Chatbot
- Category
- Conversational AI
- Interaction
- Text, voice, multimodal
- Types
- Rule-based, retrieval, generative
- Common uses
- Customer service, HR, healthcare
- Platforms
- WhatsApp, web, Telegram, app
Chatbot
- Category
- Conversational AI
- Interaction
- Text, voice, multimodal
- Types
- Rule-based, retrieval, generative
- Common uses
- Customer service, HR, healthcare
- Platforms
- WhatsApp, web, Telegram, app
Types of Chatbots
Rule-based Chatbots
Use decision trees or pattern-matching to respond. Deterministic, predictable, and easy to audit. Suitable for well-defined workflows (FAQ, appointment booking, form filling) but brittle when users deviate from expected inputs.Retrieval-based Chatbots
Select responses from a curated database based on input similarity. More flexible than rule-based systems, but limited to pre-authored responses. Used in early customer service deployments.Generative Chatbots (LLM-powered)
Generate responses dynamically using large language models. Can handle open-ended questions, nuanced context, and multi-turn reasoning. Modern examples: ChatGPT, Claude, Gemini, Microsoft Copilot. Risks include hallucination and inconsistency.Hybrid Architectures
Production systems often combine rule-based intent detection (for routing and safety guardrails) with LLM generation (for flexible response creation). Retrieval-Augmented Generation (RAG) is used to ground chatbot responses in company-specific knowledge bases.Deployment Channels
- Web widget — embedded chat window on websites
- WhatsApp Business API — dominant in Malaysia for customer communication
- Telegram bots — popular for developer communities and fintech
- Microsoft Teams / Slack — enterprise internal use
- Mobile app — native SDK integration
- Voice — phone IVR, smart speakers
Key Metrics
| Metric | Description | |--------|-------------| | Containment rate | % of conversations resolved without human escalation | | CSAT | Customer satisfaction score post-conversation | | First response time | Time from user message to first bot reply | | Intent recognition accuracy | % of user intents correctly classified | | Hallucination rate | % of factually incorrect responses (LLM bots) |- Maybank — Virtual Advisor (Maya) on MAE app; handles balance inquiries, fund transfers, loan inquiries
- CIMB — EVA (Eva Virtual Assistant) on web and app
- RHB — RHB Banking Chatbot on website
- Bank Islam — BARI chatbot for Islamic banking products
- Weizenbaum, J. (1966). "ELIZA — A Computer Program for the Study of Natural Language Communication Between Man and Machine." CACM 9(1), 36–45.
- MDEC (2023). Conversational AI Adoption in Malaysian Enterprises. Malaysia Digital Economy Corporation.