AI chatbots connected to your real data, not a script

Chatbot vs. "AI agent" — what's actually different

A chatbot is conversation-first: it lives in one channel, answers questions, and maybe books or qualifies. An agent is broader — it can call multiple tools, take multi-step actions, and run server-side without a chat window at all (the MCP calculators on this site are an example — no conversation UI involved). Most businesses actually want a chatbot: a focused, single-channel assistant. This page is about that narrower, usually cheaper build.

Where a chatbot earns its cost back fastest

Repetitive, high-volume questions — "where's my order," "what are your hours," "how do I book" — are the highest-leverage target, because the manual version of answering them is real, countable time. See the customer support agent use case for the full pattern (knowledge base + CRM/helpdesk lookup + human escalation), or the sales qualification agent if the goal is converting website traffic into booked calls instead of answering support questions. For the internal-employee version of this same pattern — IT tickets, access requests, password resets instead of customers — see service desk automation. And if what's actually needed is a broader internal assistant employees ask general questions to, not a single-purpose bot, that's the same RAG pattern at a wider scope — see RAG development, including a live example running on this site.

Built on Claude or GPT, your choice

No lock-in to a single model vendor — the underlying LLM is a configuration choice, not an architectural one, and switching later doesn't mean rebuilding the integration work (the CRM lookup, the knowledge-base search, the handoff logic) that actually took the time to build.

Cost and timeline

A single-channel chatbot runs $1,150–$2,300 and 2-3 weeks; adding a second channel or a CRM/helpdesk lookup moves that toward $4,800. See the full cost breakdown, or run the calculator with your actual channels.