Your dispatch board is training data
Every rescheduled job, parts call-back and “tech is running late” thread is a worked example of how a service business actually runs. AI labs have very little of it, and data buyers have started paying trades companies for it.
Why labs want it
Models have read the public internet. What they haven’t seen is how a 15-person HVAC company decides which truck goes where when a technician calls in sick, how an estimate turns into a work order, or how a warranty dispute gets settled over three phone-call notes and an email.
That is the kind of data buyers now list by name. The Information reported in 2026 that AI data companies are paying HVAC businesses for their data, and Handshake AI lists ServiceTitan among the sources it accepts from companies.
Where this is reported, as of October 2026. None of these companies is a SellMyData customer; they are the public record of what buyers are asking for.
- The InformationReported that AI data startups are paying HVAC companies for data, including a small Michigan contractor approached by micro1. Why Data Labeling Startups Are Paying HVAC Companies for Data.
- Handshake AILists ServiceTitan, QuickBooks and Paychex among the tools it sources company data from. Handshake AI, get started.
- MIT Technology ReviewReports that demand for companies’ operational data took off as labs hit the limits of public and synthetic data. AI labs are buying data from dead startups to train their models.
What sells in field services
The five data types →Messages
Office and crew threads: scheduling changes, parts questions, customer call notes relayed to the field.
Slack · Teams · Gmail
Code
Most trades companies have none, and that’s fine. Skip it.
Usually none
Documents
Estimates, proposals, safety checklists, maintenance agreements and the SOPs new techs are trained on.
Drive · SharePoint · Dropbox
Records
Invoices, job costs, parts orders and the customer-account history behind each job, with identities removed.
QuickBooks · ServiceTitan · Jobber
Tasks
Work orders and service tickets from first call to completed job, including reschedules, call-backs and warranty work.
ServiceTitan · Jobber · Housecall Pro
Tool names are the ones we see most in this industry. If yours isn’t listed, it almost certainly still counts: take inventory and we’ll tell you.
What stays out
- Customer identitiesHomeowner and business-customer names, addresses and phone numbers are replaced with placeholders. The job stays; the person doesn’t.
- Consumer dataWe don’t buy consumer personal data, health records tied to people, or payment card data, and we check that none is mixed into what’s in scope.
- HR and personal channelsDirect messages, HR channels, payroll files and anything about an individual employee stay out.
- Other people’s contentVendor documents, licensed material and anything marked confidential by someone else are left out during the rights check.
Full detail: what we remove and how rights are checked. We don’t buy consumer personal data, health records tied to people, or payment card data.
How it works
- STEP 01
Value it
Answer a few questions and get an estimated range. Nothing is shared yet.
- STEP 02
Share your exports
We walk you through exporting from each tool you choose, for the date range you choose. We never write to your systems.
- STEP 03
Approve the listing
We strip names, contacts and identifiers. You review the redacted preview and the price before any lab sees it.
- STEP 04
Get paid
When a lab buys a license, you get paid. We aim to pay within 7 days of the sale closing.
Today this is done by hand, with you. You export, we clean and prepare, you approve. How you stay in control.
Questions from field services
Our customers are homeowners. Isn’t that consumer data?
The customers’ identities are. Their names, addresses and phone numbers are removed before anything is listed. What remains is the work: the ticket, the diagnosis, the parts, the schedule changes and the outcome.
We mostly run on ServiceTitan and QuickBooks. Is that enough?
Yes. Work orders with their full history plus the matching invoices are exactly what buyers describe. If you also have an office Slack or shared inbox, the threads around those jobs add value.
Do we need to export photos or technician location data?
No. Photos and GPS traces are left out by default. The text of the job, its timeline and its outcome are what matter.
How far back should we go?
As far as your records are complete. Longer histories show how your scheduling and pricing changed over time, which labs value. Three or more years is typical.
Something else? Email micki@sellmydata.com.