← All speakers

Eoin Mulgrew is Head of Digital Transformation at 10 Downing Street and leads the No10 Innovation Fellowship, which recruits experienced engineers and data scientists to tackle problems across British government. His ambition is to give public institutions the technical expertise to build useful systems themselves, from legal analysis and policy modeling to safer deployment of AI in schools.

Earlier in his career, Mulgrew worked across startups, industry, government infrastructure and Brexit preparations. After the 2019 general election, he joined the Cabinet Office to lead digital and data elements of government reform. He spent three years tackling departmental data silos and technical recruitment before joining Number 10 Data Science in 2023.

He also helped develop Evidence House, where civil servants build practical coding, data science and AI skills through training, boot camps and hackathons. By June 2025, the initiative had delivered approximately 24,500 training hours to more than 1,000 colleagues. The fellowship complements that internal training by recruiting specialists from outside government, offering greater operational autonomy and assignments embedded with public-service teams.

Practical ideas and contributions

  • Forward-deployed public-sector engineering: Mulgrew places engineers alongside policy advisers, lawyers and operational staff to understand their workflows and build usable tools quickly. Projects have included analyzing the UK statute book, modeling how policy changes could affect household finances, and delivery red-teaming to identify optimism bias and weak mitigations in government project reports.
  • Accountable AI in public services: His approach includes testing models for sycophancy, training users to recognize their limitations, and evaluating educational AI against safety and cognitive-load benchmarks. Fellowship participants also supported the UK AI Security Institute; fellow Harry Coppock led work associated with its Inspect evaluation framework.
  • Cross-government delivery: Fellows contributed to Extract, the government’s planning-document digitization tool, developed by the Digital Planning Programme and Incubator for AI with Google technology. It became available to every local planning authority in England in June 2026. Mulgrew also works with teams including Justice AI, whose founder, Dan James, previously participated in the fellowship.

His next challenge is extending these methods beyond specialist teams into reusable services such as transcription and call-center support, while preserving public accountability.

Read the topics behind these talks

1 conference talk

Key ideas

Scroll to read ↓

Inside Number 10’s experiment with technical fellowships, embedded engineers and reusable AI tools—and the institutional changes needed to make their work scale.

  • What can a small team do about public-service backlogs?
    0:51 ↗
  • The constraints are organizational as well as technical
    3:23 ↗
  • An exceptional team inside the system
    5:23 ↗
  • Start with the workflow, then choose the deployment
    8:37 ↗
  • Model policy choices and make legal analysis reusable
    9:59 ↗
  • Challenge delivery reports and expose progress
    12:18 ↗
  • Build capability through partner teams
    14:32 ↗
  • Evaluate tutoring before broad adoption
    17:18 ↗
  • Embedded engineers on the frontline
    18:11 ↗
  • What if the policy tool tells users what they want to hear?
    21:25 ↗
  • Make the exception normal, then work across services
    22:41 ↗
  • The remaining gaps: motivation and shared experience
    25:44 ↗

References