The UK government has awarded KPMG and EY civil-service learning contracts with a combined maximum value of £456 million, creating a substantial outsourced training bridge while Whitehall builds a new in-house National School of Government and Public Services. The arrangements run into March 2028 and include training in artificial intelligence alongside broader professional and organisational skills.

KPMG’s portion carries a ceiling of £319 million, while EY’s is worth up to £137 million, according to procurement data reported by the Financial Times. Those amounts represent the maximum available under the contracts rather than guaranteed expenditure, since departments will draw down training according to demand. Even on that basis, the size of the framework illustrates the scale of the capability problem government is trying to address as digital systems and AI change the skills expected across a workforce numbering hundreds of thousands.

The awards arrive during an awkward transition in Whitehall’s training model. In January, government announced plans for the National School of Government and Public Services, saying it would align civil-service learning more closely with priorities including AI while bringing more provision in-house. The stated aim is to reduce duplication and dependence on external providers as the school develops, although rebuilding internal capacity takes time and departments continue to need training while that structure is established.

KPMG and EY will therefore operate in the gap between the existing outsourced model and the future system government says it wants to create. The contracts replace expiring arrangements rather than establishing a permanent commitment to the same delivery model, with the Cabinet Office describing the spending as training services rather than conventional consultancy. That distinction carries some weight because ministers have separately committed to reducing external consultancy expenditure, while large training contracts often use prime suppliers to coordinate networks of specialist learning providers rather than having their own consultants deliver every course.

AI training becomes an organisational problem

The inclusion of artificial-intelligence skills reflects how quickly AI has moved from specialist technology teams into ordinary public-sector work. Departments are already experimenting with copilots, automation, analytics, and generative systems, creating demand for training far beyond software engineering. Staff need to understand when tools are useful, how information can be handled safely, what decisions require human oversight, and where automated outputs create legal, security, or operational risk.

Those requirements make AI capability difficult to deliver through a single generic course. A policy official using a language model to analyse documents has different needs from a procurement specialist evaluating an AI supplier, a data scientist building a model, or a manager deciding whether automation should change a service process. Whitehall consequently needs a mixture of broad literacy and deeper professional expertise, while the underlying technology continues changing faster than many conventional training curricula.

The government’s proposed National School is intended to create more coherence around that problem, with a streamlined curriculum and greater internal delivery. Government Skills says the school will align training with priorities such as AI, reduce duplicated provision, and lower external spending over time. Recruitment under way this summer also shows the institution building internal evidence and evaluation capability, suggesting an attempt to strengthen how government decides which forms of training produce useful results.

Outsourcing skills while rebuilding them inside

Spending heavily through outside providers while trying to strengthen internal capability creates an unavoidable tension. Outsourcing can provide immediate scale, access to specialist instructors, and administrative capacity, particularly when demand fluctuates across departments. Yet long-running dependence on intermediaries can make government less capable of designing, delivering, and evaluating training itself, leaving the organisation buying expertise repeatedly rather than retaining it.

The contract structure will therefore be judged partly on how much capability remains after it ends. Training thousands of people in AI terminology would be easy to count but would say little about whether departments can deploy technology more safely, redesign processes, improve productivity, or challenge suppliers more effectively. A stronger measure would be whether teams develop enough practical expertise to make better technology decisions without continually turning to external advisers for interpretation.

That distinction is particularly relevant because AI procurement is likely to expand across government while the technology remains difficult to evaluate. Officials commissioning systems need to understand data access, model performance, security, supplier lock-in, intellectual property, auditability, and the limits of automated decision-making. Better internal capability could reduce dependence on consultancy elsewhere in government spending, whereas poorly targeted training could simply add another external cost alongside the technology contracts themselves.

The National School is expected to take several years to reach its intended capacity, so KPMG and EY will remain significant parts of the civil-service learning system during the transition. The £456 million ceiling does not mean that amount will necessarily be spent, but it establishes a large financial envelope around an unusually consequential workforce challenge. Whitehall is attempting to acquire AI and digital capability quickly while rebuilding the institution intended to provide more of that capability internally; whether the bridge shortens dependence on outside suppliers or merely extends it will depend on what government retains when the contracts finish.