Lesson 4

Bland Models and Architecture

TABLE OF CONTENTS▼

    The Three Bland Phone Models

    While multiple LLMs are used in Bland’s standard call flow, users can choose between three main call processing models (base, enhanced, and turbo). Each has pro’s and con’s (Turbo is not always a better selection than Base).

    Model Differences

    While multiple LLMs are used in Bland’s standard call flow, users can choose between three main call processing models (base, enhanced, and turbo). Each has pro’s and con’s (Turbo is not always a better selection than Base).

    base

    • The default model, follows scripts/procedures most effectively

    • Supports all features and capabilities 

    • Best for custom tools

    enhanced

    • Much faster latency, supports complex conversations

    • Works best with objective-based prompts

    • Supports all features and capabilities

    turbo

    • The fastest latency possible, supports sophisticated and nuanced conversations

    • Limited capabilities currently (excludes transferring, IVR navigation, custom tools)

    • Extremely realistic conversation capabilities

    LLM Architecture for Bland Calls

    During the course of a call, your Agent is continuously processing:

    • The speech from the user to text 

    • The text through multiple LLMs 

    • The text response from the LLMs to speech 

    During the second step, three different LLMS process the call transcript to help the agent make decisions. These are the Navigational, Conversational, Data Extraction models.

    The Navigational Model

    The navigational LLM is constantly trying to determine if the Agent should proceed to the next node and which pathway it should take. 

    The Conversational Model

    This LLM is responsible for speech dialog in your conversations.   This will determine what’s actually said to the user. 

    Data Extraction Models

    This LLM is responsible for identifying any specified variables in your Conversational Pathways.

    NEXT

    Making Calls

    Making Calls

    Making Calls

    Table of contents▼
      Scroll to top