AI chatbot costs: what should a small business budget for?
There is no useful fixed chatbot price without a scope. Budget separately for setup, recurring software and usage, and ongoing maintenance. A question-answering bot, a lead-capture bot and a bot that changes live records carry different requirements. Cervo Studio quotes after reviewing your information, integrations and approval rules.
An AI chatbot budget should describe what the system will do, what it will not do and who keeps it working. A headline build fee alone cannot tell you whether two proposals cover the same work.
This guide is a scoping checklist, not a Cervo Studio price list or an industry price survey. Costs depend on the agreed requirements and the providers selected.
Separate the three cost layers
Setup and implementation includes preparing information, designing the conversation, connecting systems and testing before launch. Content preparation can be substantial when answers are scattered across documents or are not yet written down.
Recurring software and usage can include the chatbot platform, model usage, hosting and connected services. Ask which accounts you will own, what the plan includes, what triggers additional charges and whether spending limits can be set.
Maintenance and support covers changes to the knowledge base, checking conversations, fixing integration issues and updating dependencies. Agree whether this is included, separately priced or handled by your own team.
Compare scope before comparing quotes
| Scope | Work to clarify | Question to ask |
|---|---|---|
| Answers questions | Approved sources, content updates, uncertainty handling | What happens when the answer is missing or outdated? |
| Captures enquiries | Required fields, validation, routing and consent wording | How does a person receive and follow up the request? |
| Takes actions | Authentication, permissions, integrations and recovery | Who approves a change, and what happens if it fails? |
These are illustrative categories, not fixed packages. A bot using a small knowledge base can still require careful testing if an incorrect answer carries a significant consequence.
The questions a useful quote should answer
- Which pages, documents and languages are included?
- Who prepares and approves the knowledge base?
- Which systems will be connected, and has access been checked?
- How are uncertain answers and requests for a person handled?
- What test cases must pass before launch?
- Which recurring charges are included, and which are paid directly?
- Who owns the accounts, configuration and customer data?
- What support, updates and exit arrangements are included?
For UK and French projects, also confirm the quote currency and how any applicable taxes are presented. Do not assume a provider's advertised subscription includes implementation or every integration.
When an existing product may be enough
A configurable product may suit a straightforward use case. Include your own setup, testing and content-maintenance time in the comparison. Check the current provider documentation and plan limits before committing; features and pricing can change.
Custom development may be appropriate where the conversation needs to connect with a specific workflow. It is not automatically more useful simply because it is custom.
Measure value without promising a payback
Record the current enquiry volume, handling time and common questions. During a pilot, track useful answers, handovers, incorrect responses and the time your team spends reviewing the system. Compare that with the full operating cost.
Do not count every automated answer as a saved sale or every minute as cash saved. The relevant outcome might be clearer enquiries or a more consistent handover.
Explore Cervo Studio's AI chatbot service, or describe the workflow you want to improve. If the problem is mostly behind the scenes, business automation may be a better starting point.