Last updated: November 12, 2026
The top companies buying Slack and email data for AI training are micro1, Mercor, Scale AI, Turing, Handshake AI, Polyshares, Endless Data, Labelbox Prism, Troveo and Surge AI. micro1 and Handshake AI pay among the largest published sums, Mercor and Polyshares take the smallest companies, and Endless Data is the main non-exclusive option.
In 2026 a new kind of buyer appeared in the inboxes of ordinary companies. Training-data firms that once paid contractors to write prompts started offering businesses six and seven figures for their old Slack channels, email archives and Jira tickets. Business Insider described it as a wild market for your Slack DMs. Forbes reported around 50 startups in the space within months.
The reason is reinforcement learning gyms: simulated workplaces where AI agents practice real tasks, from debugging an IT system to drafting a brief. Those gyms need real records of how work gets done. These are the companies buying Slack and email data for AI training in 2027, and what each one expects from a seller.
| Buyer | Type | Published payouts | Size floor |
|---|---|---|---|
| micro1 | Direct buyer | $100K to $3M+ typical | 30+ full-time staff |
| Mercor | Direct buyer | Varies by industry listing | From 5 staff in some industries |
| Scale AI | Direct buyer | $10K to $1M+ (illustrative) | 40+ staff preferred |
| Turing | Direct buyer | Up to $1M in one payment | Operating or closed companies |
| Handshake AI | Training-data company | $100K to $4M+ | 20+ staff, 3+ years |
| Polyshares | Direct license buyer | $100K to $2M+ | No headcount minimum |
| Endless Data | Aggregator, non-exclusive | Up to $4M | About 30+ staff |
| Labelbox Prism | Upfront fee plus revenue share | About $50K to $1M+ | Not published |
| Troveo | Revenue share agent | Six figures and up | About 100 employees |
| Surge AI | RL gym builder | Not published | Not published |
micro1 helps companies turn operational data into revenue through structured data partnerships with top AI labs. It buys SOPs, knowledge bases, CRM data, project histories and the decision-making patterns in team communication, and says companies keep ownership of their underlying data. Typical deals run from $100,000 to over $3 million. It generally wants 30 or more full-time staff and mainly English records, and it also runs a program for companies that have stopped trading. Its chief revenue officer has said that if a company will not sell its Slack, CRM or email records, the rest is less valuable.
Mercor partners with enterprises to anonymize and contribute operational data to top AI labs. It publishes listings by industry with what a company earns, and its floors are among the lowest, starting at five staff for law, accounting and investment firms. It connects old tools read-only and takes insolvent companies explicitly.
Scale turns anonymized operational workflows into frontier-grade training data and says it wants partnerships, not transactions. It looks for real day-to-day processes with domain depth that is hard to replicate. Illustrative figures run from $10,000 to over $1 million, and it prefers companies with 40 or more employees.
Turing’s Project Lazarus buys a company’s full working record, including chat, email, code, tickets and support history, for up to $1 million in one payment. Intake happens under NDA within a business day and it confirms a figure after sampling, usually within a week.
Handshake’s AI business went from launch to a reported $1 billion in annualized revenue in about a year, and it now goes directly to companies to license workflow data, code and support transcripts. Payouts have been reported between $100,000 and $4 million or more, for companies with 20+ staff and three or more years of history.
Polyshares buys licenses to operational records as principal: Slack and Teams, Google Drive and SharePoint, email, CRM, support tickets, design files and accounting exports. It has no size floor and takes any company with real operating history. Typical licenses run from $100,000 to $2 million and are exclusive by default. Its license does not cover what a company holds for its own clients.
Endless Data pays a fixed sum agreed before anything is shared and licenses the data to several labs. It is the only major buyer that is non-exclusive by default, so the company can license the same records elsewhere. It strips identifiers before data leaves the company and deletes it at the end of the term. It wants roughly 30 or more staff and several years of records.
Prism acquires operational data from companies and contributors and reports over $100 million distributed to more than 1,000 companies. Sellers receive an upfront fee at signing and a revenue share on downstream use.
Troveo licenses chat, email, CRM, finance and code to several buyers and pays a share on each resale. It has paid over $20 million to content owners across its categories and typically takes companies from about 100 employees.
Surge AI buys data to build reinforcement learning gyms where clients’ agents practice tasks such as debugging complex IT systems. It is less of a public intake program than the names above, but it shows how far the demand reaches.
We listed companies with a public way to license workplace records and checked each one’s own pages for what it buys, size floors, exclusivity and published payouts. Where figures come from reporting rather than the company, we say so. Read more about how we research and rank companies.
Client data held under NDA is usually not yours to license. Employee data raises privacy questions under laws such as CCPA and GDPR. Slack’s API terms also restrict using API data to train large language models without an additional agreement, and free and Pro workspaces can only export public channels. Read our guide on employee privacy and selling workplace data before you apply.
Polyshares and Prism default to exclusive; Endless Data defaults to non-exclusive. The difference decides whether you can sell twice.
If you would rather not apply to ten buyers yourself, a broker such as Vortex Bridge can match you with a vetted buyer, with the data going straight from you to that buyer. Our list of workplace data brokers compares the options.
To train AI agents on how real work gets done, often inside reinforcement learning gyms built from real company records.
A company can license records it owns within privacy law and its contracts, but Slack’s API terms restrict using API data to train large language models without an additional agreement.
Endless Data is non-exclusive by default, so the same data can be licensed elsewhere.
The buyers of Slack and email data range from venture-backed training-data firms to specialist aggregators, and their terms differ more than their headlines suggest. Compare size floors, exclusivity and who does the anonymizing, and get more than one offer. For a sense of what your records are worth, see how much your company’s Slack, email and Jira data is worth to AI labs.
For more guides and rankings on this topic, visit our Sell Company Data To AI hub.