Related: n8n vs Zapier vs Make: Which Automation Tool Is Right for Your Business in 2026?
An AI chatbot can answer customer questions 24/7, qualify leads, and take routine work off your team. But "add a chatbot" can mean anything from a $20/month widget to a custom assistant that reads your documents and books appointments. Here is how to choose, what it really costs, and how the good ones actually work.
Key takeaways
- Off-the-shelf widgets are cheap and fast but generic; custom bots are grounded in your data and brand.
- The thing that makes a bot useful is retrieval (RAG) — it answers from your content, not the open internet.
- Custom AI chatbots typically cost $1,500–$10,000 to build, plus small monthly usage fees.
- Always insist on guardrails and human handoff so the bot never invents answers or traps a frustrated customer.
- Start with one clear job (support deflection or lead capture), measure it, then expand.
Off-the-shelf vs. custom
| Off-the-shelf widget | Custom AI chatbot | |
|---|---|---|
| Setup | Minutes | 2–4 weeks |
| Cost | $20–$200/month | $1,500–$10,000 + usage |
| Answers from your data | Limited | Yes (RAG over your docs) |
| Takes actions (book, look up, route) | Rarely | Yes |
| Brand & control | Generic | Fully yours |
If you just need a smarter FAQ, a widget is fine. If the bot needs to know your products, policies, and pricing — and act on them — you want a custom build.
How a good AI chatbot actually works
The best business chatbots use retrieval-augmented generation (RAG): your documents, product info, and policies are indexed, and when a customer asks something, the bot retrieves the relevant pieces and answers from them. That is the difference between a bot that confidently makes things up and one that gives accurate, on-brand answers with sources. Add guardrails (so it stays on topic) and human handoff (so hard cases reach your team with context), and you have something customers actually trust.
What it costs
| Bot type | Build cost | Monthly usage |
|---|---|---|
| FAQ / support bot on your content | $1,500 – $3,500 | $20 – $100 |
| Lead-qualification assistant | $2,500 – $6,000 | $30 – $150 |
| Knowledge assistant (RAG over docs) | $4,000 – $8,000 | $50 – $300 |
| Agent that takes actions | $7,000 – $15,000+ | varies |
The build is the bigger number; the ongoing model/API usage is usually modest for a small business.
What to look for (and common mistakes)
Look for: grounding in your own data, a clear handoff to humans, conversation logging so you can improve it, and control over tone. Common mistakes: letting the bot answer from the open internet (it will hallucinate), no escalation path (frustrated customers get stuck), and "set it and forget it" (the best bots are tuned from real conversations).
What it costs and how long it takes
A focused, genuinely useful chatbot is usually live in 2–4 weeks for $1,500–$8,000, depending on how much of your data and how many actions it handles. Start small with one measurable goal and grow it.
FAQ
Which AI model is best for a chatbot?
It depends on the job. We use the best fit — OpenAI's GPT, Anthropic's Claude, or open models — often combined with retrieval over your data. The model matters less than how well it is grounded and guarded.
Will it use my own content?
A custom bot should. We connect it to your docs, site, or systems so answers come from your content, with a human handoff for anything it should not answer.
Can you build and host it for me?
Yes — design, build, integrate, and maintain. Get a free quote and we'll scope it to your use case.
Working with Apex Logic
We build custom AI chatbots and LLM integrations grounded in your data, with guardrails and human handoff built in. We start with one clear goal and a fixed quote. Tell us what you'd want your chatbot to do.
References
OpenAI & Anthropic documentation (2026) — model capabilities and usage pricing.
Apex Logic project data (2024–2026) — chatbot build and delivery ranges.
Comments