SafeMargin AI Cyber Insurance
Multi-sided cyber insurance platform connecting insurers, brokers, and buyers with AI-driven insights, portfolio-aware recommendations, and policy lifecycle management.
01 — The problem
What we were solving
Brokers needed faster quotes; insurers needed consistent risk signals; buyers needed transparent coverage comparisons.
- Questionnaires were long and duplicated across carriers.
- Risk scoring models were opaque to brokers trying to explain premiums.
- Policy changes generated PDFs nobody could query later.
02 — The approach
What we built
- Introduced a canonical risk graph fed by telemetry questionnaires and third-party signals.
- Dual-agent pattern: one agent summarises exposure, another drafts broker-ready explanations with citations.
- Workflow engine for binders, endorsements, and renewals with versioned documents.
05 — Outcomes
Key results
- Unified insurance workflow
- Dual AI agent system
- Portfolio-aware recommendations
- Broker enablement tools
06 — Stack
What it was built on
Representative tools and patterns — exact vendors vary per client environment.
AI
LLM orchestrationRetrieval over policy corporaEvaluation harness for answers
Core
Policy admin serviceBilling connectorsCRM hand-offs
Trust
Role-based accessImmutable audit trailPII minimisation
07 — Learnings
What we'd tell the next team
- In regulated domains, show citations and confidence - black-box scores erode trust.
- Keep humans in the loop for bind decisions; automate prep, not final judgement.
- Version everything: models, prompts, and policy PDFs drift independently.
FAQ
Questions this engagement anticipated
How did AI stay explainable to brokers?
Answers surfaced source clauses and confidence signals so premiums and exclusions could be defended in client conversations - not black-box scores alone.
What was automated versus human-gated?
Document prep, summarisation, and quote assembly were automated; bind decisions remained in human workflows with audit trails for regulators.
Compare your situation to this case.
Bring your constraints - environment, timeline, and budget. We scope before we quote.