Conversation History

The Conversation History report offers an end to end view of each customer engagement. Though looks similar to the Task History report, in contrast to the agent focus of that, the Conversation History presents information from the customer perspective.

It offers insight into engagements that did not involve human agents, e.g. completed by conversational bots, or self-served in a workflow IVR, etc.

Conversation history report layout

The top section of the report is used to filter the data presented below.

Below the filters is the tabular list with each row representing a single customer engagement, or conversation.

Filtering the conversation history

As with other historical reports, you can filter the data loaded as per the following parameters:

  • Conversation direction (inbound or outbound)
  • Media type
  • Date range, offering the last 7 days worth of data by default
  • Duration of the conversation
  • Channel the conversation was handled by
  • The workflow that routed the conversation

The list can also be ordered by the date-time, newest first or oldest first.

Conversation history records

The conversation history records contain the following fields:

  • A media type icon, e.g. phone for voice, envelope for emails, etc.
  • The date and time the conversation was created.
  • The name of the customer in the customer database (if identified).
  • The customer satisfaction rating (if provided).
  • The name of the channel that managed the conversation.
  • The workflow that routed the conversation.
  • The duration of the conversation.
  • The number of agents that were involved with the conversation.
  • Recording playback.
  • Downloading the textual transcript of the conversation.
  • A short AI generated summary of the transcript of the conversation between the customer and any bots or human agents.

Conversational bot decision trace

As Buzzeasy conversational bots rely on AI models for their decisions, it can become challenging to reveal how and using what inputs did it base its responses to individual customer questions, or why it decided to involve other bots and nodes in the workflow, and how long each internal steps it employed actually took, etc. Tracing these events for such conversations is available in text based media channels.

Do the followings to gain insight into individual conversational bot responses, the data queried and pulled from various sources like knowledge bases, custom integrations, etc.:

  1. Using the conversation filter options, find the item that involved a conversational bot either partially or throughout the entire interaction.

  2. Click the Conversation History icon for the selected item. A new modal opens up with the chronological transcript of all participant's messages, including any conversational or classic bots as well as human agents.

  3. Note that the customer messages are aligned to the left while Buzzeasy participant's are justified to the right side of the transcript. Note also that the various different Buzzeasy participant types (human agents, classic bots or conversational AI bots) are labeled differently.

  4. Click a Buzzeasy message labeled as AI. The view expands with an additional panel, listing the chronological flow of events or actions the conversational bot performed, with the time each of these took.

    Depending on the complexity of the bot's configuration and the actual customer question, the bot might have decided to utilize the Buzzeasy knowledge base or a custom integration. Therefore the list might contain one or more entries. Each of these can be expanded to reveal further details:

    • An AI Model Prompt: This shows up at least once in the event list, possibly more. The very first item represents the bot's initial assessment and comprehension of the customer question. It contains the verbatim bot instruction from the workflow node to help you identify how it fits the customer question and the eventual answer provided. If the bot is not configured for—or decided to not use—knowledge bases or custom integrations, then this item also represents the answer generation step for the customer, in which case no further bot events are shown.

      If however the bot performed at least one KB query or custom integration API call to fetch particular bits of data (see below for details), then additional copies of this event might also be seen in the list. In case such queries do not return comprehensible data, then even more copies of this event can show up representing the bot's attempt to gather the data in some other ways.

      The last item at the bottom of the list represents the bot's effort to generate the response using all the information collected.

    • Knowledge Base Search: Such events represent one query against a configured Buzzeasy knowledge base and the top 3 results it returned with the confidence level the bot identified the result qualifies as.

    • Custom Integration Call: These represent a query against a preconfigured external system's API. The event details the target URL with the data passed over in the request as well as the response code and any data sent back.

  5. The event list is closed with either the answer the bot could come up with, or if that failed for whatever reason, the name of the conversational bot node output that it routed the conversation towards in the workflow.