Chatbots

Build conversational chatbots powered by AI for customer support, FAQ, and knowledge retrieval.

5 min readAI Agents

Chatbots

A Chatbot is a conversational AI assistant optimized for answering questions and providing information. Unlike full agents that can plan and execute multi-step tasks, chatbots are focused on dialogue — they respond to user messages using their knowledge base and configured personality.

Chatbot vs Agent

ThinkFleet supports two types of AI assistants:

Feature Chatbot Agent
Primary purpose Answer questions, provide information Plan and execute multi-step tasks
Tools Optional (knowledge search, basic lookups) Full tool access (flows, MCP, APIs)
Memory Conversation history Conversation + semantic long-term memory
Autonomy Responds to messages Can delegate, reflect, and chain actions
Best for FAQ, support, onboarding Workflow automation, data processing, research

Choose a Chatbot when you need a conversational interface for answering questions. Choose an Agent when you need autonomous task execution.

Creating a Chatbot

  1. Navigate to Chatbots in the sidebar
  2. Click New Chatbot
  3. Configure the basics:

System Prompt

The system prompt defines your chatbot's personality, tone, and boundaries. A good system prompt includes:

  • Role — Who the chatbot is (e.g., "You are a friendly customer support assistant for Acme Corp")
  • Tone — How it should communicate (e.g., "Be concise and professional")
  • Boundaries — What it should not do (e.g., "Never make up information. If you don't know, say so.")
  • Knowledge context — What it has access to (e.g., "You have access to our product documentation and FAQ")
You are a helpful support assistant for Acme Corp.

Your job is to answer customer questions about our products and services.
Use the knowledge base to find accurate information. If you can't find
an answer, politely suggest the customer contact support@acme.com.

Be friendly, concise, and professional. Never make up product features
or pricing that isn't in the knowledge base.

Knowledge Base

Attach one or more knowledge bases to give your chatbot access to your documents. When a user asks a question, the chatbot automatically searches the knowledge base using semantic search (RAG) and includes relevant passages in its response.

See Knowledge Base for details on uploading and managing documents.

Model Selection

Choose which AI model powers your chatbot:

Model Best For
Claude Haiku Fast responses, simple Q&A, high volume
Claude Sonnet Balanced speed and quality
Claude Opus Complex reasoning, nuanced responses
GPT-4o Alternative provider option
GPT-4o Mini Fast, cost-effective alternative

Deployment Options

Embedded Widget

Add a chat widget to your website with a simple script tag:

<script
  src="https://your-instance.thinkfleet.com/chatbot-widget.js"
  data-chatbot-id="your-chatbot-id"
  data-project-id="your-project-id"
></script>

The widget appears as a floating chat button in the bottom-right corner. You can customize colors and position.

API

Integrate the chatbot into your own UI using the REST API:

POST /v1/projects/{projectId}/chatbots/{chatbotId}/chat
Content-Type: application/json

{
  "message": "What are your business hours?",
  "sessionId": "user-session-123"
}

The sessionId maintains conversation context across messages. Use a unique session per user or conversation.

Share a direct link to the chatbot's hosted interface:

https://your-instance.thinkfleet.com/chat/{chatbotId}

Conversation Management

Session History

Each conversation is stored as a session. You can view all sessions in the chatbot's Conversations tab:

  • Full message history
  • User satisfaction signals
  • Token usage per session
  • Session duration

Handoff to Human

Configure escalation rules to hand off conversations to a human agent when:

  • The chatbot can't answer a question after multiple attempts
  • The user explicitly requests a human
  • Sentiment analysis detects frustration

Customization

Persona

Assign a Persona to your chatbot for a pre-configured personality and system prompt. ThinkFleet includes 20+ built-in personas:

  • General Assistant — Friendly all-purpose helper
  • Customer Support — Professional support agent
  • Technical Writer — Documentation-focused
  • Sales Assistant — Product-focused conversationalist
  • Onboarding Guide — Step-by-step walkthroughs

You can also create custom personas with your own system prompts and tool configurations.

Greeting Message

Set a greeting message that appears when a user starts a new conversation:

Hi! I'm the Acme Support Bot. I can help you with:
• Product questions
• Account issues
• Billing inquiries

How can I help you today?

Analytics

Track chatbot performance in the Analytics tab:

Metric Description
Total conversations Number of chat sessions
Messages per session Average conversation length
Resolution rate Percentage of conversations resolved without escalation
Response time Average time to generate a response
Token usage Total tokens consumed

Next Steps