Speaker detection in interview processing

Overview

When you upload an interview transcript to the Data Hub, Journey AI automatically detects the speakers in the conversation and lets you configure how each one is handled before processing begins. This includes choosing which speakers to include in quote extraction and assigning personas so insights stay organized by customer segment.

PII (personally identifiable information) is stripped automatically during upload — you don't need to clean transcripts beforehand.

Before you start

  • AI must be enabled for your organization, and Remove personally identifiable information must be turned on in Settings > Details & AI. This is an admin action.
  • Your transcript file should be in .txt format or pasted directly as text.
  • Use consistent speaker labels throughout the transcript (for example, Interviewer: and Participant:) to help Journey AI detect speakers accurately.
  • It helps to have your personas set up in advance, so you can assign them during the configuration step.

Steps

  1. Go to the Data Hub.
  2. Select Add source and choose your interview file, or paste the transcript text directly.
  3. Journey AI will detect the file type as an interview and identify all unique speakers. You'll see a speaker configuration screen listing each detected speaker.
  4. For each speaker, decide whether to include or exclude them from quote extraction.
    • Toggle off researchers, interviewers, or observers you don't want included.
    • Leave customers or participants toggled on.
  5. For each included speaker, use the dropdown to assign a persona — or leave it unassigned if you'd prefer not to attribute quotes to a segment.
  6. Review your settings, then select Continue to process the file.
  7. Journey AI processes the transcript, strips PII, and makes quotes available for insight mining and journey enrichment.

Tips

  • Exclude the interviewer by default. Unless you specifically need researcher questions included, toggling them off keeps the focus on customer voices.
  • Assign personas before confirming. Speaker configuration can't be changed after processing — if you need to update assignments, you'll need to re-upload the file.
  • Map multiple speakers to the same persona if they represent the same customer segment. All their quotes will be attributed to that persona.
  • Batch similar interviews together. Processing interviews from the same segment in sequence makes it easier to stay consistent with persona assignments.