Episode 155
AI in Recruitment: Transforming Public Sector Hiring with Andy Headworth
The UK tax authority is outpacing private sector peers on AI adoption. Andy Headworth shares how HMRC built job description automation, location intelligence, and CV sifting tools—each in under two weeks—and why workforce adoption, not technology, is the real bottleneck.
Episode Key Takeaways
SkillScribe, HMRC’s job description tool, generates a complete hiring package—job ad, social variants, behavioural frameworks, interview questions, and outreach copy—in 90 seconds using a bespoke LLM trained on public guidelines and career site data. Built by one person in two weeks using MindStudio (£20/month), it’s now in use by 168 hiring managers and proves that enterprise-grade AI solutions don’t require six-figure R&D budgets or vendor lock-in.
Location intelligence flipped from a recruitment-only tool into a cross-government asset. By layering ONS census data, LinkedIn insights, salary benchmarks, and local hiring signals into a single interface, the team created a product that Cabinet Office and logistics teams now use for workforce planning and regional strategy—an unintended consequence that demonstrates how AI tools built for one function often unlock value elsewhere.
CV sifting reduced an eight-person, full-day panel review (200+ candidates) to nine minutes. The tool deselects 80–85% of unsuitable applicants and leaves the hiring decision to humans, while also generating personalized rejection feedback at scale—a model that shifts AI from decision-maker to time-liberator and improves candidate experience simultaneously.
Adoption is the constraint, not capability. Andy emphasizes that workforce resistance—recruiters hesitant to use AI, hiring managers clinging to outdated templates, teams fearing role displacement—poses a bigger challenge than building the tools themselves. Getting teams to experiment early (lunch-and-learn sessions, low-stakes exploration) builds buy-in before rollout.
Public sector is moving faster than private sector on this. At a recent conference of 50 TA leaders, only five had deployed AI; the rest were still in planning or permission-seeking mode. HMRC’s willingness to iterate quickly, use off-the-shelf platforms, and share learnings openly has positioned it ahead of most Fortune 500 TA functions.
Frequently
Asked
Questions
How do you build a job description tool without custom software?
Use a no-code LLM platform like MindStudio (£20/month per user). Feed it your public guidelines, career site copy, success profiles, and brand tone. Layer in prompts that ask for job title and grade, then generate job ads, social variants, behavioural frameworks, and interview questions in one pass. One person can build and iterate the pilot in 2–4 weeks.
Can AI safely sift CVs without bias?
The approach here is deselection, not selection. AI removes 80–85% of unsuitable candidates based on CV-to-job-spec alignment, leaving 10–15% for human hiring managers to decide. The tool also generates personalized feedback for rejected candidates, improving transparency and candidate experience while keeping final hiring decisions human-owned.
What data do you need for location intelligence?
ONS census data, LinkedIn insights (scraped with platform permission), salary surveys, council information, and local hiring/firing trends. Combine these into a single interface where recruiters can query a location by name and get a one-pager on talent availability, salary benchmarks, competitor activity, and strategic recommendations for that region.
How do you get recruiters to actually use AI tools?
Start with low-stakes experimentation: weekly lunch-and-learn sessions where teams play with AI, create images, write copy, and explore use cases together. Build buy-in before rollout. Then chunk your recruitment process into blocks (sourcing, job specs, sifting, onboarding) and show how AI solves pain in each one. Early adopters create momentum.
What's the ROI on a CV sifting tool?
An eight-person panel spending a full day reviewing 200+ candidates now takes nine minutes. Multiply that across dozens of roles annually: the time and cost savings are substantial. Add faster rejection feedback and improved candidate experience, and the business case is clear—even before accounting for hiring manager time freed up for strategic work.