good Law talking about the opt-out in the context of Palantir, and if you look at the online figures for the national data opt-out, since January those figures have gone up somewhere between five and ten times what they were in previous periods. That's an issue, because it doesn't opt you out of Palantir — you can't opt out of the FDP, that's there for direct care purposes. You can only opt out of things for research and planning. So it's ineffective to do, but people are doing it because the campaign is pointing them to do it. That then means future research projects will have biases and inaccuracies, because that data isn't there — people don't opt out uniformly across the population, certain demographics are more likely to opt out, and that creates bias within research. So it's quite harmful, what these campaigners and politicians are doing. It's not just depressing for the people working on it — it's actually got harms involved.
Tom, I understand contractually and through IG that Palantir aren't able to sell and experiment on your data in the way that's been described, but I'm curious, structurally — is there anything stopping them? I hear there's this Cloud Act where the US is able to compel US companies, even those operating in the UK, to hand over data. If Palantir were compelled at any point to hand over NHS data, is there anything structurally — encryption or other safeguards you'd have worked on — that would stop them from being able to do that?
To me, that's a bit of a rabbit hole. There are reasonable precautions in place against that, but with any software provided by another provider, you're never 100% able to say it's completely fail-safe. But that applies to all the other American companies — we've got Epic in place across the country, we've got Microsoft running all those SQL Server data warehouses, all subject to the Cloud Act. Any time I've looked at what the structural protections are, you end up in these endless circular debates. At the end of the day it's about trust, and whether you think American companies generally should be able to hold that data, given the American government could compel them to hand over a record or two under the Cloud Act.
I think it's a good point, because there are a bunch of other American companies — Epic, Microsoft, Salesforce — that work with NHS data, and I guess the criticism doesn't really come up in that context. Although I'm sure it does in your circles, which are much more astute in this world.
Actually, it hasn't come up.
Okay, so it doesn't come up. Then I guess my question — and I said it's not worth really getting into the politics — but perhaps there is a question of whether Palantir are uniquely problematic out of all these big tech companies. You see Peter Thiel come to the Oxford Union and say the NHS needs to be ripped up from the ground. You see Alex Karp publish this really controversial manifesto making a bunch of controversial statements. I don't want to ask you a political question, because it's a little unfair — that's not why I asked you on here. But is there a fair argument that Palantir are uniquely different, uniquely problematic, compared to companies like Microsoft and others?
Yeah — where people say that, I understand their point. They're talking about the origins of the company: that they were invested in specifically to build software for the military, the CIA, and things like that. Different companies have come from different origins, but you could also criticise the origins of some of those other companies. One that comes to mind is some of the EPRs that were developed in America, all around billing, not really applicable to the NHS — a different criticism, but still about origins. And some of the things some of the people say — Frank Hester made those comments about Diane Abbott, and that wasn't really treated as an issue as far as the contracts that organisation had.
I do take seriously the points people make, because the NHS has got very strong values around kindness, care, compassion — it's seen as a force for good, everybody recognises that, and it holds a completely different — almost opposite — position to a military organisation. So I understand why that is. But if you were to look practically, what could you do about it? I think this is where the debate gets it wrong. The mechanism through which we've ended up with companies like this is that lawmakers — MPs — have created a legal framework the NHS has to follow, and in any procurement they follow that very seriously and professionally. The law doesn't say anywhere that the politics of the organisation you're buying from have to meet certain criteria.
And think about what it would mean if it did — you could imagine the next life-saving drug that cures cancer, or whatever, can't be used by the NHS because so-and-so said X, Y, Z. Is it really wise to pursue that? But if it is wise to pursue it, why aren't we pursuing it? Because surely that's the answer — we shouldn't pick on results that have gone through a process when they're a natural result of that process. We should pick on the process and get the process right. That's where I'd prefer to see the attention — on the procurement process rather than the results of it. Though I'm not sure we'd ever get there. I'm sure this has been debated in the lawmaking already.
Yeah, it's a really tough problem — whatever your value system is, there's a natural conflict between wanting to uphold that and wanting access to best-in-class technology, with an objective, fair procurement process. I don't have a good answer there — I think it's up to people to make their own mind on it.
The last thing I wanted to talk about was the BMJ exposé about their work with Chelsea and Westminster. There were two criticisms. Number one was about how the data was reported on and the improvement shown — they made claims around waiting list times improving using the FDP, and the numbers were a bit cooked: they used a baseline from when things were really bad because of COVID, and then things started improving because COVID was veering off. So it was, "Look how great a job we're doing," when really things were improving everywhere. Number two was around conflict of interest — a bunch of people with pretty strong links to Palantir were also in leadership at Chelsea and Westminster. What did you make of that criticism?
“I wouldn't judge the iPhone based on a faulty calculator app. That's very specific, and I don't defend the benefit figures — maybe they're right, maybe they're wrong, I'm not defending them. But are we really saying the NHS can't be improved through the use of data?”
Tom
I thought it was quite specific criticism — I wouldn't judge the iPhone based on a faulty calculator app. That's very specific, and I don't defend the benefit figures — maybe they're right, maybe they're wrong, I'm not defending them. But are we really saying the NHS can't be improved through the use of data? Is the premise wrong? That would be my comment on that.
In terms of the conflict of interest, I never worked with Chelsea and Westminster — I've not really spoken about that publicly or formulated many thoughts on it — but the trust is responsible for its own governance arrangements. What I can speak to is whether 137 other trusts have deployed it and got benefit — got use — out of it. I think the bigger picture is important there. It would be inflating it to say that one small trust has somehow influenced the whole of the country.
What was the general opinion within your circles in the NHS — was the FDP seen as a great programme having a lot of success, with this one case being a bit of an outlier? What was the general consensus? I'm curious how people saw it from the inside.
Honestly, that particular Chelsea and Westminster example never really came up for me. If you're interested in what the engineers thought of Foundry — there was a lot of mystery around it for the engineers who came with me through NHS Digital; they weren't really sure what it was or what it did. When I met the people who'd been implementing it, and picked up line management of that team, there was a lot of enthusiasm. One guy told me, verbatim, "I'm an unapologetic fanboy." I thought, wow, that's pretty strong — there must be something to it.
When I worked with that team, the energy was really good — people felt like they were doing something worthwhile, felt like they were in the spotlight, working with the best technology. There was a real buzz in that team, and people really cared about getting it right. Then as others came on board who hadn't been part of that, there was a healthy dose of scepticism, and there were established ways of working being challenged by how Palantir encouraged the organisation to use Foundry. That did create some friction and tension.
What kind of ways of working?
A specific example: development of code on the platform is done via branching — you've got your master branch, you take a branch offline that takes the data and the code, work on it in a sandbox, get it right, then deploy it back to the master branch. That was really different to how the data engineers worked — they were more used to having a specific environment for development, a specific environment for test, then deploying to production, with nobody having access to production. It was just a different architecture for how people developed and worked, with merits on both sides — perhaps a bit of a culture clash.
Do you have five minutes for a rapid fire?
Yeah, let's do it.
I've got four questions for you. What's one thing about the FDP that most NHS England staff don't know?
Probably the hardest thing to gain consensus on in NHS England was what the ontology is. A lot of people just dismissed it as irrelevant and treated the FDP as another data warehouse tool. The potential I spoke about — to condense the data models and enhance the analytics — hasn't really been scratched at the national level. They use a lot of different data models to standardise data, when there could be one.
What's your single-sentence pitch for the FDP to a sceptical clinician?
For decades, the centre has foisted expensive systems on you that don't fit how you actually work. The FDP is a chance to build something that does.
Nice. And what's your single-sentence pitch for the FDP to a sceptical MP?
You've left the NHS to struggle with data for two decades with no effective strategy. Don't tear down the only strategy that actually works before you've got something to replace it with.
Last question — what's one piece of advice you'd give to a Palantir executive?
Their roots are in defence and intelligence, but their customers are nurses, paramedics, and clinicians. I'd encourage them to talk to them in their language, not your language.
What do you mean by that?
When you listen to those Palantir execs, they're talking very much to people in the military, in the security services — Western hard power and things like that. That's not the world of the NHS. So the invitation to Palantir is to think about what would resonate with the NHS, given they're a big customer and a big long-term investment for them.
Super interesting. Thank you so much, Tom, this was really good — I really appreciate it.
Yeah, I had a really good time, thank you. Thanks for having me.