Enable in-flight sensitive data redaction and message processing

Connect Customer supports message processing that intercepts and modifies chat messages before they reach any participant. This capability enables automatic redaction of sensitive data and custom message processing, helping businesses maintain compliance and security standards.

Processing options

The following are processing options, along with features of each option:

Built-in sensitive data redaction

Custom message processors (via Lambda)

With a custom processor Lambda, you have the freedom to transform messages in any way you choose. A copy of each message is sent to your Lambda function, and if a valid response is received within the timeout, the content in the response becomes the new canonical form of that message.

This enables use cases such as:

For more information about Lambda, see What is Lambda? in the Lambda Developer Guide.

How it works

When message processing is enabled, every plaintext, markdown, and JSON message passes through the configured processor before being delivered to participants. The processor can modify, approve, or reject the message. If processing fails, you can configure whether the original unprocessed message is delivered or suppressed.

Message processing stays active for the entire duration of the chat, even if an individual contact segment ends (for example, during a transfer). Processing does not retroactively apply to messages sent before it was enabled.

Configure message processing

You can enable message processing using either of the following methods:

To stop message processing before the chat ends, call the StopContactMediaProcessing API. For more information, see StopContactMediaProcessing in the Connect Customer API Reference.

Enable built-in sensitive data redaction

In-flight sensitive data redaction is powered by Conversational Analytics. To enable it, configure the Set recording, analytics and processing behavior flow block with the following settings:

For more information about how Conversational Analytics redacts sensitive data, see Enable redaction of sensitive data.

Create a custom message processor

You can create a custom Lambda function that transforms messages in-flight. Custom processors can perform any transformation, such as language translation, profanity filtering, or AI-powered message enhancement.

Step 1: Create your Lambda function

Your custom processor Lambda receives an input event in the following format:

{
  "version": "1.0",
  "instanceId": "string",
  "associatedResourceArn": "string",
  "chatContent": {
    "absoluteTime": "string",
    "content": "string",
    "contentType": "string",
    "id": "string",
    "participantId": "string",
    "displayName": "string",
    "participantRole": "string",
    "initialContactId": "string",
    "contactId": "string"
  }
}

Your Lambda function must return a response in the following format:

{
  "status": "PROCESSED | APPROVED | REJECTED",
  "result": {
    "processedChatContent": {
      "content": "string",
      "contentType": "text/plain | text/markdown | application/json"
    }
  }
}

The processed content replaces the original message when published to chat participants.

Step 2: Grant Connect Customer permission to invoke your Lambda function

You must grant Connect Customer permission to invoke your Lambda function:

Alternatively, you can use the CreateIntegrationAssociation API with IntegrationType set to MESSAGE_PROCESSOR. For more information, see CreateIntegrationAssociation in the Connect Customer API Reference.

Step 3: Activate your Lambda function

Choose one of the following methods to activate your custom processor:

Option 1: Use the flow block

Configure the Set recording, analytics and processing behavior flow block with the following settings:

Option 2: Use the StartContactMediaProcessing API

Alternatively, you can activate message processing by calling the StartContactMediaProcessing API. This requires your instance ID, the chat's contact ID, the Lambda processor ARN, and a failure mode. For more information, see StartContactMediaProcessing in the Connect Customer API Reference.

Additional information

Transcript storage

Important considerations