Sell email archives to AI

Why is email valuable for AI training?

AI labs are increasingly buying workplace data such as email, Slack and support tickets to train AI agents – systems meant to carry out real work tasks. Text from the web isn't enough for that. Agents need to see how people reason, negotiate and solve problems over time. In the US, shut-down startups have sold their email and Slack archives to AI companies, as reported by Fast Company, Gizmodo and Forbes, among others.

An email archive with several years of history contains exactly what is in demand: questions, decisions, follow-ups and outcomes. That is why it is often one of the most interesting data sources for anyone who wants to sell an email archive or use it for AI training through a license.

What usually determines the value:

• Length of history – more years give more complete sequences.
• Reasoning – threads where problems are discussed and solved carry more weight than newsletters and notifications.
• Language – business communication in Nordic and other smaller European languages is rare in training data and therefore in demand.

More on how value is assessed on what is your data worth.

Selling email archives – what's included and what's excluded?

Not all email is suitable for a license. A common setup looks like this.

Often relevant:

• shared function mailboxes, for example for sales, purchasing or support
• project and business communication with customers and suppliers
• internal communication about routines, decisions and problem-solving

Normally excluded:

• Personal mailboxes and emails of a private nature
• Legally privileged correspondence, for example with lawyers
• Health information – health data is entirely outside what we handle
• newsletters, automatic notifications and spam with no value

Exactly where the line goes is decided in the assessment, and you always decide what may be included.

Export formats

You don't need to export anything to get an assessment. When the time comes, the most common formats are:

• PST – common from Outlook and Exchange
• MBOX – common from many other email clients and systems
• Exports from Microsoft 365 or Google Workspace – exports of your own archives using the tools built into those services

Always check what your agreements with the provider say about export and use. We help you ask the right questions, but it is not legal advice.

De-identification and GDPR

Emails from employees and customers are personal data. Data protection authorities stress that reusing data for AI training must be compatible with the original purpose, and the European Data Protection Board has made clear that legitimate interest is not automatic support but requires a documented balancing test.

That is why we assess the legal basis and personal data first. Before delivery, names, email addresses, phone numbers, signatures and other details that could identify a person are de-identified – including in free text and attachments. Sometimes the people concerned also need to be informed.

De-identification is rarely a single step. It is often better to make a selection first – for example certain mailboxes, years or subject areas – than to try to clean the whole archive. The result is spot-checked before anything is delivered, and the work is documented. The documentation also helps the buyer meet the EU AI Act's requirement for a summary of training content (Article 53). Read more about GDPR and anonymization.

Employees and employee consultation

Emails are written by employees. Workplace privacy rules may apply when an employer uses employees' email for a new purpose. Depending on your country and any collective agreements, the decision may also require informing or consulting employee representatives or a works council before you go ahead. The same questions arise for chat data – see Slack and Teams data.

How it works

1. You describe the archive in our form – approximate number of emails, number of years and whether there is sensitive information. It takes about two minutes and needs no files.
2. We assess rights, personal data and demand, and get back to you within 48 hours on business days.
3. You approve – nothing is shared until you have said yes to a concrete offer and a non-disclosure agreement is in place.
4. Delivery and payment take place after de-identification.

By default the license is non-exclusive and limited in scope, time and use, and you keep ownership of the archive.