What Greater Manchester's Palantir Rejection Teaches Enterprise AI Buyers

The standoff

In early 2023, the UK government commissioned Palantir to build a federated data platform for the National Health Service, a programme worth more than $400 million and contracted to run until 2031. The promise was straightforward. Pull the NHS’s fragmented health records into a shared platform so care teams could plan treatment, manage waitlists, and allocate resources on something closer to complete information.

Most of the country moved toward adoption. One region did not. The integrated care board for Greater Manchester has repeatedly declined to adopt the platform, choosing instead to keep a locally built system it has developed over roughly a decade. Matt Hennessey, the board’s chief data and analytics officer, put the reasoning plainly in an interview with WIRED: even a technically strong platform will struggle to realize value if clinicians, data controllers, patients, and the public do not trust it.

That sentence is the whole story in miniature, and it belongs on the desk of every enterprise leader currently weighing a centralized AI or data platform purchase.

Why a single region said no

Greater Manchester’s objection is not that the technology fails. Palantir and NHS England argue the platform is already shortening hospital stays and wait times. The objection is about fit, trust, and control. The board believes its own platform is functionally comparable, more trusted locally, and not worth replacing.

This is a familiar shape for large organizations. A vendor arrives with a platform that works elsewhere, demonstrates impressive reference outcomes, and offers the appeal of standardization. The local team, which has spent years assembling its own tooling around its own workflows, looks at the migration cost, the loss of local autonomy, and whether the new system will actually be used, and decides the switch is not worth it.

The trust dimension is the part enterprise buyers underweight. A platform can be technically superior and still fail if the people expected to use it do not believe in it. Hennessey’s comment that full adoption would be a retrograde step is not nostalgia. It is a change-management verdict. Replacing a trusted internal system with an externally controlled one can reduce adoption rather than increase it.

The sovereignty question is no longer hypothetical

Greater Manchester is not alone in hesitating. The WIRED reporting notes that several European governments are reevaluating their relationships with Palantir, partly to reduce dependence on US-built technology. For public-sector buyers, and for any regulated enterprise, vendor origin and data residency are now strategic questions rather than procurement footnotes.

The practical lesson for enterprises is to treat platform choice as a governance decision, not merely a tooling decision. Who operates the system? Where does the data rest? What happens if the relationship sours, or if the vendor’s direction diverges from your own? Greater Manchester’s answer was to retain a capability it already controlled. That is a defensible position even when the alternative is, on paper, more advanced.

Centralization versus federation

The NHS federated data platform is, as the name suggests, a hybrid. A national data pool sits underneath, while local organizations keep their own databases and build their own tools on top. One of the platform’s advocates, Tom Bartlett, a former NHS England deputy director of data engineering, described the appeal as lift and shift. Build a tool in one region and deploy it elsewhere. The other draw, he noted, is a consistent surface for artificial intelligence to work across.

That is the central tension enterprise architects live with. Standardization creates real advantage at scale, but it assumes local needs are similar enough that one model fits. Federation preserves local autonomy, at the cost of duplicated effort and slower propagation of good ideas. There is no universally correct answer. There is only the answer that matches your risk profile, your regulatory exposure, and the strength of the local systems you already run.

What enterprise buyers should take from this

A few points survive the specifics of the NHS case.

First, trust is a deployment dependency. Before committing to a platform, map who needs to believe in it and whether they currently do. A mandate from the top does not substitute for clinician, operator, or end-user confidence.

Second, sovereignty is a design input. For regulated industries, the question of who controls the data and the roadmap belongs in the evaluation criteria from the start, not in a risk register written after the contract is signed.

Third, incumbent systems earn their keep. The system a team built in house over ten years carries context that a new vendor platform does not. A replacement has to clear that bar, not just match a feature list.

Fourth, measure the claim, not the demo. Vendors will show efficiency gains. Ask for the measurement method, the baseline, and the counterfactual. Greater Manchester’s position is essentially that the measured gains were not compelling enough to justify the switch. That is a legitimate finding, not obstruction.

The decision still ahead

The UK government has a window, reported as early next year, to terminate the NHS contract before it runs its full term. Greater Manchester’s holdout has become part of a wider debate about whether a single vendor should sit at the center of a nation’s health data. Whatever the outcome, the region has already demonstrated the most important move an AI buyer can make. It decided what it needed, assessed the offer against its own capabilities, and reserved the right to say no.

For enterprises watching the same category of decision, that discipline is the takeaway. Platforms are purchased, but judgment is not. Book a strategy consultation if you want a structured pass at your own AI and data-platform roadmap.