AI Giving Agent
Turning community-verified women's safety data into donor action. A two-stage AI pipeline that transforms incident reports into funded micro-projects — transparently, at scale.
The Challenge
Three problems. One system.
The Data Gap
Women's safety threats are hyper-local and invisible to donors. SafeCity's 130,000+ incident reports have never been translated into actionable giving opportunities.
The Trust Gap
No GiveWell equivalent exists for women's safety philanthropy. Without credible evaluation infrastructure, giving is driven by brand recognition rather than evidence.
The Friction Gap
Most motivated donors leave giving platforms without acting. Vague cause descriptions and no personalisation prevent conversion.
The Solution
The Two-Stage Pipeline
Stage One
NLP Community Intelligence Engine
The supply side. Ingests multilingual incident data from SafeCity and generates structured, community-validated, fundable micro-project briefs.
- Ingests multilingual SafeCity incident data
- Clusters incidents by location, type, and severity
- Generates structured, fundable micro-project briefs
- Safety Champions validate each brief for accuracy
- Vector embeddings stored in Supabase + pgvector
Stage Two
AI Giving Agent
The demand side. Conducts a values conversation with donors, matches them to projects via semantic search, and explains every recommendation in plain language.
- Conversational donor values elicitation
- Semantic matching to relevant micro-projects
- Plain-language explainability card for every match
- One-click giving via Every.org integration
- Real-time feedback loop to improve future matches
Get Involved
Partner with us on SafeCity
Whether you're a funder, NGO partner, or organisation looking to replicate this model — we'd love to hear from you.