Sell call center data to AI companies

What call center data do AI companies buy?

AI companies buy call center data that pairs real conversations with what happened next. A recording on its own is useful. The same recording with a transcript, the QA score, the disposition code and whether the customer called back is far more useful.

• Call recordings with transcripts and speaker labels
• Chat and email transcripts from the same queues
• QA scorecards linked to the calls they assess
• Agent scripts, call flows and knowledge base articles
• Escalation notes, disposition codes and resolution outcomes
• Calls in languages other than English, which are scarcer

Buyers training voice agents want the audio. Buyers training text models may want transcripts only, which are easier to de-identify. See call recording data and speech and audio data for how each is used.

If your platform records the agent and the caller on separate channels, keep them separate in the export. Split channels make speaker labeling and redaction much easier. Record the codec, sample rate, queue and date range for each batch in your data manifest.

How much do AI companies pay for call center data?

AI companies publish a range of $20,000 to $250,000 for call center data. That is Mercor's referral listing for call centers of 50 to 5,000 agents, as reported by ddiy.co in October 2026.

The $250,000 ceiling is the lowest of the 11 Mercor listings ddiy.co reports. Call data is plentiful, so buyers pay more for linked outcomes, specialist queues such as technical support or claims handling, and languages that are hard to source. Transcripts from regulated queues, such as debt collection or insurance claims, need more careful redaction. A free data valuation tells you whether yours stands out.

Who owns the recordings: you or your clients?

The recordings usually belong to your clients if you are an outsourcer, and to you if you run an in-house contact center. This is the first question to settle, because it decides whether there is anything to sell.

A BPO handling calls for a client normally holds callers' personal information as the client's service provider. The CCPA, at California Civil Code section 1798.140(ag), defines a service provider by a contract that prohibits retaining, using or disclosing the personal information for any purpose other than the business purposes specified. On healthcare queues, a business associate may use or disclose protected health information only as its business associate contract permits, under 45 CFR 164.502(a)(3).

What an outsourcer usually owns outright is its own material: agent training content, QA frameworks, scripts it wrote, and internal operations data. In-house contact centers face a simpler question, because the company that took the calls usually controls the recordings, but the privacy notice and recording message in force at the time still apply. Read each outsourcing contract for any rights it gives you over de-identified or aggregated data.

Recording consent laws affect a data sale because they decide whether a recording was lawfully made in the first place. The federal Wiretap Act, at 18 U.S.C. section 2511(2)(d), allows a private party to intercept a call where one of the parties has given prior consent.

Some states go further. California Penal Code section 632 makes it an offense to record a confidential communication intentionally and without the consent of all parties, with a fine of up to $2,500 per violation for a first offense. Frankfurt Kurnit Klein and Selz notes that state call-recording consent laws apply in the US, and that companies must comply with the privacy statements in force when data was collected.

A message saying calls are recorded for quality and training purposes was written with your own staff in mind, and may not extend to licensing the recording to another company. The FTC warned in February 2024 that adopting more permissive data practices, such as AI training, through a quiet retroactive change may be unfair or deceptive. These points make de-identified transcripts and your own content the easier place to start, see de-identification for AI training data.

What must be removed from call recordings?

Card details, voiceprints and sensitive personal details must be removed from call recordings before any sale. Each is governed by its own rule.

The PCI Security Standards Council's FAQ 1210 says storing card validation codes in any digital audio recording after authorization breaches Requirement 3.3.1. Under BIPA, which lists voiceprints as biometric identifiers, a negligent violation carries liquidated damages of $1,000 and an intentional or reckless one $5,000, per violation.

If you hold UK or EU data

Calls handled for UK or EU clients follow the GDPR. Article 28(3)(a) of the UK GDPR limits a processor to the controller's documented instructions, and health, religion and other special category details that callers disclose fall under Article 9, so exclude those calls or redact them by queue.

How does selling call center data work with Sell My Data To AI?

Selling call center data with Sell My Data To AI starts with a consultation about your platforms, queues, languages and client contracts. We work out which data you own, which needs client agreement and which must be excluded, before anyone else is involved.

We then run our eight vetting checks, including a re-identification test on a sample, and share a redacted sample only with buyers you approve. Book a free consultation to start.