Buyer profile
Scale AI Data Partnerships review: who qualifies and how it pays
Scale AI's Data Partnerships program, which prefers companies with 40+ employees, turns operating companies' de-identified records into training data for AI companies and, Scale says, is designed for recurring revenue.[1] as of
Scale AI
Licenses directly
- How they pay
- Partnerships designed for recurring revenue
- Published range
- $10K–$1M+ per partnership (illustrative)
- Who they buy from
- Operating companies; Prefers 40+ employees
Last checked
| What they buy | Structured, de-identified records from systems a company already runs, including data warehouses, Slack, Teams, Gmail, Outlook, Drive, SharePoint, Jira, Notion, Salesforce, HubSpot, Workday, ServiceNow, SAP and Zendesk.[1] as of |
|---|---|
| Who they buy from | Scale says 40+ employees is its preferred starting point, though it assesses every partnership case by case.[1] as of It also looks for years of records in established systems, real day-to-day workflows rather than demos or synthetic data, specialized domain depth, and an internal owner who can approve packages.[1] as of Its page lists 22 industries of interest, including Field Services (HVAC, Plumbing, Electrical), Property Management, Equipment Rental, Freight Brokerage & 3PL, Retail & Consumer, Manufacturing & Robotics, Law Firms, Legal & Compliance, Business & Professional Services and Hotel & Hospitality.[1] as of |
| Payment model | Scale says it designs partnerships so there's opportunity for recurring revenue, and describes the program as "partnerships, not transactions."[1] as of Separately, Scale pays referrers up to $50,000 for introducing a company, only if Scale signs a partnership agreement with it; one referral is paid per deal.[1] as of |
| Published ranges | Scale shows "$10K–$1M+ illustrative value per data partnership, scaling with cadence."[1] as of |
| Process | Scale describes an intro and scoping call, a limited first scope, anonymization by Scale inside the partner's environment, and the partner's sign-off on every data package.[1] as of |
| Data handling | Scale says it anonymizes everything inside the partner's environment before extraction, stripping personal information so records can't be traced to individuals or accounts.[1] as of Scale says identities, raw systems, credentials and anything outside the signed agreement stay with the company; it cites per-package approval, audit trails, encryption, and on-premises and VPC options.[1] as of |
| How to apply | Companies use the 'Book an intro' or 'Start a partnership' form on the program page; a separate 'Refer a team' form handles referrals.[1] as of |
| Not publicly stated |
|
Provider's own site: scale.com
Is Scale AI legit?
| Founded | Founded in 2016 by Alexandr Wang and Lucy Guo, according to Wikipedia.[2] as of |
|---|---|
| Leadership | Scale's About page names Francis deSouza as CEO; Wikipedia dates his appointment to July 2026. Co-founder Alexandr Wang joined Meta in June 2025 and, Scale said, stayed on its board.[3, 2, 4] as of |
| Headquarters | San Francisco, per Scale's About page.[3] as of |
| Backing | Scale said in June 2025 that Meta had invested at a valuation above $29 billion and would hold a minority of its equity.[4] as of Wikipedia describes Meta's holding as a 49% non-voting stake.[2] as of |
| Public track record | Wikipedia says Scale's commercial customers have included Microsoft, Meta, General Motors and OpenAI, and that it works with the U.S. Department of Defense.[2] as of Wikipedia reports that Scale launched Scale Labs, an expanded research division built on its safety and evaluation lab, in March 2026.[2] as of |
Scale AI’s Data Partnerships program treats 40 or more employees as its preferred starting point and presents pay as an ongoing arrangement rather than a one-time sale, with an illustrative $10K–$1M+ per partnership that it says scales with cadence, as of October 2026. Scale, founded in 2016, supplies training data and evaluation work to AI developers; Data Partnerships is its offer to businesses that want to license their own operating records.
What does Scale look for in a data partner?
Scale says 40+ employees is “our preferred starting point — though we assess every partnership on a case-by-case basis.” It lists four other criteria: years of records in established systems; real day-to-day workflows, “not demos or synthetic data”; specialized domain depth; and a clear internal owner who can approve data packages (program page). The industries it names run from software, telecom and marketing agencies to insurance brokerage, field services, staffing and hospitality.
The last criterion follows from how Scale’s program runs. The partner signs off on every data package, so someone inside the company carries that approval for as long as the partnership lasts. The buyers hub lists each program’s stated size threshold side by side.
What does “$10K–$1M+ illustrative value, scaling with cadence” mean?
Scale’s program page gives the figure per data partnership, and “illustrative” means it is not a quote. The page does not say whether the figure is per package, per year or for the life of the partnership, and it does not define cadence (checked October 2026).
Scale pairs the figure with recurring revenue. Its FAQ answers “Is this revenue recurring?” with “Yes, Scale designs its partnerships so there’s opportunity for recurring revenue,” and the page says “We are looking for partnerships, not transactions.” Scale publishes no split, payment timing or payment trigger, and it does not describe the recurring payments as a revenue share. Payment per delivered package and a share of Scale’s own sales put timing and resale risk in different places; upfront payment versus revenue share explains the difference.
How does Scale anonymize data inside my own environment?
Scale says it “anonymizes everything inside your environment, before any data is extracted,” stripping personal information so records cannot be traced to individuals or accounts. It says employee and customer identities, raw systems, credentials, source access and anything outside the signed agreement never leave the company, and it cites encryption in transit and at rest, per-package approval gates, a full audit trail, and on-premises and VPC options (program page). Partnerships “can begin with a very limited scope,” so a first package can cover one workflow rather than a whole workspace.
Other programs use different designs: Mercor says it connects to a company’s tool accounts and extracts automatically, and Replay says raw data comes to Replay for de-identification. Scale’s page does not say what it installs or needs access to in order to anonymize inside a company’s environment (checked October 2026), so that, and the IT time it takes, belong on the scoping call.
Which AI companies get the data, and will they compete with my business?
Scale’s FAQ says AI companies want the data to improve models for “certain corporate personas such as marketers or financiers,” in real rather than synthetic environments. Asked whether those companies will compete with the partner, it answers no: they model roles within the organization, not the business itself (program page). The page does not name the AI companies that receive partnership data (checked October 2026).
Meta took a minority stake in Scale in June 2025, when it invested at a valuation above $29 billion; Wikipedia describes the stake as 49% and non-voting. Scale said at the time that it remains independent and committed to “safeguarding customer data.” Owners who want to limit which companies receive their data can ask for named recipients or permitted-use limits in the agreement.
Who earns Scale’s referral bonus?
Scale offers anyone who introduces a company “up to $50,000 bonus,” paid only if Scale signs a partnership agreement with that company, with one referral paid per deal (program page). An advisor, vendor or contact who suggests the program may be eligible, so ask whether the person recommending it stands to be paid. Put the same question to any intermediary, including Data Licensing Report, which explains its own referral fees on how we make money.
What should an owner ask Scale on the scoping call?
Scale’s page leaves most commercial terms open (checked October 2026), so start with the unit and the payment: whether the illustrative range is per package, per year or for the whole partnership, what moves a company toward either end, when each payment is made and what triggers it, and whether recurring revenue means a fee per delivery or a share of Scale’s sales.
On rights, ask whether the deal is a license or a transfer, whether it is exclusive, how long recipients can use the data, which AI companies receive packages, and whether you can restrict buyers or uses. Finally, find out what Scale runs inside your environment, who maintains it, and how much staff time per-package sign-off takes.
Sources
- Scale AI program page, Scale AI. Accessed .
- Scale AI – Wikipedia, en.wikipedia.org. Accessed .
- About Scale AI, Scale AI. Accessed .
- Scale AI Announces Next Phase of Company's Evolution, Scale AI. Published . Accessed .
Offer benchmark
Have an offer from Scale AI? See how it compares
- No offer document needed
- "Unknown" is a valid answer everywhere
- Individual offers are never published