Opportunities
AI Diagnostics for HVAC
Connected through 7 “incumbent in” links and 1 “applies thesis” link.
Opportunities
Opportunities
Connected through 7 “incumbent in” links and 1 “applies thesis” link.
The gap
Wedge
Target commercial refrigeration contractors servicing grocery chains, where system downtime results in immediate product spoilage and massive liability. This niche demands rapid, accurate diagnostics and pays premium rates for immediate fault resolution. Once adopted for refrigeration, the product expands to cover standard commercial rooftop HVAC units, industrial chillers, and eventually building management system integrations.
Timing
Multimodal LLMs now accurately interpret photos of wiring diagrams, degraded nameplates, and digital multimeter readings in real-time. Previously, text-only AI failed because technicians work with visual, physical hardware constraints rather than typed descriptions.
Why This ICP
Mid-sized commercial HVAC contractors face severe skilled labor shortages and high penalty costs for equipment downtime. They employ enough junior techs to feel the acute pain of diagnostic bottlenecks but lack the dedicated internal training departments of massive enterprise fleets.
Size Of Prize
Approximately 115,000 HVAC contracting businesses in the US spend an average of $15,000 annually on senior technician time dedicated to remote phone support and unbillable return trips, yielding an addressable market of $1.7B.
Gap Narrative
HVAC technicians spend hours deciphering complex, non-standardized fault codes across dozens of equipment brands. Current manuals are static, out-of-date, and siloed, requiring endless calls to senior techs or manufacturer support lines. This creates a bottleneck where junior techs cannot independently diagnose commercial units, leading to costly return trips and unbillable hours.
Defensibility
The system builds a proprietary dataset of undocumented fault resolutions, linking specific error codes on aging equipment to the physical interventions that actually resolve them. As technicians upload photos of wiring and part configurations, the model learns brand-specific quirks entirely absent from official manuals, creating deep workflow lock-in as the database compounds.
Why This Thesis
An Agentic Copilot approach fits because the ultimate deliverable is a correct diagnostic conclusion and specific parts list, rather than another software dashboard. Contractors buy first-time fix rates directly, replacing the internal cost of escalating support calls to master technicians.
Overview
Build difficulty
Hardest Part
Mapping highly unstructured technician intake notes and proprietary, undocumented OEM error codes into a unified diagnostic graph with high enough accuracy to dictate purchasing. Recommending the wrong expensive part for a truck roll causes the contractor to lose money and immediately abandon the tool.
Min Viable Scope
Deliver a symptom-to-part recommendation engine strictly for residential central air conditioning units from the top three manufacturers using manual text input from dispatchers. Deliberately exclude commercial rooftop units, heating and furnaces, and direct integrations with IoT thermostats or hardware sensors.
Cold Start Problem
The diagnostic engine requires tens of thousands of validated symptom-to-repair pairs across diverse equipment models before it predicts accurately. Seed the initial model by scraping public HVAC repair forums, digitizing OEM manuals, and executing a bulk ingestion of historical closed work orders from two regional design partners.
Time To First Value
1-2 weeks of onboarding, gated by the ingestion and indexing of the contractor's historical dispatch notes and local parts inventory.
Data Moat Available
true
Technical Difficulty
High
Build profile
Sized prize
IllustrativeIllustrative targets and order-of-magnitude estimates — not an achieved track record. This Thing is concept-stage; real figures come from live data once operating.
SAM
~$400M-600M North American mid-to-large commercial HVAC contractors
SOM
~$15M-30M
TAM
~80k-100k global commercial HVAC contractors × ~$15k-25k/yr diagnostic tooling and labor displacement ≈ $1.2B-2.5B
Growth Rate
~10-15%/yr, driven by a shrinking pool of veteran technicians and the growing adoption of complex variable refrigerant flow systems
Paid Comparable Spend
~$1k-3k/yr per technician on traditional physical gauges and multimeters, plus ~$80-120/hr on wasted senior technician labor during manual troubleshooting
Market sizing
How you know
Kill Thresholds
Leading Metrics
What Proves Right
Technicians upload system parameters and photos of error codes on their first day of use to bypass OEM tech support calls. Service managers purchase fleet-wide licenses at $1,500 per year per technician after observing a 20 percent reduction in repeat truck rolls. Cohorts of junior technicians close diagnostic tickets at parity with veteran technicians within two weeks.
What Proves Wrong
Technicians refuse to input sensor data because manual entry into the mobile app takes longer than checking paper charts. The diagnostic model outputs generic advice that fails to identify specific component faults in complex variable refrigerant flow systems. Fleet managers cancel trials because the software requires constant cellular connectivity in concrete basements and mechanical rooms.
Win conditions