UK computer science graduates are branching out as coding's entry lane narrows

The entry route from a computer science degree into a coding job became markedly narrower in the latest UK graduate data. The Guardian reports that the share of computer science graduates finding professional work as coders or programmers fell from about 40% to 28% in the latest year. The share entering any graduate-level occupation fell from more than 60% two years earlier to 50%.

The underlying HESA Graduate Outcomes survey covered more than 350,000 former students 15 months after completing courses in 2024. That is a substantial observation base, but the figures presented in the report are descriptive outcomes, not an experiment. They show where graduates landed; they do not identify why demand, hiring or occupational classification changed.

The causal claim remains open

The sharpest disagreement is about AI. Matt Hiely-Rayner, whose consultancy compiles the Guardian University Guide, argued that cheap AI-produced routine work is difficult to separate from the decline. Charlie Ball, Jisc's head of labour-market intelligence, was more cautious: he saw a clear change in the software-developer market but said there was little evidence that AI caused it. That caution belongs in any workforce decision based on the numbers.

Other mechanisms can operate at the same time. Technology hiring expanded rapidly and then corrected; employers can move roles between occupational categories; graduates may take longer than 15 months to enter professional work; and a cohort completing in 2024 faced a particular macroeconomic market. The published account does not provide vacancy counts, applicant volumes, pay, employer size or comparisons adjusted for these factors.

The data also contains a diversification signal. Ball said computer science graduates were moving into related areas such as cybersecurity and network engineering. That makes the change more than a story about fewer programmers. It may reflect wider technical work absorbing graduates, even while the most legible junior coding title becomes scarcer.

Redesign the bridge into work

For employers, the immediate risk is not simply a talent shortage or surplus. It is the loss of the tasks through which novice engineers build judgment. If assistants produce scaffolding, tests or routine transformations, teams need deliberate ways for new hires to inspect failures, explain trade-offs and own bounded production changes. Senior review cannot substitute for a developmental task architecture.

Universities should test whether curricula expose students to adjacent destinations and to the evidence practices those jobs require. Cybersecurity, network operations, data governance and AI-assisted software delivery share foundations but differ in accountability. A generic instruction to "learn AI" is weaker than assessed practice in evaluation, debugging, secure deployment and communication. The Skills Atlas can help describe those capabilities, but providers still need local evidence about proficiency.

The best next measurement would link subject, task content, vacancy demand and progression over several cohorts. It should also distinguish permanent positions, contract work and further study. Until then, the 28% figure is a serious signal for pathway design, not proof that AI eliminated a fixed share of graduate jobs. Leaders should respond by instrumenting entry routes and destination roles, while keeping the causal diagnosis open.