You can license CRM data to AI companies when your contracts allow it and the contact details are removed first, because labs want the pipeline history (stages, activities, timing and won or lost outcomes) and have little use for the names. The hard part is legal: a work email at another company is still personal data, so a CRM export needs a rights review and a thorough scrub before any buyer sees it.
This is general information, not legal advice. Talk to a privacy lawyer before you license anything that came from your CRM.
What labs want from a CRM
A CRM records how a company sells: which leads came in, what the sales team did with each one, how long each stage took and whether the deal closed. Labs training AI to do sales and account work need that sequence, and it rarely appears in public text.
The valuable parts are:
- Opportunity history. Stage changes with timestamps, amounts as ranges and the final outcome.
- Activities in order. Calls logged, emails sent, meetings booked and tasks completed, as events rather than message content.
- Field history. Salesforce field history and similar audit logs show what changed and when, which is the raw material for episodes.
- Outcomes. Won, lost, churned, renewed, upsold. Outcome labels are what let a lab check whether an agent's decision was good.
- Account structure. How accounts, contacts, opportunities and products relate, which is useful for building a working replica of a CRM.
What labs don't need is the contact list itself. Names, emails and phone numbers add privacy risk and almost no training value.
Who owns CRM data
Your CRM vendor doesn't own your records. HubSpot's customer terms, for example, say "You own and retain all rights to the Customer Materials and Customer Data" and give HubSpot permission to use that data only as needed to provide the service.
Owning the data in a contract sense doesn't settle whether you can license it. The people in your CRM have their own privacy rights, and your customer agreements may limit what you can do with information about them. Those are separate questions, and both need an answer.
Business contacts are personal data
Many companies assume B2B contact details fall outside privacy law. In Europe and the UK they don't. The UK Information Commissioner's Office says that holding the names and details of your contacts at other businesses means handling personal data, even when those people act in a business capacity, and that they keep the right to object to direct marketing.
Under the GDPR, licensing that data to a lab is a new purpose. The purpose limitation principle in Article 5(1)(b) and the compatibility test in Article 6(4) apply, and you would need a lawful basis such as legitimate interests, which requires a documented balancing test. The GDPR guide covers this in more depth.
In California, a transfer of personal information to a third party for money counts as a sale under Civil Code section 1798.140. Data that meets the statute's definition of deidentified falls outside it, which is why scrubbing matters so much for CRM data. The CCPA guide explains the three conditions.
Your CRM may hold your customers' data
Some companies run a CRM on behalf of their clients: an agency managing a client's pipeline, or a software product whose customers store their own contacts in it. In that case you are usually a processor for those customers and can't license the data without their written permission. Can I sell data I hold for my customers? walks through how to tell.
Read your customer agreements and data processing agreements for clauses such as "used only to provide the service." A clause like that blocks a sale even after scrubbing. During our rights review we read these contracts with you, and if we can't get a clear answer for a data set, we leave it out.
How a CRM export gets scrubbed
CRM data is full of identifiers, many of them hidden in places a simple find-and-replace misses. Our scrub runs on your machine, and raw data never leaves it through the SDK. For a CRM it does the following:
| CRM content | What happens |
|---|---|
| Contact names, emails, phone numbers | Removed, or replaced with consistent pseudonyms where links must survive |
| Company names and domains | Replaced with pseudonyms such as ORG_7F3A |
| Deal amounts | Rounded into ranges |
| Dates | Generalized to the month |
| Job titles | Broadened where a rare title plus a company could identify someone |
| Email bodies, call notes, descriptions | Dropped unless reviewed |
| Owner and rep names | Replaced with pseudonyms such as PERSON_14 |
Company domain plus job title is a classic way to re-identify a person, and the scrub treats those fields as quasi-identifiers for that reason. After scrubbing, we test whether any record can still be singled out from a combination of fields and remove the ones that can. The result goes in the scrub report you review before anything is offered.
What a scrubbed CRM record looks like
After scrubbing, an opportunity might read: ORG_A21, mid-market, inbound, stage history qualified to proposal to negotiation over 11 weeks, amount $40k-50k, two discovery calls and one demo logged, closed lost to a competitor in Q2. It contains no contact, no email text and no exact figure, and it still teaches a model how a real deal moved.
Prices for CRM data
There is no public price list for CRM exports licensed for AI training, and we won't invent one. The closest public signal is reporting on shut-down startups: Forbes reported in April 2026 that SimpleClosure handled nearly 100 deals in a year selling startups' internal records such as Slack messages, Jira tickets and emails to AI companies, with payouts of $10,000 to $100,000 per company. Those were one-off sales of a closed company's archive. An operating company can sell its history and then keep earning from monthly refreshes.
Our calculator uses a base for CRM data and adjusts it for industry, years of history, team size and how clear your rights are. The estimate drops when rights are unclear, which is honest: data you can't license is worth little to a lab.
The recurring part
A CRM keeps growing as long as your sales team works. After the first license sells the history, the SDK can run monthly, pull only opportunities and activities that changed since the last run, scrub them locally and upload a new batch. Each license with a refresh term pays per accepted batch, with no extra work once the schedule is set. A bigger sales team creates more activity each month, so the batches grow. See recurring data revenue.
Get an estimate for your CRM
The valuation calculator takes two minutes and needs no access to your systems. Pick "CRM (Salesforce, HubSpot)" along with any other systems you'd consider. For how CRM history fits with industry-specific data, see the pages for software and financial services, or start at selling data to AI companies.
Frequently asked questions
Can I sell my contact list to an AI company?
We don't sell contact lists, and labs don't need them. What has value is the anonymous history of deals and activities. Contact names, emails and phone numbers are removed or replaced with pseudonyms before anything is offered.
Are B2B contacts exempt from privacy law?
In the EU and UK, no. The ICO treats business contacts' names and details as personal data. In California, a sale of personal information is regulated under the CCPA, and data falls outside it only if it meets the statute's deidentified conditions.
Do I need Salesforce's or HubSpot's permission?
Generally the vendor isn't the obstacle. HubSpot's terms say the customer owns its data. The questions that matter are your own customer contracts, your privacy notices and whether you hold the data on behalf of clients.
What about email bodies and call notes?
They are dropped by default, because free text carries names and details that pattern matching can miss. They can be included only after a dedicated review, and often the activity log without the text is enough.