"How much does an AI chatbot cost?" is the first question most businesses ask, and the honest answer is: it depends. A simple FAQ bot and a chatbot connected to five internal systems are very different projects, even though both are called "chatbots".
Instead of giving you a number that would mislead you, this article explains what drives the cost, so you can estimate sensibly and compare quotes fairly.
1. The job the chatbot has to do
The biggest cost driver is scope. Compare these three:
- Answers questions from a fixed set of documents. The simplest and fastest to build.
- Answers questions and collects information such as leads or support tickets.
- Takes action, such as checking order status, booking appointments, or updating your CRM.
Each step adds integration work, testing, and safeguards.
2. Your data and knowledge sources
A chatbot is only as good as the information behind it. Costs rise when:
- Your content is spread across many places (website, PDFs, help centre, internal tools)
- Documents are messy, outdated, or contradict each other
- You need the bot to respect permissions, so different users see different information
Cleaning and organising the knowledge base is often the most underestimated part of the work, and the part that most improves quality.
3. Integrations with other systems
Connecting to a CRM, helpdesk, payment system, or booking tool takes engineering time. A single, well-documented API is quick. Several older systems with limited documentation take longer.
4. Channels
Where will people use it? A website widget is the simplest. WhatsApp, Instagram, Facebook Messenger, mobile apps, and voice each add their own setup, rules, and testing.
5. Languages and tone
Supporting more than one language, or matching a specific brand voice, needs additional testing and tuning. This matters especially if your customers write in a mix of languages.
6. Quality, safety, and handover
A production chatbot needs more than a model and a prompt:
- Guardrails to stop it from answering outside its scope
- Human handover when it cannot help
- Testing against real questions, including tricky ones
- Monitoring so you can see where it fails and improve it
Skipping these makes a project cheaper at first and costlier later.
7. Ongoing running costs
After launch there are usually two kinds of recurring cost:
- Usage costs. Language models are typically charged by how much text they process, so busier bots cost more to run.
- Maintenance. Your business changes, so the knowledge base, prompts, and integrations need updating.
Ask any provider to be clear about both.
How to budget realistically
- Define one use case. "Answer the top 30 customer questions on our website" is far easier to price than "add AI to our business".
- Start with a pilot. Launch to a small group, learn what real users ask, then expand.
- Compare like with like. When you get quotes, check that they cover the same scope, integrations, testing, and support.
- Plan for improvement. The first version is rarely the best. Budget a little for refinement after launch.
Get a clear estimate
If you want a realistic estimate for your case, book a free consultation. We will review your use case, data, and tools, and give you an honest scope and a clear plan rather than a vague number.
Learn more about our chatbot development service and our wider AI solutions.