View generative AI-powered post-contact summaries - Amazon Connect
Get startedWhy a summary is not generated

View generative AI-powered post-contact summaries

Note

Powered by Amazon Bedrock: AWS implements automated abuse detection. Because generative AI-powered post-contact summaries is built on Amazon Bedrock, users can take full advantage of the controls implemented in Amazon Bedrock to enforce safety, security, and the responsible use of artificial intelligence (AI).

You can save valuable time with generative AI-powered post-contact summaries that provide essential information from customer conversations in a structured, concise, and easy to read format. You can quickly review the summaries and understand the context instead of reading through transcripts and monitoring calls.

Region and language availability: This feature is available only for Amazon Connect instances created in the US East (N. Virginia) and US West (Oregon) AWS Regions, and in English locales. It's available for the voice and chat channels.

The following image shows an example of generative AI-powered post-contact summaries on the Contact details page.

The following image shows an example of generative AI-powered post-contact summaries on the Contact search page.

Each contact has no more than one summary generated. Not all contacts will have a summary generated; for more information, see Why a summary is not generated.

Note

Chat contacts in supported locales and AWS Regions have summaries only for text/plain messages. By default the Test chat feature in the Amazon Connect admin website creates text/markdown messages that will not be used for summaries. To test integration, use a customized chat widget with only text/plain "supportedMessagingContentTypes".

Get started

  1. Add a Set recording and analytics behavior block to your flow.

  2. Configure the Properties page of the block:

    • Set Analytics to On.

    • Choose either Enable speech analytics, Enable chat analytics, or both.

    • If you choose redaction, the post-contact summary inherits the overall transcript redaction settings. If you choose granular settings, the post-contact summary still redacts all PII.

    • Under Contact Lens Generative AI capabilities, choose Post-contact summary.

    The following image shows a Properties page that is configured to enable chat analytics and Contact Lens Generative AI capabilities:

    The properties page of the Set recording and analytics behavior block.

Assign security profile permissions

Assign the following security profile permissions to users who need access to this feature:

  • Contact Lens–post-contact summary, View

  • Recorded conversations (redacted), View or Recorded conversations (unredacted), View

The following image shows where these permissions are located on the Analytics and Optimization section of the Edit security profile page.

The security profiles page, analytics and optimization permissions.

Why a summary is not generated

If a summary is not generated, an error message is displayed on the Contact details and Contact search pages. In addition, the ReasonCode for the error appears in the ContactSummary object in the Contact Lens output file, similar to the following example:

"JobDetails": {
    "SkippedAnalysis": [
      {
        "Feature": "POST_CONTACT_SUMMARY",
        "ReasonCode": "INSUFFICIENT_CONVERSATION_CONTENT"
      }
    ]
  },

Following is a list of error messages that may be displayed on the Contact details or search pages if a summary is not generated. Also listed is the associated reason code that appears in the Contact Lens output file.

  • Summary could not be generated due to exceeding quota of concurrent summaries. ReasonCode: QUOTA_EXCEEDED.

    If you receive this message, we recommend that you submit a ticket to increase the Concurrent post-contact summary jobs quota.

  • Summary could not be generated due to not enough eligible conversation. ReasonCode: INSUFFICIENT_CONVERSATION_CONTENT.

    For voice, there must be 1 utterance from each participant. For chat, there must be 1 message from each participant. Chat messages must be text/plain messages. If there are text/markdown messages, even if there is a lot of conversation between participants it won't be analyzed because text/markdown is not supported for this feature, and the error message will be displayed.

  • Contact Flow had invalid Contact Lens configuration for PostContact Summary, such as unsupported or invalid language code. ReasonCode: INVALID_ANALYSIS_CONFIGURATION.

    This error is returned if the enabled summary is incompatible with other contact lens settings, particularly if it's enabled for an unsupported locale.

  • Summary cannot be provided because it failed to satisfy security and quality guardrails. ReasonCode: FAILED_SAFETY_GUIDELINES.

    The generated summary cannot be provided because it failed to satisfy system safety guidelines.