What I wish MPs had heard before Thursday's FDP debate

On Thursday, MPs debated the NHS Federated Data Platform in Westminster Hall. I listened to the full 80 minutes. I could hear the passion in the room. These are people who care about the NHS, about data privacy, and about getting value for public money. I share all of those concerns.

But listening as someone who spent three and a half years leading the engineering team that built FDP, I was struck by how much of the conversation was shaped by incomplete information. Not because anyone was acting in bad faith, but because the debate was shaped almost entirely by campaign briefings and media coverage rather than operational experience of the platform. That is a gap I would like to help fill.

I set out my position earlier this week in HSJ, where I explained why triggering the break clause would generate a monumental and unnecessary task and why there is no credible alternative. I want to be clear about that position. I have no commercial relationship with Palantir. I have never been paid or encouraged by them or the NHS to say any of this. I support the principle of data sovereignty and would welcome a credible UK-built alternative. But the reality is that in 20 years, no one has come forward to even propose something like FDP. The debate on Thursday did not point to any credible plan to replace it. It did not touch on why no UK firm has built anything comparable, or what it would take to get there. I would like to share some of what I know, in the hope that it helps the conversation move forward.

What the NHS actually owns

A concern raised in the debate was that the FDP contract delivers "no software, not one line" and that all intellectual property belongs to Palantir. I understand why this reading of the contract has gained traction. The contractual language around "specially written software" is genuinely confusing. But the practical reality is different.

I led a team of 150 engineers who wrote code on this platform every day. That code was written in standard open source languages: SQL, Python, PySpark, and React, and stored in Amazon Web Services. The code belongs to the NHS. The Canonical Data Model belongs to the NHS. The linked data belongs to the NHS. When an NHS engineer writes a data pipeline or builds a clinical dashboard on FDP, that work is NHS property and can be moved to any other platform that supports those standard languages, which is virtually all of them.

What Palantir retains is intellectual property in Foundry, their own platform software. That is no different from Microsoft retaining IP in Excel while users own the spreadsheets they create.

The Minister, Dr Ahmed, confirmed this during the debate: "The NHS owns the intellectual property for all products and it is possible to migrate them to other providers." That is consistent with my direct experience.

The cost of FDP was described as "close to £500m". What I think would help the debate is visibility of the unquantified cost the NHS already bears maintaining hundreds of nonstandard systems: electronic patient records (EPR), business intelligence tools, data warehouses, and the staff to keep them all running and integrate the data between them manually. Developing FDP reduces that cost. Delaying FDP increases it. If MPs could commission that comparison, it would bring clarity to the value-for-money question that the debate rightly raised.

How the platform actually works

A theme in the debate was that FDP is a monolithic system, a single point of failure controlled by one supplier. I can see why it might look that way from the outside. In practice, it is the opposite.

The platform is federated, not monolithic. Each Trust has its own instance. The application layer is made up of many separate products built with a plethora of open source tools. The Canonical Data Model provides the interoperability standard that allows data to flow between instances. The data is stored on Amazon Web Services, but if needed it could move to Microsoft Azure, Google Cloud, or something else entirely.

One MP proposed a "modular system, a bit like Lego" as an alternative. That is a fair aspiration, and it is a reasonable description of how FDP already works. The building blocks are there. What is needed now is for more Trusts to start using them.

The Privacy Enhancing Technology layer adds another dimension. When sharing patient data between FDP instances, for example between two neighbouring NHS Trusts, the PET system acts as a membrane between them, ensuring only the right patient data is shared safely. The contract for PET was deliberately structured so that the FDP software supplier could not win it, to strengthen the integrity of the system. IQVIA provides the PET layer, not Palantir. That separation was built into the procurement design, not bolted on afterwards.

On the question of bespoke code within each Trust: the mapping of data schemas from local EPRs like Epic or RiO to the Canonical Data Model is NHS property. The Foundry connector remains Palantir's proprietary software, as it should.

The lock-in question deserves a fuller answer

I understand the concern about vendor lock-in. It is a legitimate question for any major technology contract. Here is why I believe the risk is lower than the debate suggested.

NHS England previously ran an analytics platform provided by a company called SAS which requires its users to write their code in a genuinely proprietary coding language. Users of that system are required to learn and build in that language, creating a strong vendor lock in. Migrating from SAS to FDP National was painful precisely because of SAS's proprietary code language. It was not possible to lift and shift the code. Engineers had to rewrite it from scratch in the open source languages available in Palantir's platform. Engineers who may one day face the task of migrating code from Palantir to another supplier face nothing like that because the languages used by Palantir are open source.

There is also a point that I think could transform the lock-in debate entirely. The migration pattern for a standardised system like FDP is infinitely simpler and therefore cheaper than the nonstandard systems the NHS traditionally deploys. Think of all the migrations that have happened across nonstandard systems over the past 20 years, and into the future as more nonstandard systems are migrated to more desirable systems. Consider the number of EPR migrations occurring right now. The data warehouses being moved to cloud providers. Each one a separate project, generating migration work independently, the total cost to the taxpayer remaining completely invisible. FDP, by standardising the estate, actually makes future migration easier, not harder. I would encourage MPs to explore this point further.

What the platform is delivering

The Minister provided figures that I think deserve wider attention: over 100,000 additional patients supported to undergo procedures in theatres. Nearly 94,000 people supported on their cancer journey, with a 7% reduction in cancer diagnosis times. A 14% decrease in delays to discharging patients staying more than seven days. He stated that 137 NHS Trusts are actively utilising the platform and reporting benefits, and that the programme has exceeded every single target since its go-live date in March 2024. He confirmed that external independent experts have validated these results, with a further review planned.

The Minister also gave specific clinical examples. North Tees and Hartlepool NHS Foundation Trust used the OPTICA discharge product to reduce long stays by a third, despite a 7% increase in admissions. Mersey and West Lancashire Teaching Hospitals used FDP to manage theatre lists, enabling surgeons to operate on more patients per day.

A BMJ article published this week criticised the benefits analysis of one specific FDP product at Chelsea and Westminster Hospital. That is legitimate scrutiny and I welcome it. But whoever is right on that issue, one product at one Trust does not invalidate the programme-wide benefits being reported by 137 organisations. Nor does it invalidate the potential of future products being developed now or those that will be innovated in due course. Scrutinising individual products is how platforms improve. It is not evidence that the whole programme has failed.

The Artificial Intelligence potential that was missed

A claim was made in the debate that FDP "uses external AI platforms from OpenAI and Anthropic and brings questionable value itself." I would like to offer a different perspective, because this is where the platform's future potential is most exciting.

FDP has embedded AI large language models within the platform. These are contained within the secure environment and are not linked back to OpenAI or Anthropic's servers. This provides two capabilities that could transform how Trusts work. First, it allows staff with little technical expertise to develop applications for their clinical environment, saving stretched technical resources and fuelling frontline innovation. Second, these models can enhance FDP products directly, for example by providing AI-powered analytical capability that saves Trusts thousands in data analyst time.

It is the combination of these AI models with a standardised data model and NHS-wide data that makes them so powerful. The AI could run any analysis it is asked to. The advances this brings for researchers and for service management are significant. I would genuinely welcome the opportunity to demonstrate this to any MP who would like to see it.

Being honest about pace

The pace of FDP delivery has been a concern for some time. In June 2024, Wes Streeting, then Shadow Health Secretary, described adoption as "glacial" and told Trusts to "go further, faster." In Thursday's debate, Martin Wrigley claimed that after nearly three years, only three or four of 13 core capabilities had been partially delivered. The Minister countered that the programme has exceeded every single target since its go-live date in March 2024.

I want to be honest about where I stand on this. That criticism has some basis. The programme was two years old in March 2026 and has not delivered FDP to the entire NHS. Part of the reason is that successive Governments decided to subject the service to two mergers, firstly NHS Digital with NHS England, then abolish NHS England mid-flow, which has consumed management attention and created uncertainty. Hundreds of vacancies have been frozen, and roles deleted through redundancy schemes and restructures. Part of the reason is that Trust adoption takes time, and the information governance processes required for Trusts to sign basic memoranda of understanding have been painfully slow.

Could it have been pushed harder? Yes. Mandating adoption with surge resources directly into Trusts would have accelerated things. Streamlining the IG processes would have helped and still could. Encouraging development of more FDP products sooner would have shown the potential more clearly and could have created more local buy-in. I would encourage MPs to push the Government on these points. They are solvable problems. How might MPs facilitate a more effective achievement of the vision of a joined up NHS data estate, which many MPs shared, despite their misgivings about Palantir?

Building the right comparison

During the debate, OpenSAFELY was cited as evidence that the UK can "do better." OpenSAFELY is a legitimate and impressive platform, and the team behind it deserve credit. But it is a research analytics tool built on GP data. It does not manage waiting lists, plan discharges, schedule theatres, or track cancer pathways. No one is attempting to roll out OpenSAFELY to 240 organisations. The comparison, while well intentioned, risks creating a false sense that an alternative is already available. It is not.

It was also claimed that "Palantir is in danger of getting its hands on the personal health records of every single person who has lived or died since 1948." This is not accurate. Palantir provides the software. The NHS controls the data.

On sovereignty, I share the aspiration. But the Minister made an important point that I think deserves reflection. He described his own practice in Glasgow: logging into Microsoft Windows, opening a North American system to order tests, another to read results, another to view X-rays. Four American software products before touching FDP. The sovereignty question is not unique to Palantir. It applies across the entire digital estate. Addressing it seriously would mean a comprehensive strategy for UK technology capability, not just a decision on one contract.

Could it be done in the UK?

I have thought about whether I would support a properly funded five-to-ten year plan to build UK sovereign capability in health data infrastructure. That plan would need a great deal of money, sustained commitment across electoral cycles, and protection from being cut by future governments in the way that NHS Digital was absorbed and hollowed out. I am not convinced that the public sector could deliver this alone, and the private sector has not yet stepped forward with a credible plan. There are small niche providers working close to the space occupied by Palantir, but none on the scale the NHS needs.

Looking in house, NHS England has lost 30% of its technical workforce through the NHS Digital merger and is about to lose another 50% through abolition and merger into DHSC. ICBs and local Trust HQs have been cut in half as well. The procurement capacity to run a new national contract of this scale is actively being removed right now as CSUs close down. The technical specification team does not exist. Triggering the break clause in these conditions can only produce a gap.

If MPs and campaigners are serious about a UK-built platform, I am genuinely open to that conversation and would welcome the chance to contribute. Show me the consortium, the delivery plan, the timeline, and the funding. That conversation would be worth having. It just has not happened yet.

What I would love to see next

The debate showed real passion for the NHS and real concern about data privacy and sovereignty. Those instincts are right. What would make them more effective is direct engagement with the people who build and use these systems every day.

Before joining NHS England, I worked at an NHS Trust in London. We audited over 1,000 shadow IT sources: spreadsheets, whiteboards, printed lists updated with biros. When I called for providers who could help, nobody came forward. FDP was the first platform that could address this at scale. That practical reality was not part of Thursday's debate, and I think it should be.

Some contributions on Thursday pointed in a productive direction. Jim Shannon made a reasonable point about patient empowerment through data. Samantha Niblett acknowledged that the technology is "genuinely impressive" while raising legitimate questions about trust and transparency. Julian Smith's suggestion of a more robust governance structure, perhaps involving the National Data Guardian, was constructive. These are the kinds of contributions that move the conversation forward and would benefit from a lot more focus.

I would also like to address the comments made about Ming Tang, the outgoing Chief Data and Analytics Officer at NHS England. She did not "introduce" Palantir; this is something credited to members of the Government at the time. She led the implementation and then oversaw a procurement that followed the regulations. Public servants who deliver complex programmes under intense political pressure deserve fair treatment, not personal attacks under parliamentary privilege. Getting that right matters, because the NHS will need talented people to step into difficult leadership roles in the future, and they need to know that doing the job well will not make them a target.

FDP is the first platform that connects NHS systems at national scale. It is not perfect. But it is real, it is working, and it has potential that most people have not yet seen. I would welcome the chance to show any MP, campaigner, or Trust leader what that potential looks like in practice. My door is open.

Tom Bartlett

Founder, Bartlett Data Ltd

Former Deputy Director of Data Engineering, NHS England

bartlettdata.co.uk