AI companies are paying businesses for the records of how their work gets done, such as support tickets, CRM records, procedures and code, to train AI agents. Advertised ranges run from about $50,000 for a small dataset to over $3 million for a large one. micro1, one of the buyers, asks for 30 or more full-time employees and data mainly in English.
No buyer publishes a rate card: each advertises a range, read here from its own pages on 15 November 2026.
Who buys, and on what terms
| Buyer or broker | What it is | Data it asks for | Who can sell | Range it advertises | When it pays |
|---|---|---|---|---|---|
| Labelbox Prism | Buyer | Operational data from tools such as Salesforce, Jira, SAP, ServiceNow and Slack | Companies and contributors; no size limit stated | "starting around $50K" to "$1M+": an upfront fee at signing plus a revenue share | "From first call to first wire in under 21 days" |
| Mercor | Buyer | Operational data, delivered to the buyer as an anonymized dataset | Not stated | No figure; based on "data volume, the tools you connect, and the depth of your data" | On delivery of the dataset, typically 2 to 4 weeks after signing |
| micro1 | Buyer | "data that reflects how work is actually performed within an organization" | "30+ full-time employees", data "primarily in English", US companies first, then other Western markets | "$100K to over $3M" | Review within 30 days of transfer, then payment "within 30 days of invoicing" |
| SimpleClosure AssetHub | Broker for startups that are closing | Source code and workspace data | "Founders working with SimpleClosure" | None published | Not stated |
| Troveo | Buyer | "Permissioned business data across 30+ systems of record" | Not stated | None published | Not stated |
Alphabetical. Each cell from the company's own page, checked 15 November 2026. Who pays us has no effect on which companies are listed or in what order.
What deals have actually paid
| Deal | Reported price | Source |
|---|---|---|
| Spirit Airlines' archive (100 million emails, 500 million Teams messages, 176,000 employee records), bought by Google | $10 million | MIT Technology Review, 13 October 2026 |
| One founder's codebase, sold through SimpleClosure | About $25,000, with about $15,000 more expected for the workspace data | MIT Technology Review, 13 October 2026 |
| Closed startups selling through SimpleClosure, nearly 100 deals in a year | $10,000 to $100,000 per company | Gizmodo, citing Forbes, 17 April 2026 |
SimpleClosure's AssetHub had handled over 150 deals by October 2026, according to MIT Technology Review.
What buyers want
- micro1 asks for data "that reflects how work is actually performed within an organization": procedures, project histories, CRM records, support tickets.
- Surge AI looks for archives that reflect a range of situations, from disagreements to tricky judgment calls, to build practice environments for AI agents (MIT Technology Review, 13 October 2026).
Check this before you sell
A law firm's checklist, from Frankfurt Kurnit Klein & Selz (1 October 2026). It is general information, not advice for your company.
| Check | Why it matters |
|---|---|
| Do you have the right to sell it? | "A company that holds data does not necessarily have the right to sell it." Licensed or vendor data may carry limits. |
| Other companies' secrets | Email and chat hold other firms' confidential information, often under NDA. |
| The tool's own terms | Slack's API terms bar third-party apps from using API data to train a large language model. |
| Your privacy notice | You are bound by the privacy statements in force when the data was collected. |
| Employee messages | Talk about pay and working conditions may be protected under US labor law. |
| Privacy law | Under the CCPA, selling workplace data for value is a sale of personal information unless it meets the de-identification test. The GDPR may also apply. |
micro1, Labelbox and SimpleClosure say they remove or process personal information before the data is used, and Mercor says the buyer gets an anonymized dataset. micro1: "Personal and identifying information is removed or transformed before your data is shared with third parties". Ask each buyer for a sample of what it removes, and read the contract with your own lawyer.
How a sale usually works
- You describe your data in a form or on a call. Labelbox says this step is free.
- The buyer quotes a price. Labelbox says it sends a written quote within 48 hours.
- You sign a license that sets what is removed, who may use the data and for how long.
- You transfer the data and the buyer reviews it. micro1 takes up to 30 days.
- The buyer pays. micro1 pays within 30 days of the invoice; Labelbox adds a revenue share on later use.
Questions
How much is my company's data worth?
Advertised ranges start around $50,000 (Labelbox) and reach over $3 million (micro1). Reported deals for small startups were $10,000 to $100,000. Type, volume, years covered and language move the price; no buyer publishes a rate card.
Can a small company sell its data?
Not to micro1, which asks for 30 or more full-time employees. Labelbox states no size limit.
Can an employee sell work data?
No. Only someone with authority to sign for the company can license it.
Does the company keep its data?
With micro1, yes: "Companies retain ownership of their underlying data." Check the license of any other buyer.
Our company has closed. Can we still sell?
Yes, through some buyers. See Shut down your startup? Your data may still be worth money.
Related
- What AI training experts are paid per hour
- About our data
- Companies selling data to AI labs, the same market from the worker's side
Sources
All checked on 15 November 2026 unless dated. Labelbox Prism (prism.labelbox.com); Mercor, its data licensing page; micro1, data partnerships (micro1.ai/data-partnerships); SimpleClosure, AssetHub; Troveo (troveo.ai); MIT Technology Review, "AI labs are buying data from dead startups to train their models", Matthew King, 13 October 2026; Gizmodo, "Failed companies are selling old Slack chats and email archives to train AI", 17 April 2026; Frankfurt Kurnit Klein & Selz, "Selling Company Data for AI Training: What to Know Before You Say Yes", Daniel Goldberg, 1 October 2026.