Data licensing
Yes, and the trades are the one industry where the software holding those records may already have AI training rights of its own. Housecall Pro and ServiceTitan say very different things about that, and an owner being asked for exclusivity needs to know which set he is under.
Disclosure: BLACKSIG may receive a referral fee from buyers named on this page when a company we introduce licenses its data. There is no fee to the company.
An HVAC, plumbing or roofing company can license its job data to AI companies, and the trades are the one industry where the software holding those records may already have training rights of its own. Housecall Pro's terms, last modified 07/01/2026, say that "Individual customer data entered or imported by Pros shall remain the property of the Pro" and also that "Your User Data may be used to train and enhance our proprietary AI models", with data used for AI training "retained as part of our AI model's knowledge base". ServiceTitan's terms, last updated September 16, 2026, say "Customer owns and retains all right, title, and interest in and to all Customer Data", that it may "train, tune, and improve AI models solely to provide the Service to Customer", and that it "will not (1) use Customer Data to train AI models for the benefit of any third party". You own your records under both. What differs is who else is already learning from them, and that bears directly on any exclusivity you are asked to promise. According to Polyshares' track record page, read on November 10, 2026, it has licensed a roofing company's records for $450,000.
Yes, an HVAC, plumbing or roofing company can license its job data to AI companies: a trade company owns the record of how its own technicians diagnosed, quoted and completed work, and it can license that record under a written agreement signed by an owner or officer. The trades are among the better-fitting industries for this market, because the job file carries a symptom, a diagnosis, a recommendation, a price and an outcome, and that sequence is exactly what a model cannot learn from the public internet.
The published comparison is real. Polyshares lists four selected licenses on its track record page, read on November 10, 2026: $450,000 for a roofing company, $300,000 for a general contractor, $275,000 for a solar company and $110,000 for a global marketing agency. Each is described as a "Perpetual exclusive license to the operational record. Anonymized before delivery", with an offer "within one to two weeks" and payment "net 30 to 60". The roofing figure is the highest of the four and the only published number from this part of the market.
Two conditions sit under the yes and both are checkable before you talk to anybody. The records have to be in software you can export, going back several years, with the reasoning attached rather than just the invoice. And the customer information in them has to come out, which in this trade means names, addresses, phone numbers, payment details and anything in a technician's free-text notes that identifies a household.
The condition specific to the trades is the third one, and it is the subject of the next section. Your field service platform has its own terms about AI, those terms differ sharply between the two platforms most of this industry runs on, and an owner signing an exclusive license should know which set he is under.
Your field service software may already be training its own AI on the records you are about to license, and the answer depends entirely on which platform you run. Both of the major platforms say the records are yours. They say very different things about what they do with them.
Housecall Pro's terms, last modified 07/01/2026 and read in full on November 10, 2026, state that "Individual customer data entered or imported by Pros shall remain the property of the Pro". Under the heading Consent to Data Use and AI Integration, they state that "By using our services, you acknowledge and consent to our collection, use, and processing of your information", and under AI Model Training that "Your User Data may be used to train and enhance our proprietary AI models". Under Data Retention they add that "Data used for AI training purposes may be retained as part of our AI model's knowledge base". The terms also say that once data is integrated with third-party AI models, "your User Data (including any data derived therefrom) may continue to be used, analyzed, and enhanced" by Housecall Pro and its third-party providers.
ServiceTitan's terms of use, last updated September 16, 2026 and read in full the same day, take a different line. At section 8(a), "Customer owns and retains all right, title, and interest in and to all Customer Data". At section 10(ii), ServiceTitan may "train, tune, and improve AI models solely to provide the Service to Customer", and it "will not (1) use Customer Data to train AI models for the benefit of any third party". It separately restricts customers from using its outputs to "develop, train, improve, or benchmark any AI model" that competes with the Service.
Neither of those stops you licensing your own records. What they change is what you can honestly promise. A buyer asking for an exclusive license is asking you to warrant that nobody else holds these records for training, and the answer to that question is in your software agreement rather than in your intentions.
A roofing company's published license is $450,000, the highest of the four figures on Polyshares' track record page, read on November 10, 2026. That is the only named number from this part of the market, and the buyers' own tiers put the plausible range around it.
micro1 publishes, on its data partnerships page read the same day, $100K to $500K for qualified partnerships, $500K to $3M for large-scale or multi-team data and $3M+ for highly unique data. Scale AI shows "$10K to $1M+" on its data partnership page and labels it illustrative value per data partnership, scaling with cadence. Handshake's page leads with "Earn $100K to $4M+". Every one of those is a ceiling attached to a word like unique or large-scale, not a quote for a trade company.
What lifts a trade company's number is specific and it is worth auditing honestly:
What lowers it is equally specific: a short history, records split across a platform migration with the old data stranded, jobs closed with a line item and nothing else, and heavy reliance on phone calls that were never transcribed. That last one is where a large share of this industry's actual reasoning goes, and it cannot be licensed at all unless the recordings were made with all-party consent.
Most established HVAC, plumbing and roofing companies qualify on the published size screens, and record depth is what decides an offer. Here is what each buyer publishes, read on its own pages on November 10, 2026. Where a buyer publishes nothing on a point, the table says so rather than guessing.
| Buyer | Published size screen | What it says sets the payment | Published payout range |
|---|---|---|---|
| micro1 | 30 or more full-time employees; several years of records across multiple systems preferred; records primarily in English; US first | "the quality, uniqueness, and value of the data provided" | $100K to $500K for qualified partnerships, $500K to $3M for large-scale or multi-team data, $3M+ for highly unique data |
| Mercor | Not published | "Compensation is based on data volume, the tools you connect, and the depth of your data" | Not published |
| Scale AI | "40+ employees is our preferred starting point", assessed case by case | Beyond size: established data systems, real operational workflows, domain depth, stakeholder alignment | "$10K to $1M+", labelled illustrative and scaling with cadence |
| Handshake | Not published; its estimator asks employees, tools used, years in business and country | Not published | "Earn $100K to $4M+" as a headline |
| Polyshares | Not published | Not published | Four selected licenses, including $450,000 with a roofing company |
One nuance matters for this industry more than for most. A trade company with 25 technicians and eight years in one platform is a better dataset than a company with 60 technicians and two years split across a migration, and the published screens do not see that. If you are under a headcount threshold but long on record, say so in the first conversation rather than filtering yourself out on the form. Every buyer here is read side by side in which AI company pays the most for business data.
A trade company has to leave out everything that identifies a household, and that is a bigger share of its records than owners expect, because the customer is a person rather than a business. The exclusions below come out before a dataset is scoped, not after a buyer asks.
Customer identity in all its forms: names, service addresses, phone numbers, email, payment details and anything in a free-text technician note that makes a household identifiable. Service addresses deserve particular care, because an address with a dated scope of work is identifying on its own even after the name is removed.
Call recordings made without all-party consent, which in several states means every party on the line. Many trade companies record every inbound call. That archive is attractive to buyers and it is the single most likely place for a consent problem, so treat it as excluded until somebody has checked your recording disclosure against the states you operate in. There is a platform limit here too: ServiceTitan's telephony terms say that summaries and transcripts generated through its telephony services "shall not be available for download, export or similar data tasks", so on that platform the question may be moot before consent is even reached.
Then the categories that are standard across this lane:
What remains is the part worth licensing: the diagnosis, the options, the pricing logic, the resolution, the callback, the technician's reasoning, your own SOPs and training material. That is the dataset to describe, and the discipline of pulling it out first is what lets you answer a buyer's scoping questions in one call instead of four.
Check the term before you check the number, because the term is what you cannot undo. Polyshares describes its published licenses as perpetual and exclusive, and a perpetual grant does not expire, so a single payment is the whole of what that archive will ever earn.
Permanence here runs past the contract. A model that has trained on your records cannot practically unlearn them, so whatever the contract says about deletion, the training already done is done. That is the reason to price the term rather than treat it as boilerplate, and exclusivity can be limited by time, by field of use or by scope. Our page on exclusive versus non-exclusive licenses sets out how to separate the two, and our page on what happens to your data afterwards covers the permanence question in full.
Then check four things in order. Your software agreement, for the AI terms described above and for your right to a bulk export under your own account. Your call recording disclosure, against every state you operate in. Your commercial client contracts, for confidentiality terms that reach your own files. And the buyer's sublicensing clause, because a non-exclusive license that lets the buyer pass the dataset on freely can reach more labs than an exclusive one would.
Get the export in hand before you agree to deliver anything. A live subscription is what gives you access to your own archive, and an account wound down to save a monthly fee is a harder place to license from. If your work is project-based rather than recurring, the ownership questions are different again and we cover them in can a construction company sell its data to AI companies.
The only published figure from this part of the market is the $450,000 roofing license on Polyshares' track record page, read on November 10, 2026, one of four selected licenses each described as a perpetual exclusive license to the operational record. Per the buyers' own pages read the same day, micro1 shows $100K to $500K for qualified partnerships, $500K to $3M for large-scale or multi-team data and $3M+ for highly unique data, Scale AI shows $10K to $1M+ as illustrative, and Handshake leads with $100K to $4M+. Those are ceilings rather than quotes, and no buyer publishes a rate for a trade.
You describe what systems you run and how far back the records go, the buyer scopes a dataset and makes an offer, the records are de-identified, and payment follows delivery. Polyshares publishes an offer within one to two weeks and payment net 30 to 60. Mercor says compensation is based on data volume, the tools you connect and the depth of your data. For a trade company the preparation that matters is pulling out customer identity and any call recordings made without all-party consent before the scoping call rather than after it.
Neither does. Housecall Pro's terms, last modified 07/01/2026, say that "Individual customer data entered or imported by Pros shall remain the property of the Pro". ServiceTitan's terms, last updated September 16, 2026, say "Customer owns and retains all right, title, and interest in and to all Customer Data". What differs is AI training rights: Housecall Pro's terms record consent to use data for the training of its AI models and say training data may be retained as part of its model's knowledge base, while ServiceTitan says it trains models solely to provide the Service to Customer and will not use Customer Data to train AI models for the benefit of any third party. Read your own agreement before you promise a buyer exclusivity.
Only if they were made with the consent every state you operate in requires, and that is worth checking before the archive is offered rather than after. Several states require all parties to consent to a recording. Call archives are attractive to buyers because spoken diagnosis and objection handling are scarce in written records, which is exactly why the consent question needs answering first. There is also a platform question: ServiceTitan's telephony terms say summaries and transcripts generated through its telephony services "shall not be available for download, export or similar data tasks". Where consent is sound and the records can be exported, they still need the customer's identity removed like everything else.
You can do this yourself, and the contract questions on this page are worth a lawyer's hour on your actual agreements. The commercial decision, which is what your archive is worth and what terms to accept, is a different question. This page is published by BLACKSIG, which reviews these offers with owners and takes companies to the buyer whose terms fit, so read it the way you would read any interested party's account of the work it is paid for. We are independent of every buyer named here and we are not a partner, agent or representative of any of them. The buyer pays us a referral fee when a license closes and there is no fee to the company. We are not lawyers and this is not legal advice.
We look at what your company actually holds, including what your software already claims, then take you to the buyer that fits. The buyer pays our fee, so it costs your company nothing.