Key takeaways
- micro1 asks for 25+ employees, 2+ years of records and data primarily in English.
- Every industry can qualify; patient medical records, legally privileged material and trade secrets are left out of the deal, and the company's other records can still qualify.
- Closed companies qualify with 25+ staff at their peak, but only if they export Slack or Teams, email and documents before subscriptions lapse.
- The data vendor decides whether a company qualifies from its application, an intro call and a data manifest, before any data leaves the company, and pays in full about 30 days after accepting the data.
01
Does my company qualify to sell its data to AI labs and data vendors?
The short answer
A company usually qualifies to sell (license) its data to AI labs and data vendors if it has 25 or more employees, at least two years of work recorded in software, and records that are at least 70% in English. US companies come first, then the UK, Canada and other Western markets. Companies that have shut down can qualify too, if they had 25 or more staff and still have exports of their team chat (Slack or Microsoft Teams), email and documents.
Who is buying company data
An AI model learns a task by studying many examples of that task being done. The best examples of everyday business work, such as a support agent solving a customer's problem by email or a salesperson taking a deal from first call to signed contract, exist only inside companies, in their team chat, email, help-desk tickets and customer records. Two kinds of buyers pay for those records. We explain why in why AI companies want your data.
- AI labs
- Companies that build and train AI models, such as OpenAI, Anthropic and Google DeepMind.
- Data vendors
- Companies that license records from businesses, anonymize them (remove or replace every detail that identifies a person) and turn them into training material for AI labs.
Most business owners deal with a data vendor and never with a lab. The criteria in this guide are the published requirements of micro1, the data vendor we typically refer companies to, plus what micro1's team has said in its own webinar. micro1 has committed $1B to buying enterprise data over 12 months. A buyer spending that much reviews companies of every size and industry, so its requirements are a useful guide to what other data vendors look for.
The six things a data vendor checks
Before a data vendor prices a company's records, it runs six checks to see whether the company fits. Each check is a quick way to answer one question: do the company's records show enough complete jobs, from the first request to the finished result, written in a language the vendor can read, for an AI lab to learn from? A company's industry is not one of the checks.
| Check | micro1's usual bar | Where there's flexibility |
|---|---|---|
| Employees | 25+ | The count that matters is staff whose daily work happens in software. Other data vendors set their own bars, from 20 to 50 employees, and one publishes no minimum. |
| Years of records | 2+ years | Each extra year of records raises the price. micro1's team calls 5+ years ideal. |
| Language | 70%+ English | Some records in other languages are fine. Talks with the vendor happen in English. |
| Location | US first | UK and Canada next; other Western markets case by case. |
| Work in software | Several systems | Work recorded in team chat, email, help-desk tickets, a CRM (the software that tracks customers and deals) and shared documents counts. Paper records don't. |
| Status | Operating or closed | A closed company needs 25+ staff at its peak and saved exports of its chat, email and documents. |
The diagram below sorts the possible answers to each check into three columns. A company whose answers all land in the green column is likely to qualify. One or two answers in the violet column mean the data vendor decides case by case, and an answer in the amber column means the company doesn't qualify yet.
Other data vendors set different bars
Each data vendor sets its own minimum size and history, so a company that falls just short of micro1's requirements may meet another vendor's. The figures below are from each vendor's own page, checked in November 2026. We compare the vendors in more depth in where to sell your data.
| Data vendor | Stated size | Other stated requirements |
|---|---|---|
| micro1 | 25+ employees | Mature operations, records mainly in English, US first then other Western markets (source) |
| Handshake AI | 20+ full-time employees | 3+ years in operation (source) |
| SourceX | 50+ full-time employees at peak | Several years of operating history; takes operating, acquired and wound-down companies; US first, plus Canada and the EU, and other countries when 70%+ of the data is in English (source) |
| HUMXN | No minimum published | "Any business, in any industry, with a few years of records" (source) |
A company that meets one of these bars can apply to that vendor directly. You can apply to micro1, start a SourceX application or ask HUMXN for a valuation. Each application asks about your company, not for any of its data.
Checker
Does your company qualify?
The checker asks six questions and takes about a minute. Nothing you pick here is sent anywhere.
Based on micro1's published criteria, October 2026. The buyer makes the final call after the intro call and data manifest.
02
How many employees do you need to sell company data?
The short answer
micro1 asks for 25 or more employees. The count that matters is desk staff, meaning the people whose work happens in shared software such as team chat, email, help-desk tickets and a CRM.
Headcount is a quick measure of how much work is on record
A data vendor doesn't value company size for its own sake. It uses headcount as a quick estimate of how many recorded jobs a company's systems hold, and how often those records point to each other. Twenty-five people who work together in shared tools for a few years leave thousands of chat threads, tickets, documents and deals that refer to one another, such as a ticket that links to the chat thread where a colleague explained the fix. Five people mostly leave email.
Desk employees count; field staff count less
A company with many people who work away from a computer can still qualify, because its office staff run the business through software. An HVAC company (a business that installs and repairs heating and air conditioning) with 80 technicians and 25 office staff is a good example. Its dispatchers, who decide which technician goes to which job, its estimators, who price each job, and its billing and customer-service staff record every job in software. The technicians often type their own job notes into the same system from a phone or tablet.
The test is whether the work leaves a written trail. A warehouse team that logs each order in an inventory system adds to the company's records; a crew that only talks on the radio does not.
What makes a company near the minimum a strong case
A company close to micro1's 25-employee minimum is judged mostly on what its records contain. The data vendor looks for four strengths, the same ones that raise the price of any company's data.
Many years of history give an AI lab more finished jobs to learn from. Several systems whose records link to each other let the lab follow one job from the first email to the final invoice. Records of how each job ended, such as whether a deal closed or a ticket was solved, show which ways of working succeeded.
An unusual industry or way of working is the fourth strength, because few other datasets show it. A 26-person firm with ten years of tickets in one help desk is a stronger case than a 40-person firm that started using software last year.
03
How many years of records, and how much data, do buyers want?
The short answer
Data vendors want at least two years of business records, and each extra year raises the price; micro1's team describes five or more years as ideal. micro1 publishes no fixed minimum on volume, though its team has described about a terabyte or more as a strong fit.
Why two years is the floor
An AI model learns a type of job by seeing it done many times, and a business needs time before its records show that repetition. Within two years, a support team handles the same kind of request dozens of times, a salesperson takes a customer from first contact to a renewal a year later, and managers change a process after something goes wrong. Two years is roughly the shortest history that shows all three.
More years raise the price
A data vendor pays more for a longer history because each extra year adds more finished jobs, each with a record of how it turned out, for an AI lab to learn from. Our guide to what your data is worth explains how years, systems and rarity move the price, and the data value estimate puts a range on your own company.
Volume matters least
micro1's site states that there's "no fixed minimum number of files, documents, or terabytes." A terabyte is about a thousand gigabytes, the size of a large shared drive with years of documents and attachments.
Four things move the price more than volume does. Records that link across systems, data few other companies have, clear legal rights to the records and exclusivity (a promise not to license the same data to any other buyer) all count for more. A huge pile of loose files teaches a model less than a smaller set of complete jobs, because loose files show pieces of work without showing how they fit together or how the job ended.
04
Which systems count: is your company's work recorded in software?
The short answer
The systems that count are the ones where employees do and discuss their work: team chat (Slack, Microsoft Teams), email, help desks, CRMs, project tools and shared documents. Several connected systems are worth far more than one, and work kept on paper doesn't count.
What each system records, and why buyers want it
Every system a company uses captures a different part of how a job gets done. The table lists the common ones by what they record.
| System | Examples | What it records |
|---|---|---|
| Team chat | Slack, Microsoft Teams | How people discuss problems, ask for help and make decisions |
| Google Workspace, Microsoft 365 | Conversations with customers and suppliers, approvals and follow-ups | |
| Help desk and tickets | Zendesk, Jira, ClickUp, Asana, Monday | A request from first message to resolution, with each step |
| CRM | Salesforce, HubSpot | Each customer's history, from first inquiry to closed deal or finished job |
| Documents | Google Drive, SharePoint, Notion | Procedures, playbooks, training guides, quotes and reports |
| Code | GitHub, GitLab | Software, the changes made to it and the review of each change |
A CRM (customer relationship management system) is the software a sales team uses to track every customer and deal. micro1 sorts the data it buys by business function rather than by app: software and engineering, finance and accounting, customer support, sales and CRM, legal and contracts, and operations and logistics.
Connected systems are worth more than any one alone
Records in one system often point to records in another. A support ticket mentions the Slack thread where the team agreed on a fix, and the CRM shows whether that customer renewed afterwards. By following those links, an AI lab can trace one job from the customer's first complaint to its outcome, which no single system shows. That is why a company with email, tickets, documents and a CRM is worth far more than one with any single system.
How the data leaves your systems matters
Some software companies restrict what their API (the connection other software uses to pull data out) may be used for. Slack amended its API terms in 2025 to prohibit bulk exports and the use of that data in large language models, as law firm Hunton reported. For that reason, a data vendor usually asks the company's workspace owner to download the history with the app's own admin export instead of pulling it through the API. Which exports are allowed is a question for your lawyer, covered in is it legal to sell company data.
05
English and location: which companies data vendors take first
The short answer
micro1 wants records that are primarily in English (70% or more is the working bar) and talks with companies in English. It prioritizes US companies, then other Western markets such as the UK and Canada, and considers other regions case by case.
Why English comes first
English matters because a data vendor reads every record before an AI lab sees it. The vendor anonymizes each record and checks its quality, and both steps cost more in other languages. AI labs also run fewer projects that use records in other languages, so those records are harder to place. A company that works in English with some records in Spanish or French is usually fine.
Why US companies come first
AI labs and data vendors want US companies' records most. micro1's page says it is "prioritizing US companies, followed by other Western markets," and its team has named the UK and Canada as accepted. A company with some employees overseas can still qualify, though micro1 prefers a good share of the team to be in the US, UK or Canada.
Where the data goes
micro1 also limits who receives the data after it buys it. micro1 states that licensed data goes only "to labs and enterprises based in the U.S. and allied countries."
06
Which industries qualify, and which are harder?
The short answer
Every industry can qualify to sell data to AI labs and data vendors; micro1 says "there isn't really domains that we don't accept." What makes a company harder to accept is the kind of data it holds: patient medical records and legally privileged material (confidential communications between a lawyer and a client) are mostly avoided, and trade secrets must be left out.
Why industry barely matters
AI labs are building models meant to do work in any kind of business, so they want examples of work from every kind of business. micro1's own examples range from HVAC contractors to a company that builds cooling systems for the chips that run AI, to ordinary software companies. Marketing agencies, field-service businesses (companies that send technicians to customers' homes and sites), logistics firms, manufacturers and professional-services firms such as accountants all fit.
A niche industry often helps. Records that show unusual work are rare, and AI labs and data vendors pay more for records they can't get from many other companies. That can make a small specialist firm's records worth more than those of a larger firm doing common work.
Three kinds of data that make a company harder
The exceptions are kinds of data, not kinds of company. Each one below can usually be carved out, so the rest of the company's records still qualify.
Harder
Medical records
Patient health records are protected by health-privacy law (in the US, HIPAA, the Health Insurance Portability and Accountability Act), and micro1 mostly avoids them. A clinic or health-tech company may still offer records that never touch a patient file, such as its engineering, sales, finance and operations records, if those are kept cleanly separate.
SourceX lists healthcare among the industries it buys from, limited to administrative records rather than patient files, so other vendors may go further.
Harder
Legally privileged material
Attorney-client privilege protects confidential communications between a lawyer and a client, and sharing them with a third party can compromise it. A law firm's client files are mostly avoided; its client-intake, billing and marketing records, and its practice-management records (the software that tracks cases, deadlines and hours), may still qualify.
Any company's own legal threads with its lawyers are usually excluded too.
Excluded
Trade secrets
A trade secret is information that is valuable because competitors don't have it, such as a formula, a pricing model or source code. micro1 requires trade secrets and other legally protected data to be left out.
You choose the exclusions before any data leaves the company. Leaving out a large share of records can lower the price.
The exclusions are written into the contract. The data vendor then removes or replaces personal details before anyone else sees the records. We explain that step in how company data is anonymized.
07
Can a company that has shut down sell its data?
The short answer
Yes. micro1 runs a Defunct Company Data Partnership for companies that are no longer operating, with the same 25-employee minimum and mostly-English records as for operating companies, every industry accepted, and pay of $100K to $2M or more for approved data packages. A closed company needs 25 or more staff at its peak and exports of its Slack or Teams, email and documents.
Why closed companies' data is in demand
A company's records are just as useful for training AI after the company closes. Years of Slack threads, email and tickets still show how real work got done, which is what AI labs and data vendors pay for. MIT Technology Review reported in October 2026 that 47% of the shutdowns handled by one company that winds down startups now include a sale of code or data.
The largest closed-company archive so far drew eight-figure bids. The same report says Google won the bankruptcy auction for the records of Spirit Airlines, including 100 million emails and 500 million Teams messages, with a $10M bid. micro1 later filed a $12.5M bid, Fortune reported, and as of October 2026 the court had not yet approved a buyer.
Spirit was a large airline, so its archive is a headline case. Most closed companies are far smaller, and they are paid in the same range as operating companies, $100K to $2M or more per dataset.
What a closed company needs to qualify
micro1's Defunct Company Data Partnership sets out its requirements on its referral page. The company must have had 25 or more employees and kept its records in English, it can come from any industry, and it must be based in one of 10+ countries. The countries micro1 has named for closed companies are the US, UK, Canada, India, Australia, Singapore, Ireland, the Netherlands, Germany and France.
- 25 or more employees at the company's peakHeadcount is measured at the company's largest, not at closing.
- Substantial Slack or Teams history still on handTeam chat is the record buyers value most from a closed company, followed by email and documents.
- Records mostly in EnglishThe same 70% working bar as an operating company.
- Someone with authority to sellOnly the company can license its data, never a former employee. After closing, that authority sits with whoever is winding the company down.
How a closed company's deal differs from an operating one
A closed company passes nearly the same checks as an operating one. What changes is who signs the deal, which records must still exist, and who receives the money, as the diagram shows.
What happens to a company's data when the business closes
A company's data is an asset, in the same way as its equipment or its code. In a wind-down (the process of closing a company, selling what it owns and paying what it owes), the data is usually sold for the benefit of creditors, the lenders and suppliers the company owes money to, and investors. The MIT Technology Review report notes that proceeds mostly go to them rather than to employees. Who has authority to sell depends on how the company closes: its board in an ordinary dissolution, a trustee in bankruptcy, or an assignee in an assignment for the benefit of creditors (a state-law process where an independent party sells the assets and pays the creditors).
Other buyers take closed companies too. SourceX, a data vendor that licenses business records directly and sublicenses them to AI developers, says companies that are operating, acquired or wound down can all qualify. It measures size at the company's peak, looks for 50 or more full-time employees, and pays a one-time fee per dataset.
If you're winding down
Export your Slack or Teams workspace, email and shared drives before you cancel any subscription, and don't delete the workspace until you know what it's worth. Applying takes a couple of minutes and shares no data.
08
From application to payment: how qualifying turns into a deal
The short answer
The process with micro1 runs in eight steps: a two-minute application, an intro call, a data manifest (a description of your records that shares no files), a priced offer within 1–2 days, your own legal review, a secure transfer of 1–7 days, the data vendor's acceptance check, and payment in full within about 30 days of acceptance. No data changes hands until you've signed.
What a data manifest is
The data vendor makes its final decision on whether a company qualifies after reading the data manifest. A data manifest is a short description of the data a company could license: which apps it comes from, how many years each covers, roughly how large it is, the file types, and basic company details. It works like an inventory list, describing what's in the archive without opening any of it, so nothing confidential is shared at that stage.
The eight steps, in order
Apply 2 minutes
You fill in a short form with the company's name, size, industry and the tools it uses. The form shares no data and commits you to nothing.
Intro call After review
You talk with the data vendor about whether your company fits, which systems it uses, and your questions, including how anonymization and exclusions work.
Data manifest No files shared
You describe your apps, years of records, size and file types. This is where the vendor confirms whether you qualify.
Priced offer 1–2 days
micro1's team says a priced proposal follows the manifest within one to two days. You can accept, negotiate or walk away.
Legal review Your pace
Your lawyer reviews the data licensing agreement: permitted use, exclusivity, exclusions, how long raw data is kept and when it's deleted. Don't send samples before this step.
Transfer 1–7 days
The data moves through an encrypted API or another approved method, and you can watch its progress.
Acceptance Review period
The data vendor checks that the data is complete and matches the manifest, then accepts it.
Payment About 30 days
micro1 states "We pay in full within 30 days of acceptance." Its FAQ describes the timing as within 30 days of invoicing, so confirm the exact terms in your contract.
The timeline below groups the steps into two halves. In the first half nothing leaves your company. The second half starts with your lawyer's review, and data moves only after you sign.
What a qualifying company is paid
Companies that qualify are paid $100K to $2M or more per dataset, according to the published ranges of micro1 and Polyshares, another data vendor. The largest company archives draw eight-figure bids, such as the $10M and $12.5M bids for Spirit Airlines' records described above. micro1 breaks its own range into three tiers: $100K–$500K, then $500K–$2M for large datasets or ongoing work across several teams, and $2M or more for highly unique data. The data vendor sets the final terms.
The deal is normally a license, which lets the buyer use a copy of the records while the company keeps ownership, rather than a sale, as explained in data licensing vs. selling.
09
What to do if your company doesn't qualify yet
The short answer
A company that is close on one criterion should apply anyway and say so, because data vendors make exceptions for strong or unusual data. A company that is well short (under 20 employees, under two years, or mostly paper) should keep its records in software and in English, check vendors with lower bars, and keep every archive intact.
Close on one point: apply and explain
The application is free and doesn't commit you to anything, and the data vendor makes the final call after the intro call and data manifest. In the application, say where the company falls short and what makes its records strong, such as a decade of history, an unusual industry, records of how each job ended, or several linked systems. micro1's team has said it accepts companies slightly below its bar on specific strengths.
Under micro1's headcount: check vendors with lower bars
Data vendors set different minimum sizes, as the table near the top of this guide shows. Handshake AI states 20 or more full-time employees with three or more years in operation. A company with 20 to 24 people may qualify there before it qualifies with micro1.
Not there yet: build the record now
A company that can't qualify today can still make its future records worth more. Each habit below adds to the connected history data vendors pay for.
- Move work off paper and private phonesJobs logged in a help desk, CRM or shared chat become records; calls and paper forms leave nothing to license.
- Record outcomesMark tickets solved, deals won or lost, and jobs finished, so each record shows how the work ended.
- Keep your historyCheck that your chat and email plans keep old messages; some free plans hide or delete history.
- Separate sensitive recordsKeep patient, legal and HR files in their own systems so they can be excluded cleanly later.
Mistakes that cost companies a deal
The most expensive mistake is deleting the data. Owners who close a Slack workspace, cancel Microsoft 365 or wipe a shared drive during a move or a shutdown lose the one thing data vendors buy.
Two other mistakes come up often. Only the company can license its data, never an employee acting alone, so the application should come from an owner or executive. And a sample of records sent before legal review has already shared real company data, so samples wait until your lawyer has read the contract.
Questions
Frequently asked questions
How many employees does a company need to sell its data to AI companies?
Most data vendors set a minimum of 20 to 50 employees. micro1 asks for 25 or more employees, Handshake AI states 20 or more, and SourceX looks for 50 or more full-time employees at the company's peak; HUMXN publishes no minimum. The count that matters is desk staff, the people whose work is recorded in chat, email, tickets and a CRM.
How long does a company need to have been in business to sell its data?
Data vendors usually want at least two years of records, and micro1's team describes five or more years as ideal. More years raise the price because each year adds more finished jobs and outcomes for an AI model to learn from. Handshake AI states three or more years in operation for its own program.
Can a small business sell its data to AI labs?
A small business can sell data to AI labs and data vendors if it clears the vendor's size bar, which starts around 20 to 25 employees depending on the vendor. A business with fewer staff usually doesn't have enough connected records yet. It can still build value by keeping its work in software, recording outcomes and never deleting its history.
Can a company that has shut down sell its Slack and email data?
Yes. micro1's Defunct Company Data Partnership buys data from companies that are no longer operating, asking for 25 or more employees, English records and a base in one of 10+ countries, and pays $100K to $2M or more for approved data. The company needs exports of its team chat history from Slack or Microsoft Teams, email and documents, so export them before subscriptions lapse.
What happens to a company's data when the business closes?
A company's data is an asset, like its equipment or code, and in a wind-down it is usually sold or licensed for the benefit of creditors and investors. The board, a bankruptcy trustee or an assignee decides. MIT Technology Review reported in October 2026 that 47% of shutdowns handled by one startup wind-down company now include a sale of code or data.
Can companies outside the US sell their data to AI companies?
Yes. micro1 prioritizes US companies, then other Western markets such as the UK and Canada, and considers other regions case by case. For closed companies, micro1 has named the US, UK, Canada, India, Australia, Singapore, Ireland, the Netherlands, Germany and France. Records need to be primarily in English, roughly 70% or more.
Can a healthcare company or law firm sell its data?
Often, yes, with exclusions. micro1 mostly avoids patient medical records and legally privileged material, but a clinic's or law firm's other records, such as engineering, sales, finance, billing and operations, may still qualify if they are kept separate. SourceX lists healthcare among its industries for administrative records, so other data vendors may accept more.
What is a data manifest?
A data manifest is a short description of the data a company could license to a data vendor: which apps it comes from, how many years each covers, roughly how large it is, the file types and basic company details. It works like an inventory list and shares no files. micro1 uses the manifest to confirm a company qualifies and to price an offer, usually within one to two days.
Bottom line
Close enough? Apply and let the buyer decide
The checker shows where your company stands against micro1's published criteria. Applying takes two minutes, shares no data and doesn't commit you to anything.
Not legal, tax or financial advice. Figures come from micro1's published pages and public reporting as of November 2026.