Insurance Tech & Integration Stack

Full research document: outputs/tech-blueprint-integrations-2026-08.md (Saar’s technology-and-integrations deck workstream from 2026-08-15-directions-acquisition-wedge-and-deck-structure). This page is the durable synthesis.

Definition

The technology architecture and vendor-integration map for the AI-native SMB brokerage, phased against the acquisition-wedge sequencing (2026-08-15-acquisition-wedge-before-workflows). Core thesis: every rail we need exists and is rentable; none is a moat — buy every rail, own every representation. The owned assets are the canonical business/policy data model, the conversation corpus, the NAICS→carrier-class-code mapping, and the change-detection engine — the technical expression of AX-INS-7’s “genuine post-bind continuity.”

Key points

• Architecture = three planes. (1) Acquisition & customer: branded check-up funnel with KYB prefill (Middesk/Enigma), dual-rail policy intake, coverage-picture engine. (2) Brokerage execution: rented in P0 (the partner brokerage), then neutral rater + own AMS + IVANS in P1–P2. (3) The Brain: owned data model, conversation corpus, change detection — never rented.
• Policy ingestion: canopy-connect is the only proven commercial-lines credential-pull rail (verified Aug 2026: BOP/commercial auto/WC/multi-peril, loss runs, EIN/NAICS, 500+ fields; ~$3–6marginalpercommercialpullonpublishedplans).Butitsself-reported\ast \ast 70$17.5M Series A Aug 2026 — the funded version of the data-rail hypothesis’s disconfirmer); insurgrid (acquired by Helium Ventures, Jan 2026) is worth a demo.
• Quoting: appointments are the moat, quoting tech is not. Every neutral rail (herald, Tarmika, Ivans) runs on someone’s carrier codes; everything granting instant codes (Bold Penguin, Semsee MAP, first-connect, wholesalers, networks) takes commission or has a conflicted owner. The P0/P1 architecture: herald API × the partner brokerage’s appointments, plus 3–5 direct API-first carriers (coterie, Pie, NEXT, EMPLOYERS) for a demo-able neutral comparative rater. Commission-neutral display is our choice on top; nobody grants neutral economics.
• Structural risk: applied-systems concentration. Epic + EZLynx + IVANS + Tarmika + Planck under one roof, with demonstrated willingness to cut off AI startups (the Comulate case, Forbes Jan 2026). Mitigation: momentum-amp as system of record (the only AMS with a fully public read/write API — verified), Herald/direct APIs for quoting, IVANS treated as a replaceable AL3 feed we parse ourselves. Add AMS-API accessibility to the partner-brokerage selection criteria.
• Post-bind data still flows as nightly AL3 EDI (IVANS), not APIs — the quoting/servicing asymmetry is the core constraint on an AI-run back office and part of why incumbents are slow.
• Conversation capture is the day-one proprietary asset: recall-ai ($0.50/hrrecording+$0.15/hr transcription) + Twilio → LLM extraction into ACORD-mappable risk data. Nobody owns customer-side SMB advisory-call capture; Gong is a comp, not a component. This is the customer-side twin of the forbidden “Gong for insurance” broker tool — pointed at our own funnel, it stays inside AX-INS-4.
• Doc AI is commodity (H-INS-2 confirmed from the buy side): Sensible-class extraction at ~$0.10–0.50/doc with prebuilt ACORD 25/125/126/130; 92–96% typed-field accuracy, weak on handwriting/checkboxes — fine for a reviewed check-up, not unreviewed binding. Thin in-house LLM pipeline; don’t over-invest.
• **P0 stack rents for <$2K/month\ast \ast andthe1.5-monthMVPtimelinefromthesessionisrealisticforthescopedfunnel(noquoting,noAMS,noIVANS,nolicensingtoolinginP0).Licensingnumberscorroborate[[us-producer-licensing-mechanics]]:$8–15K fees, 2–4 months to a 50-state entity (estimates), vs slow appointments.
• Regulatory boundary for P0 (counsel needed): an unlicensed tech co can take flat referral fees but cannot share commissions or solicit/negotiate; quote presentation and binding must occur under the partner’s license — or we license our own entity early, in parallel, per AX-INS-8.
• Every rail owner is converging on our product (Herald→broker AI agents; CRC REDY INTEL; bold-penguin→digital wholesale and “premier digital broker”; Renaissance→AI member platform). Keep integrations thin and swappable behind our own abstraction layer; speed to distribution beats quoting tech.
• Corollary found 2026-08-15 (insurance-distribution-aggregator-layer): every neutral quoting platform that reached distribution relevance has been acquired by a carrier, MGA, or AMS since 2021 — Bold Penguin→American Family, Tarmika→applied-systems, Talage→Mission, Sayata→Penn-America (Sept 2025), Semsee→iBynd (Mar 2026). Only herald and coverforce remain independent. Neutrality is structurally unstable in this market — which makes AX-INS-7’s incentive-alignment wedge scarce by construction, and simultaneously makes our chosen rails acquisition targets.

Evidence

• 2026-08-15-directions-acquisition-wedge-and-deck-structure — the session that assigned the workstream and fixed the integration list (Canopy/InsurGrid, comparative raters, Teams/Zoom/Meet capture, aggregation/webhooks).
• Web research (Aug 2026), load-bearing claims verified against vendor docs: usecanopy.com/commercial (commercial lines, quote-point pricing), heraldai.com/docs (appointments required, not provided), api.momentumamp.com (public REST/OData API). Full source list in outputs/tech-blueprint-integrations-2026-08.md.
• 2026-08-16-directions-broker-jobs-to-be-done-map — saar-arbel assigned to research brokerage APIs/infra providers (incl. coverforce) as part of scoping which rails Actually integrates.

Open questions

• Canopy’s real commercial carrier list and pull success rate against our target carrier mix — vendor call scheduled intent; “90% standard commercial coverage” is an unverified vendor claim.
• The two funnel kill-gate numbers (policy-handoff completion %, audit→BoR conversion) — unknowable until the pilot instruments them (policy-ingestion-requires-a-data-rail’s settling number).
• Herald’s live carrier connections vs its 86-carrier index, pricing, and precedent for the partner-brokerage model — plus the risk it becomes a competitor.
• Whether candidate partner brokerages run momentum-amp/HawkSoft (cheap integration) or Applied Epic (gated, revocable).
• First Connect and Smart Choice commission splits (undisclosed) — P2 market-access economics.

Related

• smb-insurance-portfolio-brokerage — the direction this stack serves.
• insurance-distribution-aggregator-layer — the same layer assessed as competitors and cost structure rather than vendors.
• 2026-08-15-acquisition-wedge-before-workflows — the phasing this architecture implements.
• policy-ingestion-requires-a-data-rail · us-producer-licensing-mechanics · us-insurance-distribution-economics · carrier-access-not-licensing-gates-new-brokerages
• canopy-connect · herald · coverforce · axle · insurgrid · applied-systems · momentum-amp · first-connect · coterie · recall-ai
• context-os-brain — plane 3 is the Brain pattern applied to the insurance vertical.