In 2027, AI chatbot development cost typically ranges from US$10,000 to US$40,000 for a FAQ or knowledge bot built on a low-code platform, US$40,000 to US$120,000 for a retrieval-augmented (RAG) assistant that answers from your own documents with single sign-on, and US$100,000 to US$300,000 or more for a transactional assistant that integrates with CRM, ticketing or order systems, takes actions and hands off to people across several channels. Running costs, mainly platform licences or model usage, search, hosting and content upkeep, usually add US$300 to US$10,000 a month. The best way to compare options is cost per resolved conversation, not build price alone. These are indicative planning ranges, not a quote.
Key takeaways
- Choose the chatbot tier by what it must do: answer, answer from your content, or complete tasks.
- Integrations and actions, not the language model, drive most of the cost of advanced chatbots.
- Content quality and upkeep decide answer quality; budget for both.
- Low-code platforms are fastest for simple bots; custom builds suit complex, high-volume or branded experiences.
- Measure cost per resolved conversation and containment rate from the pilot.
AI chatbot development cost by type
| Chatbot type | What it does | Indicative build cost (USD) | Typical timeline |
|---|---|---|---|
| FAQ or knowledge bot | Answers common questions from a curated knowledge base or website | $10,000 – $40,000 | 3 – 6 weeks |
| RAG assistant | Answers from internal documents and systems with citations and permissions | $40,000 – $120,000 | 2 – 4 months |
| Transactional assistant | Looks up and updates records, books, orders or resolves cases, with human handoff | $100,000 – $300,000+ | 4 – 8 months |
For how RAG assistants work, see RAG architecture for an enterprise AI chatbot. When a chatbot starts planning steps and calling tools, it becomes an agent; see AI agent development cost.
AI chatbot cost drivers
- Integrations and actions. Each system the chatbot reads from or writes to adds design, security and testing work.
- Content. Volume, quality and permissions of the knowledge it answers from. See AI-ready data.
- Channels. Website, Microsoft Teams, WhatsApp, mobile app and voice each need adaptation.
- Languages. Each language needs testing and sometimes separate content.
- Human handoff. Passing context to live agents in your contact centre platform.
- Compliance and security. Data protection, logging and safeguards against prompt injection, guided by the OWASP Top 10 for LLM Applications.
- Analytics. Dashboards for containment, satisfaction and failure topics.
Low-code platform or custom build?
| Factor | Low-code platform (e.g. Copilot Studio) | Custom build on model APIs |
|---|---|---|
| Speed | Fastest for standard scenarios | Slower to first release |
| Cost model | Licences and message capacity | Build cost plus model usage |
| Integrations | Connectors for Microsoft and common SaaS | Any system, any depth |
| Experience | Platform patterns | Fully branded and tailored |
| Scale economics | Can become costly at high volume | Often cheaper per conversation at scale |
Compare the options in Copilot Studio vs Azure AI Foundry and check current terms in Copilot Studio licensing.
Running costs
| Cost line | Indicative monthly range (USD) |
|---|---|
| Platform licences or model usage | $100 – $6,000 |
| Search index and hosting | $100 – $2,000 |
| Monitoring and analytics | $50 – $500 |
| Content upkeep and improvement | Usually 10 – 20% of build per year |
Model usage depends on conversation volume and length; see Azure OpenAI pricing and our guide to RAG cost optimization.
A worked AI chatbot estimate
Illustrative, fictional example: a support assistant for a software company answering from product documentation and checking ticket status in the help desk, on the website and in the app.
| Line | Assumption | Amount (USD) |
|---|---|---|
| Build | RAG over documentation, help desk lookup, handoff to agents, two channels | $85,000 |
| Running cost | 20,000 conversations a month at about $0.08 each, plus $900 hosting and search | $30,000 a year |
| Content and improvement | 15% of build | $12,750 |
| First-year total | $127,750 |
If the assistant resolves 40 percent of conversations without an agent, compare that with your cost per human-handled ticket to estimate return.
Measuring chatbot return
- Containment rate: conversations resolved without a human.
- Cost per resolved conversation: running cost divided by resolved conversations.
- Customer satisfaction: ratings after the conversation.
- Answer accuracy: sampled review against sources.
- Escalation quality: whether handoffs arrive with useful context.
Common questions
How much does it cost to build an AI chatbot in 2027?
Indicatively, US$10,000 to US$40,000 for a FAQ bot, US$40,000 to US$120,000 for a RAG assistant over your documents, and US$100,000 to US$300,000 or more for a transactional assistant with integrations and actions.
What is the monthly cost of running an AI chatbot?
Typically US$300 to US$10,000 a month for licences or model usage, search, hosting and monitoring, depending on volume, plus content upkeep.
Is a low-code chatbot platform cheaper?
For simple bots and Microsoft-centred organisations, often yes. At high volume or with complex integrations, a custom build can cost less per conversation.
How long does it take to build an AI chatbot?
Three to six weeks for a FAQ bot, two to four months for a RAG assistant and four to eight months for a transactional assistant.
How do we keep chatbot answers accurate?
Use well-maintained content with permissions, retrieval with citations, an evaluation set of real questions and regular review of failed conversations.
Get a chatbot estimate for your use case
Share the questions or tasks you want automated, your channels and your systems. Our conversational AI integration team will recommend the right tier and platform and estimate build and running costs.