WarrantyRail — Post-install registrar for US HVAC contractors

1. One-liner

Auto-files HVAC warranty registrations and utility rebates from install-day photos so small US contractors stop year-7 compressor callbacks.

2. Trend signal — why now?

Three things landed in the last 18 months that together make this build now-or-never:

1. The 2026 IRS PIN regime took effect. From Jan 1, 2026, every 25C-style energy credit claim relies on a 17-character Product Identification Number tied to a manufacturer-registered AHRI match. Contractors are suddenly expected to hand homeowners a paperwork packet they were never trained to produce. Manufacturers must now file quarterly to the IRS Energy Credits Online (ECO) portal. The closeout paperwork burden just stepped up an order of magnitude. (IRS QM Requirements)
2. OBBBA sunset the federal 25C credit Dec 31 2025 but installs through 2025 are still being claimed in 2026 tax season — every homeowner of those ~3M installs needs their PIN packet, and state programs (Mass Save, NYSERDA, NJ Clean Energy, TECH-CA, SRP, SMUD, NIPSCO) are now the primary incentive path going forward. The state patchwork got bigger, not smaller. (2026 Home Energy Credit Sunset)
3. Browser-Use + Skyvern made manufacturer-portal scraping a weekend project. Skyvern raised $2.7M in 2024 and the open-source ecosystem matured through 2025–26. LLM+vision agents now reliably navigate the 12+ different Carrier/Trane/Lennox/Goodman/Rheem/Bosch portals without per-site hand-coded selectors. The bottleneck has flipped from “impossible to automate” to “nobody focused on this niche.” (Skyvern)

Stacked against the underlying scale: ~3M residential HVAC systems replaced in the US each year, 78% of contractors have <10 employees, average shop revenue $480K–$1.2M (HVAC Statistics 2026). The administrative-back-office bench is one person — or zero.

Provenance:

• Signal 1: 2026 IRS PIN requirement makes contractor closeout paperwork a regulated artifact — IRS guidance — observed 2026-05-15
• Signal 2: Trane/Carrier 60–90 day registration windows + real legal cases of voided warranties — Avvo legal question + Trane warranty page — observed 2026-05-15
• Signal 3: Skyvern raised $2.7M to commoditize browser-portal automation; ESP32 + vision LLM stack now cheap — Skyvern blog — observed 2026-05-15 Category: Tech-unlock

3. The opportunity

Every HVAC install creates a ticking paperwork bomb:

• Manufacturer warranty registration within 60–90 days (Trane 60d, Carrier 90d, Goodman 60d) — miss it, the 10-year warranty drops to a 5-year base, and year 7 a $1,500–$3,000 compressor failure becomes a 1-star Google review and a small-claims summons against the contractor.
• Utility rebate filing within 60–90 days of install on programs like Mass Save (up to $10,000), NYSERDA, NJ Clean Energy, SRP, SMUD, TECH-CA. Mass Save denies applications on any missing field; processing then takes 6–8 weeks even when right.
• Homeowner 25C packet (for 2025 installs claimed in 2026, then state programs going forward) — itemized invoice, AHRI certificate, manufacturer certification statement, the new 17-character PIN per outdoor unit.
• Municipal permit closeout — final inspection sign-off, AHRI match upload to permitting portal.

ServiceTitan and Housecall Pro own dispatch, billing, and CRM. Nobody owns the post-install paperwork tail. Contractors either pay a $40K/yr admin to keep the spreadsheet, fold rebates into a higher quoted price (and pocket the delta), or skip the work and gamble. Year 5–10 the gamble loses.

The wedge: a contractor uploads the install nameplate photo and signed invoice. WarrantyRail’s vision model extracts model/serial/AHRI match, the agentic browser files the warranty on the right manufacturer portal, files the rebate on the right state portal, builds a homeowner “trust packet” PDF (warranty proof + AHRI cert + PIN + tax form + permit closeout), and chases anything that stalls. The contractor closes the install instead of dragging it into next quarter.

4. Target market

• Primary customer: Owner-operator of independent residential HVAC contractor in the US, 1–10 employees, 100–500 installs/year, revenue $500K–$5M. Geographic sweet spot: Florida, Texas, Massachusetts, NY/NJ, Pacific Northwest, Mid-Atlantic (Mass Save, NYSERDA, NJ Clean Energy, SRP, SMUD coverage).
• Why they buy: Year 7 compressor callbacks where warranty was never registered. State rebate denials that customers blame on them. New 2026 IRS PIN regime that nobody on the truck knows how to handle. Office admin already drowning — owner working evenings on paperwork.
• TAM reasoning: US has ~120,000 HVAC contracting firms. ~50,000 do residential installs. ~30,000 in residential-heavy heat-pump and tax-credit-active states. At $99–199/mo + $5–10 per install fee, capturing 5,000 of them is $7–12M ARR territory. Capturing 1,000 is $1.5–2M ARR.
• Why now for them: 2026 PIN rule is brand-new; nobody has tooling. Heat pump volume keeps rising (North America residential heat pump market $17B → $52B by 2034, 13% CAGR — Market Data Forecast). State rebate programs are getting tighter on documentation, not looser. Florida’s HB 1203 (2023) made auto-registration a legal expectation — other states will follow.

5. Product sketch (MVP)

• One-tap install closeout. Tech snaps photos of outdoor unit nameplate, indoor coil nameplate, and signed invoice. WarrantyRail extracts model/serial/AHRI match number, install date, contractor license, and homeowner contact.
• Auto-file warranty registration with the right manufacturer portal (Trane, Carrier, Lennox, Goodman, Rheem, Bosch, York, Daikin, Mitsubishi, LG — the 10 brands covering ~90% of installs). Stores screenshot of confirmation.
• Auto-file utility rebate for the homeowner’s address — top 12 state programs covered at launch (Mass Save, NYSERDA, NJ Clean Energy, TECH-CA, SRP, SMUD, NIPSCO, Xcel CO, Eversource CT, ConEd, PG&E, NV Energy).
• Homeowner trust packet — branded PDF with warranty registration screenshot, AHRI certificate, manufacturer certification statement, IRS 25C PIN (for 2025 installs being claimed in 2026), copy of permit, and a “what to do if your unit fails” page. Sent via email + SMS the day after install.
• Deadline radar. Every install has a 60d / 90d / rebate-window clock. Anything that stalls auto-pages the office admin until it clears.
• Reverse-lookup mode. Upload 12 months of past invoices — WarrantyRail back-registers anything still inside the window (a one-time win that makes a contractor feel “we just got our liability down”).
• Per-state rebate library. When a state changes forms or windows (and they do), one update propagates to every contractor on the platform.

6. AI angle — what’s load-bearing

Three layers, all load-bearing:

1. Vision extraction. Off-the-shelf vision LLM (Gemini 2.5 / Claude 3.7) reads messy nameplate photos — refrigerant type, model numbers half-obscured by zip ties, hand-scribbled install dates. Without reliable photo→structured-data, the whole flow needs manual entry and the value prop dies.
2. Agentic browser automation. Skyvern / Browser-Use style agents fill the 10–15 different manufacturer + utility portals. Selectors break monthly; an LLM-driven agent self-heals when forms change. Traditional Selenium scripts would be a full-time maintenance job.
3. Rebate eligibility reasoning. Each state program has 30+ conditional rules (panel ampacity, climate zone, equipment SEER2/HSPF2 thresholds, income tier, prior rebates). LLM is the only practical way to turn 200 pages of program rules into a per-install “yes/no/which-tier” decision.

Remove the AI and you have a wrapper around a spreadsheet plus a fragile RPA stack. Not a business.

7. Localization angle (if any)

N/A — this is a US-first play. The hook is US manufacturer registration portals, US state utility rebate sprawl, and the 2026 IRS PIN regime. There is a future Canada cut (Greener Homes Grant, BC Hydro, Hydro-Québec, Enbridge) and a UK cut (ECO4, Boiler Upgrade Scheme — fewer brands, simpler) but those are 18-month expansions, not v1.

8. Business model — path to $1M–$5M ARR

• Pricing: $149/mo base (up to 25 installs/mo) + $7 per install over plan. Premium tier $299/mo with reverse-lookup, white-label trust packet, and priority support.
• ACV: Average shop doing 200 installs/yr lands around $2,400/yr (base + per-install). Larger shops $4–6K. Blended ACV target $2,200.
• $1M ARR math: 455 contractors × $2,200/yr = $1.0M. Hit this with one good distribution channel.
• $5M ARR math: 2,275 contractors × $2,200/yr = $5.0M — roughly 7.5% of the ~30,000 addressable contractors in primary states. Plausible inside 24 months if the per-install fee model holds.
• Expansion path: (a) charge homeowners $19 for the trust packet (contractor pockets margin or comps), (b) sell extended-labor-warranty referrals (JB Warranties pays 10–15% on policies sold), (c) add municipal permit autofiling, (d) Canada + UK cuts at year 2.

9. Go-to-market wedge — first 100 customers

1. The “back-register your last 12 months” hook. Cold-email 2,000 owner-operator HVAC contractors from state contractor-license registries (publicly downloadable in TX, FL, CA, NY, MA, NJ). Offer: “Upload your last 12 months of invoices, we’ll back-register every warranty still inside the window, free. After that it’s $149/mo.” Expect 8–12% conversion on the free audit, 30% of those convert to paid. → ~50 customers from one campaign.
2. HVAC-Talk + Facebook contractor groups. “Independent HVAC Contractors” and “Heat Pump Installers” Facebook groups + the HVAC-Talk forum have hundreds of owner-operators arguing weekly about warranty disputes and Mass Save denials. Drop a free state-rebate-deadline tracker as a lead magnet → email funnel → demo. Slow but compounding.
3. State-utility partnership. Mass Save and TECH-CA both maintain “participating contractor” lists. Offer the state program a no-cost contractor portal in exchange for a co-branded launch email to their installer list. Mass Save alone has ~1,800 participating contractors.
4. Bird-dog ServiceTitan / Housecall Pro integrations. Build a one-click integration with HCP’s API, get listed in their marketplace. HCP has tens of thousands of HVAC seats; even 0.1% installation conversion is meaningful.
5. Insurance-side referral. JB Warranties and similar extended-labor-warranty providers want contractors with clean registration trails — they currently bear the loss when warranties die. Co-marketing with them puts WarrantyRail in front of their existing 5,000+ dealer network.

100 customers in 6 months is achievable with channels 1 and 2 alone.

10. Build complexity — justification

Medium. Off-the-shelf stack: vision LLM (Gemini/Claude), browser-agent framework (Skyvern open-source or roll your own with Playwright + LLM), Next.js app, Postgres, Twilio/SendGrid. Custom work is real but contained: per-manufacturer portal templates (10 brands × ~2 days each), per-state rebate program templates (12 states × ~3 days), rule engine for eligibility, AHRI directory integration. A 2-person team — one full-stack, one domain expert with HVAC sales/install background — should reach paid pilot in 10–14 weeks and v1 in 4–5 months. The 6th-month risk is portal drift maintenance: if 3 manufacturers each refactor their portal in the same quarter, the AI agent had better self-heal.

11. Gating checklist

Gate	Pass?	Note
Legal in target market	✅	Filing on behalf of contractor under their credentials — same as a $15/hr admin would do
Ethical — no harm / dark patterns	✅	Reduces homeowner risk; transparent trust packet
Market exists (evidence above)	✅	3M installs/yr, real legal cases, $10K Mass Save rebates
1–5 person team can build this	✅	2 people, ~4 months to paid v1
Launchable with <$50K / ₹40L	✅	~$15K LLM credits + $20K founder runway covers MVP

All five pass.

12. Feasibility score

Axis	Weight	Score	Notes
Problem intensity	20	17/20	Year-7 warranty void = real lawsuits + 1-star reviews. Rebate denial = customer rage. Owner already losing sleep. Not 18 because it’s “future pain” — owner doesn’t feel it until a claim comes in.
Demand evidence	15	13/15	Three independent signal types (legal cases, manufacturer policy, state programs, IRS rule). No direct competitor doing all three together; adjacent spend (JB Warranties, Housecall Pro) confirms wallet. Not 15 — no public revenue from an identical play.
Build feasibility	15	11/15	Vision + agentic browser are mature, but 10 manufacturer × 12 state portals is real integration grunt. Portal drift maintenance is forever.
Distribution clarity	15	12/15	Multiple named channels (state license registries, HVAC-Talk, Mass Save partnership, HCP marketplace). The back-register hook is a strong wedge. Not 14 — cold-email to contractors converts but is grindy.
Revenue mechanics	15	12/15	Pricing benchmarked against HCP/ServiceTitan; contractor is used to paying for back-office tools. ACV $2,200 is realistic. Per-install fee may face pushback from owners who’d rather pay flat.
Time to first revenue	10	8/10	8–12 weeks to paid pilot via the back-register hook — contractor pays after seeing the audit.
Defensibility	10	7/10	Soft moat: state-by-state rebate rule library compounds; manufacturer portal templates compound; deadline-tracking workflow lock-in. Copyable by anyone willing to do the grunt work — execution moat.
Total	100	80/100	STRONG GO

13. Qualitative modifiers

Founder-fit tags

technical-heavy · domain-expertise-required

A solo non-technical founder can’t build this. Browser agents, vision extraction, and rule engines need real engineering. But pure engineering without HVAC domain knowledge ships a product contractors don’t trust. Best team: a senior full-stack engineer + an ex-HVAC operations manager or a service-tech-turned-trainer. Bonus if one has worked at JB Warranties, ServiceTitan, or a state utility program office.

Key assumptions to validate (3–5)

1. Assumption: Owner-operators feel post-install paperwork pain badly enough to pay $149/mo even when their current process is “we mostly skip it.” How to test: 30 cold-call interviews with FL/MA/NJ HVAC owners (license registries are public). Ask: “When did you last register a warranty?” “Have you ever had a year-7 callback?” “What did Mass Save denials cost you last year?”
2. Assumption: Manufacturer portals tolerate agentic submission without rate-limiting or banning. How to test: Build the Trane + Carrier + Lennox flow first. Submit 50 real registrations across 30 days from different IPs. Measure: any blocks, any TOS objections, any rejected registrations.
3. Assumption: Back-register hook converts at >5% from cold email. How to test: Run a 2,000-shop campaign with a free 1-month audit offer. Track: open / reply / audit-accepted / paid-conversion. Kill if paid-conversion <3%.
4. Assumption: State rebate library can be maintained by 0.5 FTE once built. How to test: Time the initial build of 3 state programs (Mass Save, TECH-CA, NJ Clean Energy). If maintenance per state per month exceeds 4 hours, the economics break above 20 states.
5. Assumption: Liability framing (vs. revenue framing) is the right wedge. How to test: Run two landing pages — “Stop year-7 callbacks” vs. “Capture $2K rebates you’re leaving behind” — and measure which converts demo bookings 2× the other.

Risk flags

1. Platform dependency: Manufacturer portals can change ToS to forbid third-party submission. Mitigation: position as the contractor’s tool acting under their credentials, not a third-party service. Florida HB 1203 precedent of legal auto-registration is helpful tailwind.
2. Regulatory drift: OBBBA could pull more state rebate programs (Trump admin already threatened HEEHRA). If federal funding evaporates from state programs, half the rebate value proposition vanishes overnight. Warranty side stays intact.
3. Vertical-software incumbents: ServiceTitan or Housecall Pro could bundle this in 12 months. Mitigation: own the deadline-tracking + rebate-rule library moat, partner with HCP rather than compete head-on.
4. Per-state grunt work: 50-state coverage is a 2-year slog. Stay focused on top-12 states for first $2M ARR; resist national-coverage demands from prospects too early.

14. Structured verdict

Score:                  80/100
Verdict:                STRONG GO
Confidence:             Medium
Best-fit builder:       Technical founder + HVAC ops domain advisor (ideally ex-service-tech, ex-state-utility-program-officer, or ex-warranty-provider)
Time to revenue:        8–12 weeks to first paid pilot via back-register hook
Capital to launch:      $30–50K ($15K LLM credits, $5K state-data + Lyft for contractor visits, founder runway 4 months)
Top 3 assumptions to validate first:
  1. Owner-operator HVAC contractors pay $149/mo for back-office paperwork autopilot — validate with 30 cold interviews + 1 paid pilot inside 6 weeks
  2. Manufacturer portals (Trane, Carrier, Lennox) accept agentic submission at scale without blocks — validate by submitting 50 real registrations in 30 days
  3. Back-register hook converts cold contractors at >3% paid — validate via 2,000-shop email campaign
Kill criteria:
  - Abandon if <3 of 30 cold interviews say they'd pay $149/mo
  - Abandon if 2+ manufacturer portals actively block agentic submission and contractors won't share login credentials
  - Abandon if back-register email campaign paid-conversion <2% after 1,000 sends
  - Abandon if ServiceTitan or Housecall Pro ships a native warranty+rebate autofiler in v1's 6-month build window

15. Next step — 1-week validation sprint

• Day 1–2: Pull public contractor license registries for FL, MA, NJ, TX. Build a list of 200 owner-operator HVAC shops (1–10 employees, residential heavy). Draft cold-call script focused on year-7 callbacks and Mass Save denials.
• Day 3–4: Run 30 cold calls. Ask: “When did you last register a warranty?” “Have you ever lost a Mass Save / utility rebate to a paperwork mistake?” “If a tool back-registered your last 12 months of invoices for free, would you pay $149/mo to keep it running going forward?” Record verbatim quotes.
• Day 5: Manually back-register 3 real warranties (Trane, Carrier, Lennox) against the live portals using a friendly contractor’s credentials. Time it. If each takes >10 minutes manually, the unit economics for AI automation are great. If <2 minutes, the wedge is weaker.
• Decide go / no-go: GO if ≥6/30 interviews say “I’d pay $149/mo today” AND at least one contractor agrees to a paid 30-day pilot AND manual back-registration takes ≥5 min per warranty.