How to Sell Company Data to AI Companies

Short answer: Most companies license a de-identified copy of operating records (email, chat, files, CRM, tickets), they don't sell the originals. Size, years of history and the systems you use decide which buyers will talk to you. Start with published eligibility rules, inventory what you own, keep regulated and client-confidential data out, and only share samples under a written scope.

On this page

Will they see my confidential data?
Who qualifies
What buyers pay for
Step 1. Inventory your systems
Step 2. Check ownership and rights
Step 3. Pick the right kind of buyer
Step 4. Scope before any export
Step 5. Sample, offer and license terms
Step 6. Get paid
Where to start
Frequently asked questions
Sources

Will they see my confidential data?

This is the question most owners ask first. Programs say identifying details are removed before anything reaches an AI lab, and that you choose what's included. In their own words:

Polyshares says names, emails, phone numbers, account numbers and addresses are removed or replaced by “a specialist PII partner before Polyshares or any buyer sees the record,” and that once a license is signed, “we do not disclose which labs receive which record.”
Lunchbreak says names, emails, and client and sponsor identifiers “are removed automatically before any person reviews a file,” a specialist then checks for anything a pattern can miss, and “you approve the anonymized dataset before it moves.” You can “exclude any project, client or record.”
Mercor says “The extraction and anonymization process is fully automated. No AI lab buyer or third-party vendor ever sees your raw data,” and “No data leaves your systems until you approve it.”
micro1 says “Personally identifiable information (PII) is removed when relevant,” the data scope and anonymization requirements are agreed before participation, and companies “are given the opportunity to review representative samples before prepared data is used.”

Per Mode's site, Mode exports and de-identifies the agreed copy after you agree the systems and date ranges, and no passwords are handed over. None of this is a guarantee. Read each program's contract and data-handling terms before anything leaves your systems. Client-confidential or regulated records (legal, medical, financial) may need to stay out entirely. Our risk checklist covers what to leave out.

Who qualifies

Published size and history rules differ by program. A short map:

Program	Published minimum	Where
Mode	10+ legal; 15+ accounting, medical, finance; 50+ other; 3+ years; 5+ systems	US, CA, UK, AU
micro1	30+ employees; mature operations	US prioritized
Lunchbreak	20+ staff at some point; 2+ years	US, Canada, UK, EU
Mercor	By industry (for example law 5–300, accounting 5–200, software 15–500)	Not stated
Polyshares	“Any company with real operating history” (its own wording)	Not stated
Scale AI	“40+ employees is our preferred starting point” (case by case)	Not stated

Sources: each program's own pages, also summarized on our eligibility guide and the home-page who-qualifies table. Mode's FAQ lists staffing, insurance, and title and escrow as not a fit (Mode affiliate FAQ).

Closed, acquired or winding-down companies have their own path: see buyers for closed companies and our guide on closed startups.

What buyers pay for

Buyers want the record of how work moves through a business: email and calendar, chat, files, CRM, tickets, calls, billing and work tools. Mode's published rules ask for at least 5 systems (Mode FAQ). micro1 lists SOPs, knowledge bases, CRM metadata, project histories and QA processes (micro1).

Published ranges (not offers): micro1 publishes tiers of $100K-500K, $500K-2M and $2M+ (micro1); Scale AI cites $10K–$1M+ illustrative value per data partnership (Scale AI); Handshake AI cites $100K to $4M+ (Handshake AI). Mode shows its own estimate on its site (See Mode's own estimate). Mercor and Lunchbreak do not publish a dollar figure on their pages; Polyshares' pricing is not quoted here. Final offers follow each buyer's review. Full list with sources: how much is my company's data worth and the buyer directory.

Step 1. Inventory your systems

List email, chat, files, CRM, calls, billing and work tools, with date ranges. Note who owns each system and whether you can export it. Mode's published rules ask for 3+ years and 5+ systems (Mode affiliate FAQ). A deeper checklist is in how to prepare your records and what systems buyers want.

Step 2. Check ownership and rights

Confirm you own the records, or that your contracts let you license a copy. Customer, vendor and platform terms matter; so do privacy notices and employee policies. Frankfurt Kurnit's checklist starts with ownership and contracts (Frankfurt Kurnit). Keep health records, payment card data and children's data out of scope entirely: each has its own legal rules, and a buyer's de-identification doesn't change that.

This is general information, not legal advice. Talk to your own attorney or compliance advisor before you share or license any records.

Step 3. Pick the right kind of buyer

Match your size, industry and status to published rules, then browse by situation:

Every buyer we track (sourced directory, CSV and JSON).
Professional services (law, accounting, consulting and similar).
Code and software.
Engineering and CAD.
Under 30 employees.
Outside the US.
Closed or winding-down companies.

Program-by-program comparison: AI data programs compared and best programs roundup. Rankings follow published rules, not referral fees.

Step 4. Scope before any export

Per Mode's site, you agree the value, terms, systems and date ranges before any export; your team signs in to the approved systems and no passwords are handed over. Lunchbreak describes “a written offer after a confidential sample review” (Lunchbreak). Mercor says no data leaves your systems until you approve it (Mercor). Put the systems, date ranges and exclusions in writing before a sample leaves.

Step 5. Sample, offer and license terms

Expect a sample or scoping review, then a written offer and a license agreement. The clauses that matter most after price are scope, exclusivity, who does de-identification, deletion, payment timing and publicity. Buyers differ: Labelbox Prism says licenses are “exclusive by default”; Scrimdata says it is “non-exclusive by default”; per Mode's site, the agreement defines its scope and any exclusivity. Read terms to expect, exclusivity explained and how deals work, step by step before you sign.

Step 6. Get paid

Payment timing is set in each license. Ask when you're paid (signing, delivery or acceptance), whether any of it can be clawed back, and what happens if part of the data is rejected. Published ranges are not offers. Companies never pay DataWorth AI anything; if we introduce you to a program we may earn a referral fee from that program. See our disclosure.

Where to start

Two free paths, no data or logins needed:

Take the 2-minute quiz to see which programs' published rules you meet and each one's published range or link to its own estimate.
Browse the buyer directory and open the buyer pages that fit your size and industry.

Questions: hello@dataworthai.com.

What could your company's records be worth?

Answer 6 questions. See the published range for a company like yours and which program fits. Takes about a minute.

Start the free check

Frequently asked questions

Is this a sale or a license?

Programs describe licensing a copy. Per Mode's site, companies keep their original records and systems. Read the agreement's exact words.

Do I need 50 employees?

Only for Mode in industries outside legal (10+), accounting, medical and finance (15+). micro1 asks for 30+. Lunchbreak counts 20+ staff at some point. Polyshares describes any company with real operating history. Check each program's page.

Can a solo founder or tiny team sell?

Most staffed-company programs set headcount floors. Some buyers focus on code, closed companies or operating history without a high headcount; start at /buyers/under-30-employees/ and /buyers/closed-companies/.

Will anyone know we licensed data?

Ask the buyer whether it will name your company publicly or to its AI customer, and put a no-publicity clause in the agreement if you need the deal kept quiet.

Does this cost us anything?

The check and directory are free. Companies never pay DataWorth AI anything. Programs pay any referral fee, not you.

Sources

Referral-fee disclosure: we may be paid by a program if your company signs with it after using this site. General information only, not legal advice.

Compare programs

Best AI data licensing programs for companies (2026)

Side by side from each program's own pages. Rankings follow published rules, not referral fees.

Buyer directory

See every AI data buyer we track, from their own pages, with sources and dates.

All all industries guides · Every guide

How to sell company data to AI companies