How many of these companies exist, what they already pay for software, and whether the gap Dirk described is real. Sourced where possible; everything modelled is marked est.
HVAC alone is 120,000 businesses. Add plumbing, electrical, appliance repair, garage doors, pest control, landscaping, and the commercial and oil field variants, and the field service universe runs to several hundred thousand US firms. Almost none of them are large.
The fragmentation matters more than the headline. IBISWorld notes the sector has low market concentration — there is no dominant operator, which is precisely the condition under which a horizontal tool can spread. It also means no single customer will ever be large enough to carry us, so the model has to be repeatable from the start.
The number we don't have and should get. How many field service firms sit in the $2–3M revenue band specifically. Census County Business Patterns can be cut by employment size but not cleanly by revenue. Albert is the fastest route to a real answer — he works in this band and will know roughly how many are within reach of a Houston-based team. Worth asking in the first conversation.
| Product | Entry price | Serves | Notes |
|---|---|---|---|
| Jobber | $29/mo | 1–10 people | 14-day free trial. Tiered, grows with users. |
| Housecall Pro | $59/mo | 1–10, residential volume | 14-day free trial. Tiered, grows with users. |
| ServiceTitan | Quote-led | 10+ employees | No published pricing. Annual contract required. No free trial. Reported around $200+/mo per technician for 10+ shops. |
Pricing per a July 2026 comparison from LeadDuo; ServiceTitan does not publish rates, so its figure is reported rather than verified.
The recurring criticisms across the 2026 reviews are consistent and useful to us: per-user fees that escalate as the team grows, separate onboarding charges, and cancellation penalties. One comparison is titled, plainly, "The $299/mo Trap." The pattern is a low advertised entry price that becomes a serious line item at 15 or 20 technicians — which is exactly our target band.
The hole, stated precisely: below roughly 10 employees there are good cheap products. Above roughly 50 there is ServiceTitan and real ERP. In between sits a company paying escalating per-seat fees for scheduling software that still does not talk properly to its accounting system — and whose technicians still capture job detail badly or not at all.
Construction and field trades are at the bottom of every AI adoption table — 13.9% of US construction firms reported using AI in any business function in the July 2026 Census Business Trends and Outlook Survey, against 43.7% in Information and 37.6% in professional services. In the transaction data, JPMorgan Chase Institute puts construction at 8.9% of firms having actually paid for an AI service.
That is usually read as a reason to avoid the sector. It is better read as timing. Construction showed the fastest relative jump of any sector in the most recent BTOS reading, and the same survey found 27.9% of Texas businesses say they don't know whether they'll use AI in the next six months — a larger group than either the adopters or the refusers, and the group most available to be led.
Dirk tried an adjacent version of this at HCL and his own assessment was that it was "before AI, and doing it was clunky." Two things are different now: vision models can read a serial plate or a failed part from a phone photograph, and building a configured agent is days of work rather than months. The problem didn't change; the cost of solving it collapsed.
The barrier to SMB adoption is not price. Across the 2026 surveys, data security (33%) and lack of trust in accuracy (31%) both beat cost (24%), and 78% of small businesses say they don't fully trust AI to handle basic tasks without human oversight. That is an argument for a service with a named accountable human on it — not a product.
The clock we're on. The window is open because integration is hard and the incumbents are complacent. Both of those decay. ServiceTitan can move down-market, and AI makes integration cheaper for everyone including our competitors. Assume 18 to 36 months of structural advantage est, and decide early what we own at the end of it — client relationships, proprietary trade configurations, or the data itself.
Dirk is in Houston. Eduardo is in Dripping Springs, outside Austin. Both are in the top tier of US markets for field service density, and the two cities are three hours apart — close enough to run a single sales territory.
| Austin MSA — verified counts | Establishments |
|---|---|
| Construction, all | 6,153 |
| Specialty trade contractors | 3,804 |
| Plumbing, heating & air-conditioning | 825 |
| Electrical contractors | 642 |
| Residential remodelers | 843 |
Census County Business Patterns 2023. Houston is a substantially larger market again — worth pulling the equivalent cut before the Albert conversation.
Texas also removes a friction: no state income tax, straightforward entity formation, and both founders already resident. One caution carried over from the Meridian work — Texas taxes data processing services, including hosted software and website work, at an effective 6.6%, while pure consulting and advisory is not taxable. If we sell a managed platform to Texas clients, that distinction has to be built into the invoice from day one.
Dirk's method rather than a single bet: put out three, see which bites, then commit. These are the three that fit what we discussed — deliberately different in customer, not just in size.
The classic case. High job volume, low ticket, acute capture pain, Albert's likely territory. Largest number of targets and the fastest path to a reference story.
Fewer, larger jobs. Contractual SLAs, warranty claims, and compliance documentation — where bad capture costs real money and the buyer can quantify it.
Raised on the call. Houston's home industry, very high value per job, heavy documentation and safety requirements. Hardest to enter, best economics if it lands.
Recommendation for the Monday conversation: run Rod 1 and Rod 2 in parallel, because Albert plausibly opens both, and hold Rod 3 until there is one working reference. Rod 3 has the best economics and the longest sales cycle — it is the wrong place to learn.