Harper

Overview

San Francisco AI-native licensed commercial insurance brokerage (YC W25; founders Dakotah Rice & Tushar Nair). $46.8Mcombinedseed+SeriesAledbyEmergenceCapital(Feb2026;PeakXV,Antler,10XFounders;$54M total). Autonomous placement across 160+ carriers — workers’ comp, GL, professional — with 1–2 day turnaround vs a traditional broker’s 5–7 (company-claimed). First customers October 2024; **5,000+ businesses and just over $6Mannualizedpremiumacross35statesinthefirst13months\ast \ast .TheproofthatVCconvictioninAI-nativeSMBbrokerageisalreadypricedin—and,at$1.2K average premium per account, the live test of whether micro-accounts can carry an AI brokerage.

Relationships

• coverwatch — the portfolio-optimization-first counterpart; Harper optimizes at placement/renewal instead.
• next-insurance · embroker — incumbent digital supply it places against/alongside.

Notes

• Model: classic commission brokerage (BoR) with AI doing intake, carrier navigation, and placement. Focus is fast acquisition + placement of new coverage, not ingesting and continuously optimizing an existing multi-carrier portfolio — the seam left open for the smb-insurance-portfolio-brokerage thesis.
• Confirmed 2026-08-17: the public site has no existing-policy collection flow at all. The onboarding form (business name, revenue, coverage type, state) is a pure new-quote intake — no carrier login, no document upload, no mention of pulling current policies anywhere on the marketing site. This is direct evidence for, not just an inference about, the placement-vs-portfolio distinction above: Harper is not doing what a check-up-funnel-style product does at any visible layer.
• Vendor stack unconfirmed. Checked homepage, TechCrunch’s Feb 2026 raise coverage, and public job-posting search (2026-08-17) for any mention of canopy-connect, insurgrid, herald, or axle — none found. Given Harper doesn’t appear to ingest existing policies at all, it may simply have no need for a policy-data vendor in the first place, which would explain the silence better than a hidden integration would.
• ⚠ GTM corrected 2026-08-19 (web pass). The wedge is hard-to-place risk, not moment-of-new-risk. The site sells to businesses “told your business is too tricky to insure, or too late,” explicitly targeting firms “other insurers turn down” and niches needing specialized underwriting. The 2026-08-10 deck session recorded Harper’s GTM as moment-of-new-risk outbound (reach businesses as they hire/expand); the public positioning does not support that reading. It is a supply wedge — win where the market says no — which buys them less price competition and a reason to exist beyond speed.
• Ten named verticals: construction, restaurants, auto sales, garage services, childcare, adult care, manufacturing, hospitality venues, security, transportation/logistics. Founder framing in TechCrunch: “middle America,” “real-world businesses” — daycares, manufacturers, car dealerships, local bars and restaurants. Notably this is the industrial/main-street field, not the startup/VC field.
• ⚠ Unit economics — the number that matters most. Harper wrote 5,000+ businesses in its first 13 months and just over $6Minannualizedpremiumacross35states\ast \ast \to anaverageofroughly\ast \ast$1,200 premium per account, i.e. ~$120–240ofcommissionperaccount\ast \ast at10–20$25K-premium / ≥$3K-commissionfloorthisKBtreatsastheviabilityline(H-INS-1,fromVerifl{y}^{\prime }sfailureandCoverWalle{t}^{\prime }sdismantling).At15$900K of annual commission revenue on $47M raised. Harper’s entire bet is that AI throughput makes micro-accounts viable — the direct contradiction of H-INS-1. One of the two is wrong, and Harper is the live test.
• ⚠ “1,000+ new customers/month” is a capacity claim, not an acquisition rate. Both TechCrunch and Antler frame it as what the system can process versus a traditional broker’s 20–30/month. Actual acquisition is ~5,000 customers over 13 months (~385/month). Earlier versions of this page recorded it as an achieved rate; that was a misreading.
• Company-claimed 98.78% renewal rate on the public site (2026-08-19). This is the first retention figure published by anyone in the AI-brokerage wave — unaudited, oddly precise, and over a book barely old enough for one full renewal cycle. Treat as marketing, but note it exists: H-INS-8 says nobody has published retention through a cycle, and Harper now nominally has.
• They tried the sell-to-brokers model first and abandoned it. Rice and Nair originally planned Harper as an outsourced sales service for existing insurance agencies; brokerages resisted, viewing it as a threat to their own producers, so the founders used the technology to build their own brokerage instead. This is the category’s clearest empirical data point on the broker-tooling path: the channel rejected it before the economics were ever tested.
• Stated long-game: Rice — “become the voice for entrepreneurs, starting with their insurance, but over time becoming a focal point for all types of things related to risk, compliance, and their entire back office.” Target of 100,000 customers by end of 2026 (from ~6–7K at the Feb 2026 raise) — a ~15x year.
• Intake form (re-checked 2026-08-19): name, business name, email, phone, state, revenue band, industry, coverage-type checkboxes, marketing consent, plus an “existing Harper customer” checkbox for renewals. Still no existing-policy collection of any kind — confirming the 2026-08-17 finding. Carriers named on-site now include Progressive, GEICO and AIG plus “hundreds more” specialty underwriters.
• Part of a 2025–26 YC wave of agentic brokerages: Kinro, Panta, Casey, Fernstone, Acolite (none has portfolio ingestion as the wedge; see smb-insurance-competitive-landscape).
• Web-verified 2026-08-19: harperinsure.com (positioning, verticals, intake form, renewal-rate claim); TechCrunch 2026-02-25; Emergence Capital thesis post; Antler founder story (original GTM, traction); YC company page. Earlier passes: 2026-08-01, 2026-08-17.
• Assessed head-on as the B2C model to beat, 2026-08-19. Saar’s read of the premise: not the cheapest — the most suitable for you, plus ongoing service; the B2C play is to copy it and do it better. His warning in the same breath: going head-to-head against Harper and many others on experience or a marginal automation gain is naive, ends in a CAC bloodbath, and reaches a dead end without a real niche advantage. Guy’s counter-read is that Harper executes badly — no information or quote available before speaking to a rep — leaving room for a genuine direct-to-consumer flow where AI carries the customer far before any human. Also noted: Harper is generalist on coverage, and they are “not an efficiency play” in OpEx terms.

Mentioned in

• 2026-08-03-directions-smb-insurance-value-chain-mapping — wave-3 flagship in the insurtech-waves analysis.
• 2026-08-03-directions-round-2-super-broker-vision — live research: “Harper is us — same promise, exactly what Huri talked about for SMB”; ~1,000 carrier markets, verticalized landing pages.
• 2026-08-07-directions-nizan-saar-ai-brokerage-thesis-and-pitch — positioned as the “speed” archetype the category-creation frame beats; named in the four-rings competitor deep-dive assignment.
• 2026-08-10-directions-huri-pitch-deck-build — its GTM decoded as moment-of-new-risk outbound (reach businesses exactly when they hire, expand, or open a new activity); used as a GTM contrast in the deck’s competition chapter.
• 2026-08-19-directions-b2c-vs-b2b-verticals-and-mga-principles — the B2C benchmark and the CAC-bloodbath warning.