Whitehall’s recruitment process does not reliably establish whether candidates can do technical jobs, former government AI head Laura Gilbert told MPs. Her proposed practical alternative is a four-hour coding test based on a realistic scenario—but it is one example, not a complete hiring model. The issue matters because government’s plans to improve public services and productivity depend on building and retaining capable digital teams.
What did Laura Gilbert say about civil-service tech hiring?
Giving evidence to the House of Commons Science, Innovation and Technology Committee, Gilbert said CV keyword checks and behaviour-based interviews can be poor indicators of whether someone can perform technical work. “It’s very difficult to hire technologists well,” she said. “The way the civil service hires for this sort of role is not suitable for this sort of purpose. It’s really hit and miss.” Computer Weekly reported her evidence on 25 March 2025.
The concern is about job validity: a CV may include the right programming languages and terminology without demonstrating that its author can use them to solve the problems a role requires. Gilbert said the civil service was “still doing behaviour-based interviews and trying to apply that to technical roles”. She described a gap between signals that help candidates present themselves well and evidence that they can do the work.
Gilbert said her AI incubator uses a four-hour coding test based on a real scenario. That gives applicants an opportunity to demonstrate practical ability, rather than relying only on credentials or interview answers. Her testimony does not establish that this test has been evaluated against other methods, nor does it show a measured improvement in hiring outcomes.
Recommended Free Tools
#1 Best Overall
How do the approaches compare?
| Hiring question | CVs and behaviour-based interviews | Gilbert’s practical test example |
|---|---|---|
| Evidence of technical ability | Credentials, keywords and interview responses; Gilbert said these can be hit and miss for technical roles. | Applicants demonstrate coding on a realistic scenario during a four-hour test, as reported by Computer Weekly on 25 March 2025. |
| Candidate time and effort | The cited account does not state a standard number of interview stages or their duration. | Four hours for the coding test; Computer Weekly’s 25 March 2025 report does not state whether this is paid or what other stages are involved. |
| Consistency between departments | Not established in the cited account. | Not established; Gilbert described her AI-incubator process, not a government-wide standard. |
| Roles beyond coding | Behavioural interviews may address some non-coding competencies, but the account does not evaluate their effectiveness. | A coding test alone cannot assess every leadership, data, security or non-coding capability. |
A practical assessment can make technical ability more observable, but it also asks candidates to commit substantial time and may test only a slice of the job. A realistic, role-specific exercise is different from a generic puzzle: the former can reveal how someone approaches relevant work, while neither format by itself establishes whether a person can lead a team, communicate well, handle sensitive data or deliver within government constraints.
Why is technical hiring part of a wider government reform?
In January 2025, the government published a blueprint for a refreshed Government Digital Service (GDS), intended to consolidate digital, data, AI and related capability across government. The plan is set against a large existing workforce and budget: the government’s State of digital government review reports more than £26 billion in annual public-sector digital-technology spending and nearly 100,000 digital and data professionals.
The same review estimates that full digitisation could unlock £45 billion a year in unrealised savings and productivity benefits, equivalent to 4–7% of public-sector spending. This is a government estimate of potential benefits, not a guaranteed saving, a budget reduction already achieved or a forecast that can be attributed to hiring changes alone.
Recruitment is one link in the delivery chain. The government also has to address unmaintained legacy technology, make data exchange clearer, provide effective leadership and coordinate implementation across departments. Richard Pope, a former government digital adviser, told the committee: “There’s a lot of unmaintained technology the government relies on to do its job. It’s not good enough at the moment, and we obviously need to fix that.”
What a stronger technical hiring process would need to establish
Gilbert’s testimony makes the case for better evidence of technical skill; it does not set out a full replacement for civil-service recruitment. A credible system would need to match assessment to the role and combine practical evidence with other relevant checks.
- Test the work the role actually requires. A coding exercise may suit a software engineering post, but data, AI research, architecture and technical leadership roles need assessments that reflect their distinct responsibilities.
- Use consistent standards without making every role identical. A refreshed GDS could help departments share expectations and good practice, but the cited evidence does not specify a government-wide assessment standard.
- Keep the candidate burden proportionate. A four-hour exercise is a substantial commitment. Employers should be clear about its purpose and how it fits into the rest of selection; the reported account does not say whether Gilbert’s test is paid.
- Assess the conditions that affect delivery. Technical skill cannot by itself resolve retention, pay, security-clearance delays, legacy-system constraints or cross-department dependencies. These factors shape whether a team can recruit and deliver successfully.
The reported evidence does not quantify how many government technologists are unsuitable, and it does not show that coding tests alone would solve recruitment. Gilbert’s warning is narrower and consequential: when hiring methods do not reliably show who can perform a technical job, expanding digital teams or labelling more roles “digital” will not guarantee the capability needed to improve services.
Quick Recap
Rank #4
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
