AI chatbot developmentA chatbot that answers like your best employee, and knows when to fetch one.
Most chatbots are an FAQ page with a typing animation. Ours is grounded in your own pricing, policies and documents, scoped on what it may say and do before a single prompt is written, and tested against real questions before it meets a customer.
/ In plain terms
What is an AI chatbot?
An AI chatbot is a chat assistant on a website or app that uses a large language model to answer questions in natural language, drawing on the business's own content instead of a fixed script. A well-built one answers common questions, collects the details of a lead, books an appointment, and passes the conversation to a person when it reaches the limit of what it should handle.
/ Not ready to book a call?
Get a free teardown of your website first.
Send us your address and we email you the three changes we would make first. Specific, free, no pitch attached.
/ The tell
Find the sentence that describes your week.
The things you would already have noticed if an AI chatbot applied to you.
- The same ten questions arrive by email every week and each one costs someone ten minutes.
- Enquiries that come in after hours wait until morning, and some have gone elsewhere by then.
- You tried an off-the-shelf bot and it confidently made things up.
- Your contact form collects a name and an email and nothing useful about the job.
- Support and sales answers live in documents only two people know how to find.
/ What lands in your hands
What you actually receive.
Concrete things, not a statement of intent. This is what is included in an AI chatbot build.
Knowledge base grounding
The bot answers from your own content using retrieval, and says it does not know when the answer is not there.
Lead qualification
It asks the questions your sales call would ask, then sends a structured summary to your inbox or CRM.
Booking and actions
Calendar booking, order lookups or ticket creation through your existing tools, with permissions scoped per action.
Human hand-off
A clear route to a person, with the conversation so far attached, whenever the visitor asks or the bot is unsure.
Evaluation set
A test set of real questions with expected answers, run before every change so quality is measured, not assumed.
Cost and guardrails
A per-conversation cost budget, rate limits, and rules for the topics the bot must decline.
/ Proof
Work you can open, not a promise.
Products we have shipped with a language model doing real work inside them: a conversational property assistant, a resume evaluator and an AI workspace.
/ How it runs
Four stages. No surprises.
3 to 8 weeks depending on integrations. You get a fixed timeline and a fixed price in writing before any code is written.
- 01
Scope the job
Week 1We agree what the bot answers, what it may do, and what it must pass to a person. This is written down before any build.
- 02
Ground and build
Weeks 1 to 3Your content is indexed, the conversation flows are built, and the integrations to calendar, CRM or helpdesk are connected.
- 03
Test against real questions
Weeks 3 to 4We run the evaluation set, fix wrong or vague answers, and try to break it before your customers do.
- 04
Launch and tune
OngoingConversations are reviewed after launch. Questions it could not answer become new content or new rules.
/ Before you buy
Rule-based bot, AI chatbot or AI agent?
They are sold under the same name and do very different jobs. Pick by what the conversation needs to end with.
| Criteria | Rule-based bot | AI chatbot | AI agent |
|---|---|---|---|
| How it answers | Fixed buttons and scripted replies | Natural language, from your content | Natural language, and it takes actions |
| Handles unexpected questions | No | Yes, within its knowledge base | Yes, within its tools and permissions |
| Typical job | Routing and simple FAQs | Support answers and lead capture | Bookings, updates and multi-step tasks |
| Main risk | Frustrating dead ends | Wrong answers if not grounded | Wrong actions if permissions are loose |
/ The awkward questions
The ones you were going to ask anyway.
Answered here so you do not have to spend a call on them.
AI Chatbot Development FAQ · reviewed
How much does it cost to build an AI chatbot?
Projects start from $1,000. There are two costs: the build, and the running cost of the model. The build depends on how many systems the bot connects to and how much content it has to learn. Running costs are charged per conversation by the model provider and are usually small for a typical business site. We give a fixed build price and an estimated monthly running cost after the discovery call.
Will the chatbot make things up?
A model with no grounding will. We reduce that by making the bot answer only from your content, instructing it to say when it does not know, and testing it against a set of real questions before launch. No AI system is perfect, so anything with consequences, such as pricing commitments or medical and legal topics, is routed to a person.
Can it connect to my CRM, calendar or helpdesk?
Yes. Common integrations include GoHighLevel, Cal.com, Slack and email. If your tool has an API or webhooks, the chatbot can usually send leads to it or read from it.
Which AI model do you use?
We choose per project between models from OpenAI and Anthropic based on answer quality, speed and cost for your use case. The bot is built so the model can be swapped later without rebuilding it.
How is a custom chatbot different from a plug-in widget?
A plug-in widget is quicker and cheaper to start, and for a simple FAQ it can be the right choice. A custom build makes sense when the bot needs to take actions in your systems, follow specific rules, match your product closely, or when you need control over where conversation data is stored.
What happens to our customers' conversation data?
We tell you exactly which providers process it and what they retain. Conversations can be stored in your own database, and we configure providers so your data is not used for training where that option exists.
/ Notes from the foundry
Further reading
- AI chatbot vs live chat vs contact form: which should you use?A contact form, live chat and an AI chatbot solve different problems. Here is how to choose, and when the honest answer is to use two of them.
- AI integration for small business: where to actually startNot with a chatbot. Start with the task your team already dreads doing manually every single week.
- AI features that earn their keepMost AI features ship as a demo and die as a product. The difference is almost never the model — it's the integration.
/ Same team, different problem
Other services
Still not sure this is the one you need?
Describe the business problem rather than the technical solution. Mapping it to the right service is our job, and the call is free whether or not we end up working together.
Book a discovery call→