Catch maiden names
and other PII, automatically.

A different approach to PII detection.

A mother’s maiden name is the surname she held before marriage, and it is Personally Identifiable Information (PII): a unique identifier tied to one specific person, the kind of detail financial institutions, insurers, and healthcare providers have used as a security question for decades. That makes it valuable to an identity thief and, increasingly, a compliance liability the moment it is spoken out loud on a recorded line. Most PII policies were written for text and forms. Very few were written for the moment a caller says a maiden name into a live phone call.

Why maiden names are still a security question

Verification calls have not caught up with how much of this information now lives on a recording. A claims agent asks for a maiden name to confirm identity, a bank rep asks for the same, a healthcare intake line asks for a date of birth right after it. The words get transcribed. The PII inside those words rarely gets flagged, so it sits in a shared transcript, an export, or a training file, unredacted, long after the call that generated it is forgotten.

How Speak AI reads for PII, not just words

Speak AI listens to every call the way a compliance officer would, at machine speed. Each recording is transcribed in your language, with 100+ supported, and then read on three layers: the words themselves, the tone and energy behind them, and the visual or contextual cues on a video call. A stated maiden name, social security number, date of birth, or account number is recognized in the audio, not just the transcript text, and flagged as a structured field your systems can query through the MCP server from Claude, ChatGPT, or Cursor.

What gets flagged on a verification call

• Mother’s maiden name, spoken as a security answer
• Social security or national ID numbers
• Date of birth and account or policy numbers
• Home address and contact details given for verification
• Payment card or banking details read aloud mid-call

From a spoken answer to a protected record

The result is a transcript that flags its own PII instead of hiding it. Redaction happens at the field level, so the rest of the call, the actual complaint, the claim details, the resolution, stays intact and searchable. Trends across hundreds of verification calls become a report instead of a guess, and dashboards you can customize and white-label track PII exposure and redaction rates over time, month over month. One legal intelligence firm ran this across 5,100+ hours of carrier calls; read the large-scale comparative analysis on how it saved $700K+ without adding headcount.

And because verification calls rarely happen in isolation, the same engine scores and codes those calls on your own criteria, connecting PII detection to call scoring and coaching across every conversation your team has.