AWS S3 integration

Automatically sync call recordings and transcripts to your S3 bucket for long-term storage and compliance.

Automatically sync call transcripts and audio recordings to your own AWS S3 bucket for long-term storage, compliance archiving, and integration with internal data pipelines.

What gets synced

Data type	Format	Description
Transcripts	JSON	Full conversation transcripts with one entry for each utterance
Audio recordings	WAV	Complete call recordings

PolyAI syncs transcripts for voice calls only. Calls from the Agent Studio test panel have no recording, so PolyAI does not sync a transcript for them.

File structure

Each call has two files in the same folder:

• <conversation_id>.wav: the call recording.
• transcript-<conversation_id>.json: the call transcript.

The <conversation_id> value is the same in both file names. Use it to match a transcript to its recording.

The folder path depends on the bucket structure of your account:

Date and hour folders

<account_id>/<project_id>/call/<YYYYMMDD>/<HH>/<conversation_id>.wav
<account_id>/<project_id>/call/<YYYYMMDD>/<HH>/transcript-<conversation_id>.json

<YYYYMMDD> and <HH> are the date and hour when the call started, in UTC.

Flat folder

<account_id>/<project_id>/call/<conversation_id>.wav
<account_id>/<project_id>/call/transcript-<conversation_id>.json

Your PolyAI representative confirms the bucket structure and the destination prefix in your bucket during setup.

Transcript schema

A transcript file is a UTF-8 JSON array. Each object in the array is one utterance. The array is in chronological order.

transcript-<conversation_id>.json

[
  {
    "message_id": "0199a1b2-7c3d-7e4f-8a9b-0c1d2e3f4a5b",
    "participant": "agent",
    "created_at": "2026-09-17T11:00:00.123000+00:00",
    "transcript": "Hello, how can I help you today?"
  },
  {
    "message_id": "0199a1b2-8d4e-7f5a-9b0c-1d2e3f4a5b6c",
    "participant": "user",
    "created_at": "2026-09-17T11:00:04.456000+00:00",
    "transcript": "I want to book a table for two."
  }
]

Field	Type	Description
message_id	string	The unique ID of the utterance, in UUID format.
participant	string	The speaker. The value is user, agent, or live_agent.
created_at	string	The time of the utterance, in ISO 8601 format and UTC.
transcript	string	The text of the utterance. The value is an empty string when the caller says nothing, for example after a no-input timeout.

The transcript does not include internal system messages.

When to use S3 sync

• You need to retain call data for compliance (HIPAA, GDPR, PCI-DSS)
• You want to run your own analytics or ML pipelines on conversation data
• You require downloadable audio files (the Studio UI supports playback but not download)
• You process high call volumes and need scalable, automated data transfer

How to get started

1. Contact PolyAI

   Reach out to your PolyAI account manager or contact PolyAI to discuss your data transfer requirements.

2. Provide your S3 bucket details

   Share your AWS account ID, target S3 bucket ARN, and preferred region with your PolyAI representative.

3. Configure IAM policies

   PolyAI will provide an IAM policy template. Apply it to your bucket to grant PolyAI write access.

4. Test and validate

   PolyAI runs test transfers to your bucket. Verify the data format and file structure meet your needs.

5. Go live

   Once validated, PolyAI enables continuous sync for your production environment.

Related pages

List conversations

Retrieve conversation metadata programmatically through the API.

Studio transcripts

Review transcripts and play recordings directly in Agent Studio.