BramForgeLabs / Products / Insurance AI Radar
Insurance · AI Solutions
Not another market report. A two-layer radar: scan the value chain for production vs theatre, then take a buildable agent spec into a POC that would survive an audit. Life-first. Vendor-neutral. Regulation designed in.
The Problem
It has a production problem. Most "AI in insurance" writing is a model-type essay or a vendor brochure. Neither tells a chief underwriter, claims lead, or CRO where to start — or how to build something a regulator can reconstruct six months later.
Two layers
The Radar
The build spec
The interesting number is not 39 ideas. It is 6 in production and 13 that should not be built as autonomous agents yet. The Radar's job is to stop a year spent in theatre.
Architecture, not a template
Every detailed spec uses the same orchestration the IT playbook uses for knowledge work — retrieval, synthesis, verification, escalation — because insurance AI that cannot cite its source is not a system, it is a liability.
Retrieval Agent → Synthesis Agent → Verification Agent → (low confidence / ungrounded / automated-decision risk) → Escalation Agent
Worked example — Use Case 1.4 APS Retrieval & Clinical Summarization. The agent does not issue, decline, or rate. It produces a cited clinical packet a human underwriter can use.
| Agent | Responsibility | Output |
|---|---|---|
| Retrieval | Page-aware chunks from the APS and e-App impairments | Top-k spans, page numbers, source hashes |
| Synthesis | Structured clinical packet from retrieved context only | JSON + narrative; every finding has a source span |
| Verification | Grounding, purpose-limitation, prohibited fields | Verified packet, or escalate with reason |
| Escalation | Human UW queue when confidence is low or Law 25 is in play | Work item + immutable trace |
The Catalogue
Depth is life insurance, with a Canadian regulatory overlay. P&C is a second pane, not a second book.
| # | Domain | Use cases | n |
|---|---|---|---|
| 01 | New Business & Underwriting | App triage, KYC/AML, evidence procurement, APS summarization, risk synthesis, Law 25 disclosure, STOLI detection | 7 |
| 02 | Claims | FNOL triage, document ingestion, adjudication assist, SIU/fraud, leakage/subrogation, settlement explanation | 6 |
| 03 | Policy Servicing | Request orchestration, billing exceptions, loans/withdrawals, beneficiary change, contract assembly | 5 |
| 04 | Distribution & Advisor | Licensing/commission, illustration assist, suitability check, case-status communication | 4 |
| 05 | Actuarial & Reserving | Experience-study narrative, assumption packs, IFRS 17 commentary (human-signed) | 3 |
| 06 | Customer, Retention & Conduct | Lapse/save, complaint/TCF routing, consent-limited next-best-action, fairness monitoring | 4 |
| 07 | Risk, Compliance & Governance | E-23 inventory, model drift, shadow AI, Law 25 consent/PIA, sanctions/PEP | 5 |
| 08 | P&C Operations | CV damage assessment, cat-event surge, telematics, repair-cycle, contents inventory | 5 |
First three POCs
High value, medium effort, audit-friendly. The house pattern. Free preview on this site.
Production-shaped, easy, visible to operations in two weeks.
The E-23 evidence cell that makes the other two defensible.
The 11-section template and the house pattern, without the private fleet roster. Open Method →
Who this is for
Packets, not copilots. Cited findings a human can sign.
FNOL and fraud that already work; coverage agents that can point at the policy.
E-23, Law 25, purpose limitation, and an audit trail as the design, not the appendix.
A reusable radar + spec set. One purchase, every carrier engagement.
Set expectations
Before you buy
No email required. The full APS summarization spec — agent table, failure modes, Section 11 how-to-build.
Pricing
$89 CAD
One-time · 192-page PDF · no DRM
Email checkout until Lemon Squeezy is wired · same 30-day guarantee as the IT playbook
D·A·D subscribers get first notice when card checkout goes live. Subscribe free →