Chatbot: 24/7 customer support

10 years ago chatbots were rule-based: 'if user types X, reply Y.' Result: outside the limited set, the chatbot became useless.

Modern chatbots are LLM-based (Claude, GPT) with natural-language understanding and response. Company-specific knowledge integrated via RAG makes them yours.

Through RAG, the chatbot reads your FAQ docs, product catalog, manuals, pricing pages. When a user asks, it finds the relevant pieces and the LLM synthesizes a reply.

Sources can be diverse: Notion, Google Docs, Confluence, your CMS, PDFs, web pages. We push them through ETL to a vector database; the chatbot searches in real time.

A well-designed chatbot knows its limits. Complex topics, frustration signals, critical decisions — auto-escalate and hand off to a human agent.

On handoff, the full conversation history goes to the agent; they don't have to 'start over.' Improves customer satisfaction and agent productivity.

Chatbot development cost: fixed-scope dev + runtime LLM API usage (per message). A typical B2C chatbot at 10K–100K messages/month is in the $100–$1,000 LLM-cost range.

Prompt caching (Anthropic feature) and template-based fallback (ready answers for simple questions) reduce cost by 50%+. Part of the standard package.