Sell Construction Data
Estimates, RFIs, change orders, schedules and daily logs are the operational records that AI training-data buyers now license from established firms. This manual lists each buyer's posted pay, entry rules and data handling, plus the contract checks a GC or sub should run before any talk of price.
What counts#
Buyers want records of real work: questions asked and answered, scope changing, plans slipping and recovering. A closed project file holds all of it, usually across Procore, Buildertrend, Autodesk Construction Cloud, an accounting system and shared drives.
| Record | What it shows | Closest listed category |
|---|---|---|
| RFIs | A field question, the design answer, who answered and how long it took | Scale AI: email threads; documents with metadata |
| Change orders | Scope change, cost and time impact, the approval chain | Nyne: estimates, invoices, AP/AR; micro1: accounting records |
| Schedules | Sequence, durations, revisions and the reasons for them | Nyne: jobs and calendars; Scale AI: calendar events |
| Daily logs | Crew counts, weather, work in place, delays, incidents | micro1: operations records |
| Estimates and bids | Quantities, unit costs, scope assumptions, win or loss | Nyne: estimates |
| Submittals | Review cycles, rejection reasons, resubmissions | Scale AI: documents with metadata |
| SOPs and QA checklists | How the firm runs work, written down | micro1: SOPs, process documentation |
No buyer page read on 7 November 2026 names construction records. The right-hand column is the nearest category each buyer does list, in its own words.
Who is buying#
Four buyers post terms a contractor's records could fit. A fifth program, Mercor's, is left out: its page would not load at our 7 Nov 2026 check, and we print nothing secondhand.
| Buyer | Type | Listed records a contractor holds |
|---|---|---|
| micro1 Enterprise Data Partnerships | Training-data company | Lists operations records, SOPs, process documentation, accounting and legal records, internal knowledge bases |
| Nyne Company data partnerships | Training-data company | Lists accounting data (estimates, invoices, AP/AR), jobs and calendars, Asana or Jira projects, service histories |
| Labelbox Prism | Licensing marketplace | Lists nothing construction-specific; nearest is cross-system interaction data |
| Scale AI Data Partnership | Training-data company | Lists files with metadata, warehouse tables, email, calendars, ERP, Jira and Confluence records |
Paid-referral notice: a listed buyer may pay this site if one of our introductions ends in a signed license. Rows are sorted by published pay floor; fees play no part in order or wording.
Published pay#
| Buyer | Published figure | Basis |
|---|---|---|
| micro1 | Headline $100K–3M+; tiers $100K–500K, $500K–3M, $3M+ | Per company |
| Nyne | $100K+ is the common size, per Nyne; varies with data type and volume | Per company |
| Labelbox | From about $50K (small, targeted sets) to $1M+ (large enterprise deals); fee at signing, then revenue share | Per license |
| Scale AI | $10K–$1M+, which Scale labels illustrative and not guaranteed; rises with delivery cadence | Per partnership, illustrative |
Paid-referral notice: a listed buyer may pay this site if one of our introductions ends in a signed license. Rows are sorted by published pay floor; fees play no part in order or wording.
Not an offerHeadline ranges taken from buyers' own sites. Read them as no offer: the licenses contractors actually sign can land far below. No buyer posts a price for construction records.
For scale only: per Reuters (17 August 2026), Spirit Airlines' workplace data (email, Teams chat, spreadsheets, calendars) drew a $10 million Google bid and a $7.5 million Mercor bid, with the court yet to approve a sale when the story ran. A whole airline's workspace is far from a typical contractor deal.
Eligibility#
| Buyer | Employees | Years | Where |
|---|---|---|---|
| micro1 | 30+ | Not published | US, UK, Canada strongest; elsewhere case by case |
| Nyne | 20+ | 3+ in business | US only (W-2 staff) |
| Labelbox | Not published | Not published | Not published |
| Scale AI | 40+ | Not published | Not published |
Worked case: a drywall sub in Kansas, 25 on W-2 payroll, trading eight years. It clears Nyne's floor, falls short of micro1's 30, and sits under Scale AI's preferred 40. Count your own full-time staff only; crews from a labor broker are not yours.
Ownership#
Licensing rights follow authorship and contract terms, and on one project they can differ from one record to the next.
| Record | Made by | Rights often sit with |
|---|---|---|
| Estimates, bids, cost data | Your firm | Your firm, subject to bid confidentiality terms |
| Daily logs, internal schedules | Your firm | Your firm, unless the owner contract claims all project records |
| RFIs and responses | You ask; the design team answers | Shared; the answers may be the designer's work |
| Change orders | You price; the owner signs | A contract document; check the owner agreement |
| Drawings and specifications | Architect or engineer | Usually the design professional; leave out |
| Sub's records on your job | The subcontractor | The sub, unless the subcontract assigns them |
Legal
- Sell only records you own or are licensed to sublicense. Material your clients own, and personal data about customers, stays out unless both contract and law permit it.
- Depending on the contract, project documents can belong to the owner or to the GC. Check the prime contract and every subcontract.
- Get independent legal review of any data agreement before signing.
Removing client and site details#
Every buyer that describes its process includes de-identification. Do your own pass first regardless, so you decide what leaves the building.
| Detail | Where it hides | Action |
|---|---|---|
| Customer names | Cover sheets, RFI headers, change orders, email signatures | Replace with a role token: Owner-01 |
| Site addresses | Title blocks, permits, daily log headers, photo metadata | Remove; keep state and building type at most |
| Payment details | Pay applications, lien waivers, invoices | Remove bank, card and account numbers entirely |
| Worker names | Daily logs, sign-in sheets, certified payroll | Replace with trade and role; leave payroll out |
| Photos | Daily logs, punch lists, closeout | Leave out unless faces, plates and signage are removed |
Steps#
- Inventory. List each system, the years it covers and the record types in it.
- Contracts. Pull the records, ownership and confidentiality clauses from your owner contracts and subcontracts.
- Scope. Choose record types and years; list projects to leave out entirely.
- Fit. Compare headcount, years and country with each buyer's published rules.
- De-identification. Agree who removes personal and client details, where, and whether you see samples first.
- Agreement. Your lawyer reviews scope, exclusivity, permitted use and deletion terms.
- Transfer and payment. Data moves only under the signed agreement; payment follows the buyer's terms.
Questions#
Can a construction company sell its data to AI companies?
Provided the firm holds the rights, yes. micro1, Scale AI and Nyne each run a licensing program for operational records from established firms. None names construction, yet their categories (operations records, files with metadata, estimates, project records) span most of a GC or sub’s project file.
Which construction records are most useful?
Records that show work and decisions in sequence: RFIs with their answers, change orders with the approval chain, schedules with revisions, daily logs and estimates. A single project file links all of them. See what counts for the full table.
How much do buyers pay?
As posted on 2026-11-07: micro1 $100K–3M+ a company; Nyne says $100K+ is typical; Scale AI shows $10K–$1M+ and calls it illustrative. Marketing ranges, not offers: real license fees often land far lower.
Is my firm big enough?
Size floors as posted: Nyne asks for 20 W-2 staff and 3 years trading, micro1 for 30 full-timers, and Scale AI prefers 40 but will look at smaller firms. Run your numbers through the tier estimate.
Do I need the owner’s permission?
Sometimes. Many owner contracts and subcontracts say who owns project documents and what stays confidential. Drawings and specifications usually belong to the design team. Read the records and confidentiality clauses, and have your own lawyer review before you sign. See ownership of project records.
Does any data leave my systems before I agree?
Never via this site, which has no access to company data at all. Among buyers, Scale AI is the one that states the cleaning happens on your own systems, ahead of extraction, with your sign-off on each package. The rest leave the location unstated.
Sources and read dates: 6 pages
- micro1.ai/data, as of Nov. 7, 2026
- scale.com/data-partnership, as of Nov. 7, 2026
- nyne.ai/sell-company-data, as of Nov. 7, 2026
- prism.labelbox.com, as of Nov. 7, 2026
- mercor.com/data: unreadable at check, no terms used
- Reuters, 17 Aug 2026, Spirit Airlines data bids (syndicated copy)
Get a quote
ZIP code first, then a reply address, then a handful of optional facts about your project records. Our answer names each buyer whose posted entry rules your firm clears, and nothing moves further without your say-so.
Disclosure: introduction service, not a buyer. A buyer that signs a license after our introduction may owe us a referral fee; your company is never charged.
Keep project files, exports and samples out of this form. We never take custody of company data or open it.
Got it. Here is what follows.
- Your headcount, country, records and ownership answer get compared with every posted buyer rule we track.
- A reply follows by email: buyer names that fit, each with its posted terms, or a plain note that none fits yet.
- Your details stay with us unless you approve one named buyer; only that buyer then hears from us.