What does an AI chatbot cost in the UK?
There are two broad ways to get an AI chatbot: subscribe to a platform and configure it yourself, or have one built. The ranges below are the ones SpiderHunts quotes for custom builds, as published on our AI chatbot development page. They exclude VAT, and every project gets a written quote with scope and exclusions agreed up front.
| Route | Typical build cost | Typical timescale | What you get |
|---|---|---|---|
| Platform subscription | Your own time, plus the plan fee | Days to weeks | A configurable bot on the vendor's platform |
| Focused custom chatbot | £5,000 to £15,000 | 2 to 4 weeks | One use case on one or two channels, trained on your content, with human handover |
| Multi-channel custom chatbot | £15,000 to £40,000 | 6 to 8 weeks | Web, WhatsApp, Teams, Slack or SMS from one core, with CRM, helpdesk or store integrations |
Platform plan fees vary widely by vendor and change often; our comparison of conversational AI platforms links to each vendor's current pricing.
What drives the build cost up or down?
Two chatbots that look similar in a demo can differ several times over in cost. The differences are almost always in these areas:
- Number of jobs. Answering FAQs is one job. Answering FAQs, qualifying leads and booking appointments is three, each with its own logic and testing.
- Integrations. Reading from and writing to a CRM, booking system, helpdesk or store is often the largest single cost, and depends on how good those systems' APIs are.
- Actions versus answers. A bot that only answers is simpler than one that changes bookings, raises tickets or updates records, which needs verification and safeguards.
- Channels. Each channel, such as website, WhatsApp, Teams or SMS, has its own set-up, rules and testing.
- State of your content. Tidy, current documentation is quick to index. Contradictory or scattered content needs work before the bot can answer from it.
- Data requirements. UK or EU processing, redaction of personal data before it reaches a model, or strict retention rules add design and build time.
- Languages and voice. Several languages, or a phone channel with speech, add testing and tuning.
- Testing depth. A bot in a regulated or sensitive setting needs a larger set of test questions and more review before launch.
What are the running costs of an AI chatbot?
The build is not the whole bill. Budget for these running costs, which we estimate in every quote rather than leaving as a surprise later:
- Model usage. Each conversation calls a language model, which is billed by the provider, usually by the volume of text processed. It is typically billed to your own account, and it rises with the number and length of conversations and the model chosen.
- Hosting. The application, the search index over your content and the conversation logs need hosting, usually a modest fixed monthly cost.
- Channel fees. WhatsApp business messaging has Meta's own charges for some message types, and SMS and phone channels have per-message or per-minute costs.
- Maintenance. Content updates, reviewing transcripts, adjusting prompts and keeping integrations working when other systems change. You can do this in-house or through an optional retainer.
We do not quote generic monthly figures here, because they depend so heavily on your volumes and choices that any single number would mislead. A good supplier should estimate them from your expected conversation volume before you commit.
Why does a platform cost less than a custom build?
A platform spreads its development cost across thousands of customers and leaves the configuration to you. That is excellent value when your needs match what the platform does well: answering from a help centre on one channel. You pay with your own time and with per-conversation or per-resolution fees as volume grows.
A custom build costs more up front because it is shaped around your systems and rules. It earns that cost when answers depend on live data in your own systems, when the bot must take actions, when you need several channels from one knowledge base, or when per-resolution pricing would add up at your volume. If none of those apply, a platform is probably the right choice, and we will tell you so.
How can you keep the cost of an AI chatbot down?
- Start with one use case and one channel, and add more once it works.
- Tidy your content first: one current answer for each question, not three conflicting ones.
- Use integrations that already exist before commissioning custom ones.
- Choose the smallest model that answers your test questions accurately; larger is not always better.
- Agree what goes to a person, so the bot does not need to handle every edge case.
- Review transcripts monthly so small content gaps do not turn into rework.
What should a chatbot quote include?
A useful quote is specific. Look for:
- The use cases and channels in scope, and what is explicitly excluded.
- Each integration, and whether it reads data, writes data or both.
- Who prepares the content, and how it is kept current after launch.
- Estimated running costs at your expected volume, and who pays each provider.
- How the handover to a person works.
- How data is processed and stored, with a data processing agreement where needed.
- Who owns the code, prompts and conversation data.
- What support after launch costs, if you want it.
If a quote is a single figure with none of that, ask for the detail before comparing it with anything else.
Is an AI chatbot worth the cost?
It depends on what it replaces. The honest way to judge is to estimate how many conversations it could resolve without a person, what those conversations currently cost in staff time or missed enquiries, and compare that with the build and running costs. Our AI automation ROI calculator helps with the arithmetic using your own figures. If the numbers only work with optimistic assumptions, start smaller or wait. If you want a bot that also works on WhatsApp, factor in the channel fees from the start.