Lesson 2

Common Prompt and Pathway Mistakes

TABLE OF CONTENTS▼

    Common Problems In Pathways

    The majority of Pathway issues fall into the following categories:

    Stuck and Looping Agents

    If an agent can’t seem to get the info required,  and has no alternative route for their pathway you will see long looping transcripts that don’t make much sense.  This is easily monitored by call-times (if you know your expected average call duration),  and becomes very obvious while reading transcripts. 

    Variable Extraction Related Errors

    Making sure that your Agent pulls out the right information from the conversation, in the right format can be a major source of inefficiencies in your Agent’s conversation.  Often, the Agent’s script is good,  but it doesn’t handle the data in the way you need for your pathway (ie an external API request).    This is seen through good API logging practices, and ‘bad data’ type responses from your external APIs.   These will show up as ‘Errors’ in your Bland logs. 

    Unforeseen Conditions

    Sometimes in our design of a Conversational Pathway we forget to account for a specific condition or request by the User that’s called.  While a well-written Agent prompt can seamlessly handle variations within a conversation,  sometimes a request so unforeseen occurs that throws them off the script. 

    Bland Agents are designed to always work to find a solution, and if unprepared (via prompts)  there is always the possibility of deviating way off script and not being useful to the caller. These types of situations can be addressed by providing more/better fine-tuning dialog examples for your nodes.

    NEXT

    A Poor Examples and How We Fix Them

    A Poor Examples and How We Fix Them

    A Poor Examples and How We Fix Them

    Table of contents▼
      Scroll to top