How much does it cost to build an AI chatbot?
CodeSWI ·
At CodeSWI, AI chatbots typically range from $10,000 for a proof of concept to around $210,000 for an enterprise AI agent that acts in your systems. Where you land depends on your data, your integrations, and how carefully answers must be controlled.
An AI chatbot can be a prototype built in days or a production system connected to half your business. The price follows from four things: the data it answers from, the systems it connects to, how its answers are checked, and how many people use it.
What drives the cost
| Driver | Lower cost | Higher cost |
|---|---|---|
| Knowledge | A few documents or an FAQ | Many sources that change often and need syncing |
| Integrations | None, or one read-only system | Order systems, CRM, ticketing, with actions (refunds, bookings) |
| Answer control | Internal users, low risk | Customer-facing, regulated or high-stakes answers |
| Hand-off to humans | Not needed | Routing to agents with full conversation context |
| Channels | One web widget | Web, mobile, email, messaging apps |
| Languages | One | Several |
Example scopes
These are our typical price ranges, not quotes. Your scope sets the real price, agreed in writing before work starts.
| Scope | Typical timeline | Typical price (USD) |
|---|---|---|
| Proof of concept on your documents | 1–2 weeks | $10,000–$30,000 |
| LLM-based support assistant with help-desk hand-off | 8–14 weeks | $24,000–$60,000 |
| Enterprise AI agent that takes actions in your systems | 16–28 weeks | $72,000–$210,000 |
For market context: across AI projects reviewed on Clutch, the most common budget is $10,000–$49,999, while the average project costs about $120,600 and runs around 10 months, because a few large builds pull the average up.
Running costs after launch
Plan for roughly $400–$6,000 a month to run a production chatbot, most of it model usage. It breaks down into:
- Model usage: you pay the AI provider per request, so cost grows with conversations and answer length.
- Hosting for the application and any search index over your documents.
- Monitoring and improvement: reviewing conversations, fixing wrong answers, updating content.
How to keep cost down
- Start with one high-volume question type, measure it, then expand.
- Use the smallest model that answers well, and cache common answers.
- Keep humans in the loop for anything risky instead of automating it on day one.
At CodeSWI we usually start with a working prototype on your own documents, typically in 1–2 weeks, so you can judge answer quality before committing to a full build. Tell us your use case and we will scope it and give you a written estimate.
Sources
- Clutch — AI Development Pricing Guide (updated Sep 21, 2026) · accessed October 17, 2026
General information, not legal, financial or professional advice for your situation. Timelines and costs are illustrative estimates, not quotes or guarantees. Third-party figures are as published by their sources on the dates shown. See our terms of use.