Intelligent Summaries for conversational data
Functionality, requirements, workflows, and field usage for Intelligent Summaries for conversational data in Medallia Experience Cloud
Intelligent Summaries use generative artificial intelligence (AI) to describe a conversation between an agent and customer in a single paragraph that provides actionable insights.
Each summary is a single, unstructured string of text that includes these logical components:
- Whether the customer's issue was resolved.
- The primary reason for contact.
- The overall resolution or outcome of the call.
- A summary of the conversation.
When the Intelligent Summaries feature is enabled, Experience Cloud generates summaries for new records that contain valid conversation data. Experience Cloud modules can display those summaries wherever the original transcript is shown. Summaries can be rated for accuracy by the end-user. No configuration is required for summaries to be generated, but additional optional configurations can be used to add summaries to reports and dashboards.
Requirements, languages, availability, and data compliance
Intelligent Summaries supports Medallia Speech transcripts and imported chat interactions.
To be summarized, Speech and chat transcripts must be in a supported language, and the transcript must be between 550 and 32,000 characters. Additionally, the Speech transcript must be generated from two-speaker, supported audio, and the record's Survey language (a_survey_language) System field must be a supported language.
Supported languages include English, Spanish, French, and German. Summarization output is in the transcript's original language. For more information about language support with genAI features, see Language support for generative AI.
Intelligent Summaries can be enabled according to data center availability. Contact your Medallia expert for details.
In compliance with GDPR requirements, when a record that has been summarized is deleted, associated summary data is also deleted.
Role permissions
Users must have the View Intelligent Summaries permission enabled for their role to see Intelligent Summaries.
Summaries may be edited if the user’s role has the Edit Summary permission enabled.
For more information about adding role permissions, see Roles screen and Reporting permissions.
Summarization workflows
Summarization processing and configuration vary by transcript type:
Medallia Speech transcripts
-
Medallia Speech generates a transcript from two-speaker English audio. Audio may be imported by Experience Cloud or sent to Experience Cloud by an ACD/CCaaS platform.
-
With the Intelligent Summaries feature enabled, Experience Cloud monitors incoming data for new records that contain a Speech transcript in the Medallia Speech transcript (q_speech_comment) feedback field.
-
Experience Cloud evaluates and queues the record for summarization.
-
Experience Cloud summarizes the transcript.
-
Experience Cloud stores the summary and updates Intelligent Summaries System fields.
-
Summaries, summarized content, and feedback and editing options are displayed as configured in Experience Cloud reports and modules.
Restriction: Feedback and editing options are available only with Responses Form and Responses Feed.
Chat transcripts
-
A chat interaction generates a transcript. Chat transcripts may be imported by Medallia Experience Cloud or sent to Experience Cloud by an ACD/CCaaS platform.
-
With the Intelligent Summaries feature flag enabled, Experience Cloud monitors incoming data for new records that contain a chat transcript in the Conversational transcript (q_conversational_transcript_comment) feedback field.
-
Experience Cloud evaluates and queues the record for summarization.
-
Experience Cloud summarizes the transcript.
-
Experience Cloud stores the summary and updates Intelligent Summaries System fields.
-
Summaries, summarized content, and feedback and editing options are displayed as configured in Experience Cloud reports and modules.
Restriction: Feedback and editing options are available only with Responses Form and Responses Feed.
Fields used
These categories are useful for understanding how Intelligent Summaries uses System fields:
-
Summarized fields contain content to be summarized. This original transcript persists throughout the summarization process and remains available afterward.
-
Summary fields contain the summary of the content stored in a summarized field.
-
Extracted summary fields contain a logical component of a summary stored in a summary field, such as the primary reason for contact.
-
Summary support fields do not contain AI-generated data, but provide information related to the summarization system process, such as a completion timestamp.
Intelligent Summaries summarizes a Medallia Speech transcript stored in the Medallia Speech transcript (q_speech_comment) feedback field or a chat transcript stored in the Conversational transcript comment (q_conversational_transcript_comment) feedback field. These are summarized fields that store original transcripts.
Summaries are stored as metadata in the summarized feedback field and in one of two System fields (A-fields) depending on the originating transcript: Speech transcript summary (a_speech_comment_summary) for Speech transcripts or Conversational transcript summary (a_conversational_transcript_comment_summary) for chat transcripts. These are summary fields.
Extracted summary fields
Extracted fields contain logical components of summarized content, such as the primary reason for contact. By default, extracted summary fields are shown with the summary itself in a summary box.
Medallia Speech extracted summary fields
- Medallia Speech transcript primary reason
- Primary reason for contact. Consists of one word minimum to four words maximum.
- Medallia Speech transcript resolved?
- Whether or not the primary reason for contact was resolved during this interaction.
- Medallia Speech transcript outcome
- Outcome of the interaction, including how the primary reason for contact was resolved, or why the primary reason for contact was not resolved.
Chat extracted summary fields
- Conversational transcript comment primary reason
- Primary reason for contact. Consists of one word minimum to four words maximum.
- Conversational transcript comment resolved?
- Whether or not the primary reason for contact was resolved during this interaction.
- Conversational transcript comment outcome
- Outcome of the interaction, including how the primary reason for contact was resolved, or why the primary reason for contact was not resolved.
These System fields are used to support summary fields:
Medallia Speech summary support fields
- Medallia Speech transcript is summarized?
- Whether or not the Speech transcript summary field has been summarized.
- Medallia Speech transcript summarization date
- Date and time when the summarization processing is complete.
- Medallia Speech transcript summarization status
- Indicates the summarization processing status, which may be:
- Pending
- The referenced field has been marked as summarizable and its record has persisted in the database queue.
- In-queue
- The field's record is queued for summarization processing.
- Completed
- The record was summarized successfully.
- Unexpected_failed
- Summarization failed for the given field because of an internal error or unexpected behavior.
- Expected_failed
- Summarization failed because the record or comment field does not contain valid content. For example, if the comment field contains more than the maximum valid number of characters, the field is not summarized.
- Not_requested
- The record was not queued for summarization processing. For example, a record may not be queued for summarization because the record itself is too old, or because the record's language or comment field is not supported.
Chat transcript summary support fields
- Conversational transcript comment is summarized?
- Whether or not the Conversational transcript comment field has been summarized.
- Conversational transcript comment summarization date
- Date and time when the summarization processing is complete.
- Conversational transcript comment summarization status
- Indicates the summarization processing status, which may be:
- Pending
- The referenced field has been marked as summarizable and its record has persisted in the database queue.
- In-queue
- The field's record is queued for summarization processing.
- Completed
- The record was summarized successfully.
- Unexpected_failed
- Summarization failed for the given field because of an internal error or unexpected behavior.
- Expected_failed
- Summarization failed because the record or comment field does not contain valid content. For example, if the comment field contains more than the maximum valid number of characters, the field is not summarized.
- Not_requested
- The record was not queued for summarization processing. For example, a record may not be queued for summarization because the record itself is too old, or because the record's language or comment field is not supported.
To use these fields for report filtering, ensure they have been added as filters from the Admin Suite > Control Panel Configuration > Primary Filters screen. For more information, see Filters and control panels.
Supported reports, modules, data export, and filtering
Summary fields may be used with the following reports, or in any other Alchemy Experience Reporting module that supports System fields (A-fields). For example, extracted summary fields ending in ...resolved? could be used to filter score data by whether the customer's issue was resolved.
- Responses List
- Any Intelligent Summaries System field (A-field) can be used as a column. Ensure the fields to be used as columns are enabled with the Admin Suite > Reports > Responses List > Filter columns > Fields screen. Feedback options are not available.
- Responses Feed
- When a Medallia Speech or chat transcript field is added to Responses Feed, if a summary exists, the summary box is displayed before the transcript in the Comment Data section, and feedback options are available. For example, the following image shows an interaction with a summary in Responses Feed.
- Responses Form (Medallia Speech)
- When a Speech transcript field is added to Responses Form, if a summary exists, the summary box is displayed before the transcript, and feedback options are available. For the Speech transcript to be displayed, the Medallia Speech transcript (q_speech_comment) feedback field must be configured in the Responses Form Comments module. For example, the following image shows a Responses Form summary box.
- Responses Form (chat)
- Conversational transcript comment (q_conversational_transcript_comment) does not need to be added to the Responses Form configuration for a chat transcript summary to display. The chat transcript UI is shown automatically, and if a summary exists, the summary box is displayed before the transcript, and feedback options are available.
- Recent Responses
- Users can hover on a response, typically represented by a face icon, to see a summary of the interaction in the transcript field, and can click through to see the full record. Feedback options are not available. For example, the following image shows an interaction summary in the hover pop-up.
The summary fields Speech transcript summary (a_speech_comment_summary) and Conversational transcript summary (a_conversational_transcript_comment_summary) may be configured to display the summary without the full text of the original transcript. In this case, feedback options are not available.
These fields can be used for data exports:
-
Medallia Speech transcript
-
Medallia Speech transcript summary
-
Medallia Speech transcript primary reason
-
Medallia Speech transcript outcome
-
Conversational transcript comment
-
Conversational transcript comment summary
-
Conversational transcript comment primary reason
-
Conversational transcript comment is resolved?
-
Conversational transcript comment outcome
Report and dashboard data may be filtered by these fields:
-
Medallia Speech is summarized?
-
Medallia Speech summarization date
-
Medallia Speech summarization status
-
Medallia Speech transcript is resolved?
-
Conversational transcript comment is summarized?
-
Conversational transcript comment summarization date
-
Conversational transcript comment summarization status
-
Conversational transcript comment is resolved?
Other summary fields can be configured as indexable and become searchable to filter for all summaries containing certain content.
Medallia Speech transcript summary and Conversational transcript comment summary cannot be used as filters.
Edit Intelligent Summaries
When the user's role has the Edit Summary permission enabled, that user may edit full summaries and extracted summary fields. For more information about enabling the Edit Summary role permission, see Roles screen and Reporting permissions.
User-made edits to summary and extracted field are saved to the record and are displayed instead of the AI-generated summary when other users view the record. Only the latest edited versions of summary and extracted fields are stored and displayed. When a record's summary and extracted fields have been edited, the text shown ends with (edited).
Summary and extracted field edits provide Medallia with examples of user updates to inaccurate or otherwise erroneous records for future model training. Summary and extracted field edits do not affect the system that generated the summary.
AI-generated data transparency
When AI-generated content or data is displayed in Experience Cloud, the generative AI icon displays with the module title or data, and a text disclaimer displays in the module's footer or after the relevant content. For example, the following image shows an AI-generated summary with the generative AI icon.
When all fields in a module are populated with AI-generated data, the generative AI icon displays with the module title. For example, the following image shows a module populated entirely with AI-generated content, including the icon and disclaimer.
When a text field contains AI-generated data, the generative AI icon displays next to that text field's label, and the disclaimer displays after the relevant content. When only some text fields in a module are populated with AI-generated data, the icon is displayed on each text field but is not displayed with the module title. For example, the following image shows a module with only one row with AI-generated data, so the row header shows the icon, but the module title does not.
Model data, feedback, and training
Medallia generative AI products use large language models trained on Medallia experience data. Medallia models are not trained with client data.
When its original data sources are displayed, AI-generated content includes a feedback prompt for the user asking whether the generated content is helpful. Only one feedback response can be submitted per user per session for each unique piece of AI-generated content.
Only the full summary may be rated for accuracy; individual extracted fields cannot be rated.
Summary feedback provides Medallia with examples of accurate and inaccurate records for future model training. Ratings and feedback affect neither the summary itself nor the system that generated it.
