The engagement
When a major incident happens, the charity's trustees need a clear recommendation fast: activate a national appeal, support a local response, or keep monitoring. Assembling the evidence for that decision took hours of manual research under intense time pressure, with no consistent structure from one emergency to the next. We designed and configured a governed decision-support agent around an existing AI model. It scores each incident against a six-pillar evidence framework and draws on a structured library of more than 130 documented historical emergencies alongside live official sources. Every output separates confirmed from unconfirmed information, logs its sources and timestamps, and ends in a recommendation, never a decision.
What was delivered
- Six-pillar evidence framework covering impact, urgency, resource strain, public attention, aid gaps and mobilisation potential
- Structured library of more than 130 documented historical emergencies
- Board-ready assessments reported by the team in under 15 minutes, previously several hours
- Supported monitoring and triage across around a dozen major incidents since 2025
- In one recent case, a full assessment was ready within 100 minutes of a major incident
- Confirmed and unconfirmed information separated, with a source log on every output