Leading Hongkongers in Britain
Directing a grant-funded UK non-profit community organisation — scaling from a volunteer response to structural impact, with AI-assisted drafting on the casework side.
One of three directors since July 2021.
I joined the board of Hongkongers in Britain as a director in July 2021, when it was a volunteer response to a new wave of UK-bound migrants. It is now a non-profit community organisation (a company limited by guarantee) with staff across every UK region and policy engagement with three arms of central government.
What we do
Settlement and integration support for Hong Kong-background migrants. In practice: enquiries on settlement matters (immigration, housing, employment, daily life), English-language education, LGBTQ+ inclusion work, mental-health referrals, and advocacy with local authorities and parliament.
The numbers
Scaling the operation
A 40-person team of staff and volunteers, distributed across the UK, running several funder-restricted programmes at once.
- Programme-by-funding discipline. Each programme is a funder-bounded unit with its own deliverables, budget, and reporting rhythm. Shared overhead is allocated transparently. Multi-programme cases get warm-handed-off, not dual-charged.
- Advocacy as a distinct workflow. Policy and casework share evidence but have different outputs, audiences, and clocks. Treated as adjacent teams with a shared data layer.
The technical dual axis
Most of non-profit leadership at this scale is systems work.
- AI-assisted enquiry drafting. After government funding for the enquiry service ended, a volunteer team kept it running — with turnaround limited by volunteer capacity at around seven days. Retrieval-augmented generation over historical casework and authoritative sources now drafts first responses; volunteers moved from being caseworkers to being knowledge organisers (curating the reference corpus) and response reviewers (editing AI drafts before they go out). Turnaround: seven days → under two. Commodity models (GPT-5) behind a thin orchestration layer.
- Interactive AI acrostic-couplets service. A Lunar New Year 2025 experiment: personalised acrostic couplets in Chinese and English — a form notoriously hard to compose because each line must satisfy couplet rules (tonal balance, parallel structure) and a fixed first-character constraint simultaneously. 300+ users at community events.
- Reporting pipelines over Python and Power BI. The numeric sections of quarterly funder reports are generated automatically from the pipeline.
Policy impact
- UN Committee on Economic, Social and Cultural Rights (CESCR) — research submission reflected in the CESCR concluding observations on the United Kingdom at the 77th session (early 2025). Frontline casework data evidenced a pattern no single household could have surfaced alone.
- Ministry of Housing, Communities & Local Government — integration-funding recommendations adopted.
- Strategic Migration Partnerships — service-design engagement across English regions.
- MPs and local authorities — advocacy on integration support for Hong Kong-background immigrants in the UK, and on human-rights protection in Hong Kong.
Frontline casework → patterned evidence → policy briefs → recommendations → changes to what authorities do. That pipeline is HKB’s real product.
Notes from the role
- Funders trust applications that demonstrate measurement infrastructure, not just promised outcomes.
- Research submissions backed by internally consistent casework data carry weight that op-eds do not. Both are needed; the former is rarer in the sector.
- A non-profit at this scale is mostly systems work. Most of what moved the organisation forward was a data pipeline, a workflow change, or a small piece of automation — not a speech or a campaign.