Directions Brain

AI Brokerage TAM Model

4 min read

AI Brokerage TAM Model

Definition

The honest, haircut-by-haircut TAM math for smb-insurance-portfolio-brokerage, built to survive investor diligence (full derivation + linked sources in the research-hub artifact, “TAM — honest math” section; compiled Aug 7, 2026). Rule zero: never quote premium as TAM — a brokerage’s revenue is the commission slice only.

Key points

  • The waterfall: US commercial premium 50–60B ⚠ (all sizes, incl. unaddressable enterprise) → **TAM: small-commercial commission pool ~175B premium × 10–15%; recurring at ~90% retention) → **SAM ~200–500 commissions that can’t carry CAC) and heavy-specialty lines → *annually contestable ~2B stream, and the gap audit’s strategic job is widening the stream by giving settled accounts a reason to move.
  • **10-year revenue scenarios (penetration of the 105M/yr (≈ embroker’s plateau — what no-differentiation buys). Base 2% → ~1.1B raised with carrier tailwinds; ≈3x clark after a decade in EU consumer). Bull 5% → ~$1.05B/yr (requires the digital share of small commercial — ~10–13% after a decade of insurtech ⚠ — to roughly double, with us taking a third of the shift; unprecedented for a broker).
  • The hidden product requirement inside the Base case: ~117K accounts at 10% audit→BoR conversion ≈ 1.2M audits over the decade ≈ ~330 audits/day — only achievable with a genuinely self-serve/viral audit funnel, or via the MGA/roll-up shortcut. TAM slides and product architecture are the same decision.
  • What makes it venture-scale anyway: (1) AI-native margins — at 40–55% EBITDA ⚠ (vs industry 26%), Base-case 170–230M EBITDA ≈ 18–26B commission pool toward the 20B annuity, we keep 2–3x more of every dollar in it than incumbents can, and the data ladder opens the $500B pool behind it.”*
  • Critic’s row (why even this may be generous): TAM measured at the top of a rate cycle (CIAB Q1-26: first decline in 33 quarters); the retention that makes our book an annuity makes every incumbent’s book a fortress (5–10-day BoR veto, >90% saves); and the same contestable stream is fished by harper (120M, who buy accounts), and 39K incumbents being armed with AMS-native AI.
  • Per-beachhead-segment commission pools ⚠ (from insurance-beachhead-segment-sprint): robotics integrators ~100–300M/yr, EV/BESS installers ~10–40M/yr. No beachhead alone is a venture market — every pitch must present the beachhead as chapter one of the sequence.

Evidence

  • Built entirely on us-insurance-distribution-economics anchors (S&P/NAIC premium, Big “I” channel share, Munich Re $175B small-commercial, Agentero commissions, Reagan retention/P&L, MarshBerry multiples, US Census business-formation ~5M/yr, CIAB Q1-26). Per-figure links in the research-hub artifact.
  • Flagged unknowns that decide the model: premium split by account-size band (no public dataset — SAM haircut is our estimate), audit→BoR conversion (unmeasured anywhere — the concierge pilot manufactures it), true US small-commercial switching rates (public data does not exist).

Open questions

  • Does the flat-fee alternative (coverwatch’s bet) change the contestable-pool math by monetizing accounts that never flip BoR?
  • At what premium-under-management does the MGA conversation with a carrier realistically open — 200M?

Graph View

Backlinks