George Eliadis spent six summers pressure washing houses in upstate New York, missing calls from the top of a ladder because he couldn't hear his phone over the spray. Every missed call was a job gone — and he was the whole company. Years later, inside a 260-technician shop, he watched the same thing happen at scale: not one missed call but hundreds of jobs a day that someone has to catch, place and route before they slip away. That function has a name — dispatch — and Eliadis just raised $40 million from two of the most prestigious venture capital funds in the world — Andreessen Horowitz (a16z) and Sequoia Capital — to dominate it across a $700 billion industry.

His company, Probook, gets filed under dispatch software. That’s technically true, but wildly incomplete. The dispatch board is the one place in a home-services business where you can watch every job happen, across every truck, in real time. If you own that seat you own the only complete record of how the business actually runs, and that seat is exactly what a16z and Sequoia bought.

The Seat, Not The Software

The timing isn’t an accident. Private equity has been rolling up local HVAC, plumbing and electrical businesses, stitching together fragmented operators into regional platforms. The thesis is straightforward: buy local businesses, professionalize operations, centralize back-office functions and expand margins. This roll-up wave hit 88% growth year over year through mid-2025, but this only works if you can actually take cost out of businesses that are mostly labor. For decades, that was extremely hard. These companies were local, operationally messy and deeply dependent on people who knew the rhythm of the shop: which technician could handle which job, which customer needed a callback, which appointment was likely to run long, which route made no sense on a Friday afternoon.

AI is the first technology in a generation that can credibly touch that mess, and so the industry went shopping. The result, in many cases, was five different AI point solutions stapled onto one business: a voice agent for missed calls, a chatbot for the website, a follow-up bot for unsold estimates, a scheduling tool, a customer-service assistant. Each tool owned a slice of the workflow. But private-equity-backed platforms don’t want five AI vendors. They want one connected operating layer, wired into the place where utilization, revenue and customer experience are actually won or lost. In home services, that place is dispatch.

The proof is already on the board : An Indiana shop with 14 locations and 260 technicians booked 2,542 jobs in its first month on the platform without a human touching a single booking. A Kansas operator grew its average job revenue 20% on a leaner team. Each new location makes the seat more valuable, because the record gets deeper and harder to rebuild from outside.

Probook isn't the only one who's figured out what physical-world data is worth. The entire AI market is waking up to the same problem: the internet trained language models, but the physical world did not conveniently upload itself.

A German startup called MicroAGI is taking a very different route: Through an app called Shift, it’s offering free home cleanings in New York in exchange for permission to record the cleaning. Every cleaner wears a head-mounted camera, and the footage gets sold as training data to robotics labs. The pitch is almost disarmingly simple: they clean, you pay nothing, and in return they record the work. It runs more than 10,000 camera-wearing operators across 15 countries at roughly $20 an hour, and it has said the next categories are plumbing and cooking.

It sounds strange because it is strange. And revealing.

Strip away the cameras, and Shift and Probook are chasing the same thing: operational data from physical work that doesn’t exist in a model’s training set. The difference is the direction of the money. Shift pays to harvest the data: It sends workers into homes, subsidizes the labor, records the task and sells the footage downstream — “Data collection as a service”.

Probook gets paid while the data compounds. Customers buy the product because it solves an immediate pain: missed calls, underutilized technicians, inefficient scheduling, lost revenue. The dataset is a byproduct of usefulness.

Which brings up the company everyone in the trades is actually watching: ServiceTitan, the $6.3 billion public incumbent with roughly $960 million in revenue and its own AI scheduling product. For now, Probook is listed as a ServiceTitan partner — the two technically work together. The open question is how long a multibillion-dollar incumbent stays comfortable next to a startup that owns the one screen its customers can't run the business without.

What Probook is building in the meantime is hard to copy. a16z’s David Haber calls it a "years-old structural moat": the accumulated dispatch record, plus the trust to sit in the seat in the first place. Eliadis was his company's only salesperson until February; He's been to customers' weddings and slept on their couches. That's pretty difficult to copy.

The trades are the test case, not the whole story. The same pattern will show up anywhere physical work creates valuable operational data: logistics, home health, construction, field service, manufacturing, elder care. The strategic question will be the same in each market: who sits in the operating seat, and who's paying whom for what it sees. The durable businesses are the ones whose customers pay them while a proprietary real-world dataset quietly compounds.

But keep an eye on the human labor line underneath it all. The exact same automated seat that flawlessly books 2,542 jobs untouched is the seat that can compress a dispatch team from 22 people down to 10. The data captured to make the trades run smoother today is the data that trains the systems that will run them tomorrow.

The trades were supposed to be the corner AI couldn't reach. Instead they're the clearest preview of where this goes, and whoever's in the chair when the work comes through doesn't just run the business — they own the record of how it runs. Everyone else is basically renting.