Chatbot development cost: what a chatbot costs to build and run
Chatbot development cost is set by what the chatbot has to know and do: the content it answers from, the systems it must look things up in, the channels it works in, how it hands over to people, and how many conversations it will have. Unlike most software, an AI chatbot also has a running cost that grows with every conversation. This guide explains the build cost, the running cost, the work on content that people forget, the hidden costs, and how to compare chatbot pricing.
It does not quote our own prices. It explains how a chatbot budget is built, so any quote you receive, from us or from anyone else, can be compared on the same basis. For what we build, see chatbot development; for AI budgets in general, see artificial intelligence cost.
The short version
- Build: design, the conversation flows, connections to your systems, the hand-over to people, testing.
- Content: collecting, cleaning and structuring the material the chatbot answers from.
- Running: model usage per conversation, hosting, monitoring and a subscription if ready-made.
- Upkeep: reviewing answers, updating content and adjusting as products and policies change.
What drives chatbot development cost
| Driver | Why it adds work | How to keep it in check |
|---|---|---|
| Content | Answers are only as good as the pages, documents and policies behind them | Start with the questions people ask most; fix that content first |
| System look-ups | Order status, bookings or account data need secure connections and permissions | Read-only look-ups first; actions later |
| Channels | Website, messaging apps and internal chat each have their own interface and rules | Launch in the channel with most questions |
| Languages | Each language needs content and testing | Add languages where demand is proven |
| Hand-over to people | Passing a conversation to staff with its history needs integration with the help desk | Use the help desk you already have |
| Testing | A set of real questions with expected answers is needed to measure quality | Build the test set from past enquiries |
| Privacy and security | Personal data in conversations needs retention rules, access control and agreements with providers | Collect only what is needed; keep data in the EU |
Running costs: priced per conversation
Hosted language models are usually priced by the amount of text processed, both what is sent to the model and what it writes back. Each conversation therefore has a cost, which grows with the length of the conversation, the amount of your content included in each answer and the model chosen. Ask any supplier for an estimate of running cost per conversation at your expected volume, and for a monthly cap or alert. Providers publish their current model prices on their own pricing pages; check them on the day you compare, because they change.
Ways to keep running costs down: send only the relevant parts of your content with each question, route simple questions to smaller models, cache frequent answers, and keep a monthly budget alert.
Ready-made or custom chatbot
| Option | Where the money goes | Fits when |
|---|---|---|
| Feature in a tool you already use | A licence or add-on to an existing help desk or website platform | Answers come from help pages; standard hand-over is enough |
| Ready-made chatbot product | A subscription, often by conversations or seats, plus set-up | A website assistant with light integration needs |
| Custom chatbot | Build, then hosting, model usage and upkeep | Your own systems, rules, channels or data location matter |
A ready-made option is often the cheaper start, and a good supplier will say so. Compare over two to three years at your expected volume, including the content work, which costs the same either way.
Hidden costs of a chatbot
- Content clean-up. Outdated or conflicting pages give wrong answers, and fixing them is work.
- Review time. Someone has to read conversations each week and fix what went wrong.
- Usage growth. A successful chatbot has more conversations, and running costs follow.
- Model changes. Providers retire models; switching needs retesting.
- Compliance. Under the EU AI Act, systems that interact directly with people must let them know they are dealing with an AI unless that is obvious (Regulation (EU) 2024/1689, Article 50(1); Article 113 sets 2 August 2026 as the general date of application; read 12 November 2026). A data processing agreement is needed with every provider that handles conversation data (GDPR Article 28).
The people's time in a chatbot project
Building and testing a chatbot is labour, and labour costs follow location. Eurostat's release of 31 March 2026 puts average hourly labour costs across the whole economy in 2025 at €34.9 in the EU and €38.2 in the euro area, from €12.0 in Bulgaria to €56.8 in Luxembourg (read on 12 November 2026). Your own staff time matters as much: the people who know the answers are needed to write and check content.
How to compare chatbot pricing
- Which questions, systems and channels are included in the build?
- What is the estimated running cost per conversation, and at what volume?
- Which model is used, and can it be changed later?
- How is quality measured before launch, and how often after?
- How does hand-over to people work, and with which help desk?
- Where are conversations stored, for how long, and who can read them?
- Are code, prompts and the content index in your company's accounts?
- What is excluded, in writing?
When a chatbot becomes an agent
A chatbot answers and looks things up. When it starts to take actions, such as changing bookings, issuing refunds or updating records, it becomes an AI agent, with more work on permissions, approvals and testing. See AI agent development cost, and for the difference itself, chatbot vs AI agent. All our services are on the services page, and every guide is in the guides library.
Questions people ask
How much does a chatbot cost?
A chatbot cost has three parts: building it (or setting up a ready-made product), the content it answers from, and running it. Running costs for AI chatbots follow usage, because hosted language models charge for the amount of text processed. A website assistant answering from a few dozen pages and a chatbot that checks orders in your systems across several channels are very different budgets.
How much does a chatbot cost per month?
For a ready-made product, a subscription, often by number of conversations or seats. For a custom chatbot, hosting, model usage that grows with every conversation, monitoring and content upkeep. Ask for running costs per conversation at your expected volume, not only the build price.
Is a ready-made chatbot cheaper than a custom one?
At the start, usually. A ready-made chatbot is the right choice when it answers from your website and help pages and hands over to people in a standard way. Custom work pays off when the chatbot must use your own systems, follow your rules, work in your channels and keep data where you decide.
What makes chatbot running costs grow?
More conversations, longer conversations, larger documents included in each answer, more capable and more expensive models, and extra channels. Caching, shorter context, routing simple questions to smaller models and good content all keep the cost per conversation down.
Does the EU AI Act affect the cost of a chatbot?
It adds a small amount of work for most chatbots. Systems that interact directly with people must be designed so the people know they are dealing with an AI, unless that is obvious (AI Act, Article 50(1)). Chatbots used in areas the Act lists as high-risk need much more.
Sources
- Regulation (EU) 2024/1689 (AI Act), Articles 50 and 113 (read 12 November 2026)
- Regulation (EU) 2016/679 (GDPR), Article 28 (read 12 November 2026)
- Eurostat: EU hourly labour costs ranged from €12 to €57 in 2025 (31 March 2026, read 12 November 2026)
Your chatbot scope checklist
- The twenty questions people ask most, from past emails and tickets.
- The content that answers them, and who keeps it current.
- Systems to look things up in, read-only at first.
- Channels and languages for the first version.
- How and when a person takes over.
- Expected conversations per month.
- Personal data in conversations, and how long it is kept.
- How quality will be measured.
- A monthly running-cost budget.
- Ownership: code, prompts and content index in your company's accounts.