TABLE OF CONTENTS
Introduction
This document explains the different response modules available in the chatbot product NLP, ExoMind, and GenAI (Bot Mandate), including how they work, how to enable them, and when to use each.
Our chatbot platform supports multiple response modes:
NLP (Natural Language Processing) - Default mode for predefined, intent-based workflows. Handles intent-based queries, that is, when the bot already knows the possible user intents and has pre-defined responses or workflows mapped to those nodes.
ExoMind (ExoInsights) - AI-driven module using Retrieval-Augmented Generation (RAG) to understand, fetch, and answer from trained documents or websites.
GenAI Bot (Bot Mandate) - Pure Generative AI bot using an LLM-driven architecture to create dynamic, free-flow conversations or connect with ExoMind sources based on the prompt provided.
We allow an NLP bot to have GenAI flows or vice versa, which allows both structured flows and AI-driven responses for FAQ in a single bot.
Understanding the Bot Types
NLP Chatbot
What is NLP Bot
It is used for rule-based and intent-driven flows for both static and dynamic use cases, where the intent detection is based on the pre-trained keyword matching. based
This can be used in all cases, such as basic workflow, FAQs, User data collection flow, or having multiple API integrations.
Use Case Examples: “Apply for Loan”, “Check Balance”, “Track Order”.
How It Works
The chatbot checks if the user’s query matches the predefined intent trained on keywords. We do string matching to detect the user query.
If a match is found, the bot executes the corresponding flow using the pre-configured responses.
If not, the query proceeds to show closest suggestions (did you mean) or provides a configured bot break failure message.
Otherwise, the query proceeds to the fallback logic of providing responses from ExoMind or GenAI if enabled.
Availability
It is available by default in all bots.
How to Enable and Use
Available by default on all bots.
Manage intents and flows directly from the Bot Builder - Manage Intents section.
ExoMind (ExoInsights)
What is ExoMind or ExoInsights
ExoMind is a Retrieval-Augmented Generation (RAG) service that enhances your chatbot’s intelligence by enabling it to search, understand, and respond using data from your own documents or websites.
It acts as an AI-powered knowledge layer that works alongside your chatbot’s traditional NLP system.
When It Works
If enabled on a bot, it works when there is no NLP flow or NLP fails to identify an intent available in the bot.
How It Works
User asks a query - The system first checks if the question matches any existing NLP intent (e.g., “Apply for Loan”).
If no match is found, the query automatically falls back to ExoMind.
ExoMind searches through your trained data sources uploaded PDFs, or URLs and retrieves the most relevant “chunks” of text.
These chunks are sent to an LLM (Large Language Model) along with a prompt, which then generates a natural, contextual answer for the user.
Key Features
Train multiple data sources: PDFs and website URLs.
Optionally show source links in answers.
Configure custom system prompts for improved contextual responses.
Availability
Available on all the bots, needs to be enabled.
How to Enable
Please read the steps here: Enabling and Configuring ExoMind
Or reach out to your respective Account Managers or raise it to our support team (help@exotel.com) for any issues.
How to Use
GenAI Bot (AI Agent)
What is Gen AI Bot:
The Gen AI Bot, also referred to as the AI Agent, is an advanced AI-driven chatbot that uses LLMs instead of traditional NLP for understanding user queries and its intent and responding to user queries.
It allows your chatbot to move beyond predefined intent-based journeys and instead generate natural, context-aware, and dynamic responses using AI.
The Gen AI bot requires a detailed prompt followed up with configurations for which response model to use:
ExoMind (upload PDFs or URL) - Bot will respond from these sources only.
API - A Customer's API can be integrated with a sample request response to extract a dynamic response.
Internet - Bot will respond from the information available on the Internet
How it Works
Works on a two-tier hierarchy, one Delegator (parent) and multiple Goals (children).
Goals can connect to ExoMind for factual answers, or to the internet for open responses, or to the API for customer-specific dynamic data.
Delegator: Acts as the central brain. It determines the correct “Goal” (sub-agent) for the user’s query.
Goals: Child nodes that define specific objectives, responses, and prompts. Each goal represents a micro-agent responsible for handling a particular topic or function.
Runtime Workflow
User asks a query - System checks NLP for a known intent flow.
If found → NLP flow executes.
If not found → passes to Gen AI Bot.
Gen AI evaluates the configuration:
If connected to ExoMind → retrieves relevant trained content.
If connected to the API → It will fetch the details from the API.
If not connected to either of them → It will rely on Internet-based reasoning.
LLM generates a response based on retrieved data and configured prompts.
Response returned to the user, optionally with source links (if ExoMind was used).
How to Enable
Request must be raised to help@exotel.com with the subject:
"Enable GenAI for <bot_name> - <customer name>".Add your Account Managers in cc.
How to Use
NLP to ExoMind/ GenAI Switch Works
When your chatbot has ExoMind or GenAI enabled with NLP flows, the system automatically manages how user queries are routed between different response engines.
This switch is a default behavior that ensures your chatbot can handle both structured (NLP-based) and unstructured (AI-driven) conversations seamlessly. It is an NLP first bot.
Understanding How the Query Handling Logic Works
Every user query follows a decision flow managed by the system in real time:
NLP check (first priority):
The chatbot first tries to match the query against its predefined NLP intents.
If a matching node is found, the corresponding flow or journey runs as usual.
This is typically used for transactional or rule-based use cases (e.g., Apply for Loan, Check Order Status).
If no NLP match is found, it will fall back to the LLM Intent Engine:
The system will run the query via the LLM Intent Engine to use the LLM-based model for understanding user intent in natural language
It tries to detect the user’s intent if not caught by the NLP Intent Engine.
If recognized → Trigger corresponding intent flow.
If not → Fallback to Bot Madate.
The Bot Mandate decides how to handle it:
If ExoMind is configured, it retrieves the most relevant chunks from trained data (documents or websites).
If ExoMind is not configured, the response is generated directly by the GenAI model based on its prompt setup or internet-enabled configuration.
AI response generation:
The retrieved content (if any) and prompts are passed to the LLM.
The LLM generates a natural, contextual response and returns it to the user.
User experience:
The user seamlessly receives an answer, regardless of whether it came from NLP, ExoMind, or GenAI.
No manual switching or setup is required - the routing happens automatically in the background.
Limitations
During an active NLP flow, Gen AI or ExoMind-based queries are temporarily disabled to maintain flow consistency.
Once a user exits an NLP flow, it cannot be resumed mid-way.
AI responses depend on the quality and completeness of trained ExoMind data.
Context from NLP flows (like variables or user inputs) is not carried over into GenAI responses.
Response time for ExoMind/GenAI answers may be slightly longer due to document retrieval and LLM processing (Same as you see on ChatGPT).
When both are active, NLP takes priority; GenAI only responds if NLP fails to detect an intent.
If ExoMind is not connected, GenAI may generate open-ended or hallucinated responses.
Flow Transition Limitations
Choosing the Right Bot Type
FAQs
General FAQs
ExoMind supports both document and website training.
Training time depends on file size (1–5MB = few minutes).
Documents can be retrained anytime.
Overlapping data is handled by LLM; use custom prompts for priority control.
Bot Mandate FAQs
Disabled by default as it changes routing.
GenAI can run without ExoMind, but using ExoMind improves factual accuracy.
Missing config will result in incomplete or failed responses.
GenAI Orchestrator FAQs
ExoMind can work independently without the Orchestrator.
When both NLP and GenAI detect intents, NLP is prioritized.
Technical / Integration FAQs
Check document encoding for failed training.
Upload only compliant, non-confidential data.
Only one Orchestrator per bot is allowed.
ExoMind may slightly impact response time due to external chunk retrieval.