Sell your customer support tickets for AI training

What AI buyers want from tickets

Customer service is one of the main jobs AI agents are built to do. A ticket shows a real request, the steps an agent took, the policy applied and the outcome. Exceptions and escalations are the most useful part, because that is where agents fail.

• Full threads with resolution, tags and status changes
• Escalation paths and who approved exceptions
• Macros and help-center articles the agents used
• QA scorecards and reviewer notes

What makes ticket data worth more

Sunset prices data on "data richness," meaning how well records link across systems (Forbes). The same logic applies to tickets. A linked ticket is a complete lesson. A standalone ticket is a snapshot.

Price signals

No public deal sets a per-ticket price. These are the closest public figures:

• Published enterprise tiers: $100K+ for qualified companies, $500K+ for large-scale or multi-team, $1M+ for highly unique data. These cover a company's whole data package.
• Troveo added an "enterprise workflow" category for business data (IT Brief) and says business-data deals "typically start at six figures" (Troveo, company claim).
• Data-labeling startups have paid HVAC companies $150,000 for data (The Information).

See how much company data is worth for the full picture.

Privacy and legal risks

• CCPA. Making customer personal information available for money may be a "sale" unless it meets the de-identification standard: reasonable measures, a public no-re-identification commitment and contracts binding recipients (Cal. Civ. Code 1798.140).
• Your own promises. If your privacy policy did not cover this use, changing it quietly could be unfair or deceptive, per the FTC (FTC). The FTC also says "there is no AI exemption from the laws on the books" (FTC).
• Healthcare. Patient tickets may contain PHI. HIPAA bars selling PHI without authorization. De-identified data is not PHI, via Safe Harbor or Expert Determination (HHS).
• Financial services. GLBA requires a privacy notice and opt-out before sharing nonpublic personal information with nonaffiliated third parties (FTC).
• Card data. Under PCI rules, sensitive authentication data must never be stored after authorization. If it shows up in tickets, purge it.
• Client tickets. If you run support for clients, the tickets may belong to them. See client data vs. company data.

General information only, not legal advice. See CCPA and AI data licensing.

How to prepare ticket data

1. Export tickets with metadata: tags, timestamps, status history and agent IDs.
2. Replace customer names, emails, phones, addresses and order numbers with consistent placeholders.
3. Strip free-text fields that often hold sensitive data: attachments, screenshots, payment notes.
4. Replace agent names with role labels.
5. Bundle the matching SOPs, macros and QA rubrics.
6. Agree on sample review before full delivery. Enterprise programs state they review samples and removes PII where relevant.

Operating vs. shutting-down companies

Ticket data fits operating companies well, because current SOPs and QA processes give the tickets context. Enterprise programs look for companies with 30+ employees and established documentation. A closing company can sell historical tickets, but it may lose the people who can explain its processes. Shutdown services such as SimpleClosure exclude employee and health records from what they sell (01net), so expect similar limits on ticket data that touches either.

Curious what your help desk is worth? Take the 60-second estimate.

Frequently asked questions