Compliance first: We never take HIPAA/PHI data or personal customer data. Everything is de-identified before it leaves your control.
Home / Guides / Pillar guide · Selling data
Pillar guide · Selling data

How to Sell My Business Data: A Step-by-Step Guide for Owners

How to Sell My Business Data: A Step-by-Step Guide for Owners

How to sell my business data, step by step: find the operational records AI builders pay for, send a header sample, compare competing bids, de-identify everything and license it without giving up your company or your customers' privacy.

Most owners never think of their inbox, help desk or job board as an asset. Yet the years of routine work sitting in those systems are exactly what machine learning teams struggle to find: real examples of how a business takes a request, routes it, argues about it and closes it out. This guide walks through the whole sale, from the first inventory to the day the money lands.

Short answer: yes, you can sell your business data

You can sell, or more precisely license, the operational data your company has generated, as long as personal and regulated information is stripped out first and you have the right to share what remains. What gets sold is not your customer list. It is the pattern of your operations: the shape of tickets, the flow of deals, the rhythm of invoices, the back-and-forth of project tasks.

AI and ML builders want that pattern because public internet text is running thin. Researchers at Epoch AI estimate the stock of human-generated public text could be fully used by frontier model training somewhere between 2026 and 2032. Private, operational records are one of the few large pools of real-world data left.

In practice, the sale has four moving parts:

  1. You describe what you have (tools, years, seats, volume).
  2. You share a header sample: column names only, no records.
  3. Buyers bid and you see the competing offers.
  4. The data is de-identified, licensed and delivered, and you get paid.

The rest of this guide unpacks each step and the decisions behind it.

What counts as sellable business data

Sellable data is the operational data produced by running the company, not the private details of the people you serve. Think in categories rather than files.

  • Email and team chat: how work is requested, escalated and resolved. See email and team chat.
  • CRM: stages, timings and outcomes of deals in tools like Salesforce or HubSpot.
  • Support and tickets: problem types, routing and resolution paths, for example in Zendesk.
  • Projects and tasks: planning, slippage and hand-offs in tools like Jira or Asana.
  • Accounting: transaction categories, cycles and seasonality from QuickBooks and similar ledgers.
  • Phones: call volume, timing, routing and outcomes.
  • Industry software: vertical systems such as ServiceTitan or Procore, often the most specialised and valuable records of all.

What is never part of the sale

We never take HIPAA/PHI data or personal customer data. Everything is de-identified before it leaves your control.

For healthcare, dental and veterinary practices that means the admin side only: scheduling patterns and billing operations, never patient records.

Why AI builders pay for operational records

Large models learned to write from the public web. They have not learned how a plumbing dispatcher juggles a Monday morning, how a law firm moves a matter from intake to invoice, or how a support team decides what to escalate. That knowledge lives inside companies.

Builders use operational data to:

  • train agents that can carry out multi-step business tasks;
  • model workflows so automation tools know what "normal" looks like;
  • forecast volume, delays and cash cycles;
  • evaluate whether an AI system behaves sensibly in a realistic setting.

Public reporting shows the appetite is real. Google reportedly agreed to pay Reddit roughly $60 million a year for access to its content, and News Corp's agreement with OpenAI was estimated by The Wall Street Journal at about $250 million over five years. Those are giant publishers, not small firms, but they show that data with a clear licence has become a line item for AI labs. Founder Jeff Cline puts it bluntly: "Every industry is a geek away from being uberized." The firms whose records teach those systems can be paid for it.

Step 1: Take inventory of what you have

Before anyone can price your data, you need a rough map of it. You do not need exact counts; ranges are fine.

For each system, note:

  • Tool name (for example Microsoft 365, Slack, Clio, Toast).
  • Years of history: when the oldest records start.
  • Seats: how many people have used it.
  • Volume: emails, messages, tickets, tasks, transactions or call hours.

Each tool page tells you exactly where an admin can find those numbers. The FTC's data security guidance for businesses starts with the same advice, "take stock" of what you hold and where, and it pays off twice here: you learn what is valuable and you spot sensitive fields that must never leave.

Once you have the list, enter it at request a bid. It takes a few minutes.

Step 2: Send a header sample, not your data

A header sample is the list of column or field names in an export, with no rows underneath. It tells a buyer what kind of information exists without revealing any of it.

For a help desk, that might look like: ticket id, created at, channel, priority, status, group, first response time, resolved at, tag list. For an accounting ledger: transaction date, account, class, amount, memo category, vendor type.

Why it matters:

  • Buyers can judge structure and richness immediately.
  • You expose zero records, zero names and zero contents.
  • It flags fields that need masking before any deal moves forward.

Our header sample walkthrough shows how to pull one from the common tools.

Step 3: Collect bids and compare offers

Once buyers can see what you have, they price it. According to founder Jeff Cline, bids typically come back within 72 hours when the data is described correctly and a header sample is attached, and owners usually see offers from multiple buyers so they can pick the highest bid. Results vary; not a guarantee.

When comparing offers, look past the headline number:

  • Scope: which systems, which years and which fields are included.
  • Exclusivity: a non-exclusive licence lets you license again; an exclusive one usually pays more but closes the door.
  • Payment terms: upfront, milestone or staged on delivery.
  • Usage rights: training only, evaluation, or broader use.
  • Deletion and audit: what happens to the data at the end of the term.

Our licensing agreement guide covers these terms in more depth, and your attorney should review anything you sign.

Step 4: De-identify, license, deliver and get paid

The final step turns an accepted bid into a delivered dataset. De-identification happens first, inside your environment, so raw records never travel.

How de-identification works

Direct identifiers such as names, phone numbers, email addresses and account numbers are removed or replaced with consistent tokens. Indirect identifiers, like a rare job title combined with a small town, are generalised or dropped. NIST's guidance on de-identification (SP 800-188) stresses that simple masking is not always enough and recommends testing for re-identification risk, which is the standard a careful process should aim at. For any health-adjacent admin data, HHS describes two recognised approaches under HIPAA, Safe Harbor and Expert Determination; in our process PHI is excluded outright.

Licensing and delivery

The data licensing agreement spells out what the buyer can do. Delivery typically runs through a controlled data room. In Jeff Cline's experience, deals typically close in about 90 days from first submission. Results vary; not a guarantee.

What makes one company's data worth more than another's

Value comes from a handful of factors that buyers weigh together:

  • History: five or ten years of consistent records beat a few months.
  • Volume and seats: more activity across more people means richer patterns.
  • Structure: clean fields, timestamps and statuses are easier to train on.
  • Uniqueness: a niche industry or unusual workflow that few others can supply.
  • Industry demand: some verticals, such as IT services, legal and call centres, are asked for more often.

According to founder Jeff Cline, some small companies have received over $100,000 for their data. Results vary; not a guarantee. For a directional figure, try the valuation calculator, and read how to value your data for the reasoning behind it.

Why timing matters

The window for this kind of sale will not stay open forever. Once models have learned the patterns of a given industry, the next dataset of the same kind teaches them less, and prices fall. Unique data is king today because it is still scarce; the companies that hold the keys can profit while it stays that way.

If you are curious, the cheapest move is also the fastest: list what you have and get a free bid. You can always say no.

How to sell your business data

  1. Inventory your systems. List each tool you use, how many years of history it holds, how many seats use it and its rough volume.
  2. Submit your inventory. Enter those details on the request-a-bid form so buyers know what exists.
  3. Send a header sample. Export the column or field names only, with no records, so buyers can judge structure.
  4. Compare competing bids. Review offers on scope, exclusivity, payment terms, usage rights and deletion, with your attorney.
  5. De-identify inside your environment. Remove or tokenise names, contact details and other identifiers before anything leaves your control; exclude PHI entirely.
  6. License, deliver and get paid. Sign the licence, deliver through a controlled data room and collect payment on the agreed terms.
Our compliance promise: We never take HIPAA/PHI data or personal customer data. Everything is de-identified before it leaves your control.

Frequently asked questions

Can a small business really sell its data?

Yes. Small businesses can license de-identified operational data such as tickets, deal stages, tasks and transaction patterns. Size matters less than history, structure and uniqueness; a ten-person firm with years of clean records in a niche industry can be very attractive to buyers.

Do I have to hand over customer names or personal information?

No. We never take HIPAA/PHI data or personal customer data. Everything is de-identified before it leaves your control. Buyers want the operational patterns, not who your customers are.

What is a header sample and why do buyers ask for it?

A header sample is the list of column names from an export, with no records underneath. It lets buyers judge what your data contains and how well it is structured without seeing any actual information, which is why it speeds up pricing.

How long does it take to sell business data?

According to founder Jeff Cline, bids typically return within 72 hours when a correct description and header sample are provided, and deals typically close in about 90 days. Results vary; not a guarantee.

Am I selling the data or licensing it?

Usually you are licensing it. A licence grants a buyer defined rights to use a de-identified copy for set purposes, such as model training, while you keep your original records and often the right to license again if the deal is non-exclusive.

Is it legal to sell my company's operational data?

Often yes, once personal and regulated data is removed, but it depends on your contracts, privacy promises, vendor terms and jurisdiction. Review those with your attorney before signing anything.

What if I only use one or two tools?

That is fine. A single system with long, consistent history, such as an accounting ledger or a field-service platform, can be sellable on its own. List what you have and let buyers price it.

Sources

  1. Epoch AI: Will we run out of data? Limits of LLM scaling based on human-generated data
  2. Engadget: Reddit is licensing its content to Google to help train its AI models (Feb 2024)
  3. Engadget: OpenAI will reportedly pay $250 million to put News Corp's journalism in ChatGPT (May 2024)
  4. NIST SP 800-188: De-Identifying Government Datasets: Techniques and Governance
  5. HHS: Guidance Regarding Methods for De-identification of PHI
  6. FTC: AI (and other) Companies: Quietly Changing Your Terms of Service Could Be Unfair or Deceptive
  7. FTC: Protecting Personal Information: A Guide for Business

Updated November 19, 2026.

OpenDataPolicies.org connects businesses with buyers of de-identified operational data. We are not a law firm, broker-dealer or investment adviser, and nothing on this site is legal, tax or financial advice. Statements about amounts, timing and outcomes are Jeff Cline's statements based on past transactions. Results vary; nothing here is a guarantee of any bid, price, buyer or closing. Every deal depends on the data, its quality, its rights and buyer demand at the time.