AI chatbot development, for assistants that understand what the visitor wants
Most business chatbots fail in one of two ways. The old kind is a menu: fixed flows, buttons, canned replies, and no idea what the visitor is actually trying to do. The new kind is the opposite problem: a general model wired to a website that answers from information the business never gave it, invents prices, and sounds confident when the right business answer is unknown. We build the thing in between. A controlled layer around the model that knows your business, detects what the visitor wants, answers only from what you have approved, says it does not know when it does not, captures a lead when there is one, and hands to a person when it should.
Who this is for
- - A business whose website chatbot collects contact details but cannot tell a buyer from a browser, and loses the buyer.
- - A SaaS company that wants a support assistant on its own documentation and product data, answering accurately and escalating what it cannot handle.
- - A company that connected a general model to its site, got burned by a made up answer, and now needs it done properly.
- - An internal team that wants a private assistant over company knowledge for employees, with no lead capture and strict permissions.
What you get
- - Intent detection: the assistant works out whether the visitor is researching, comparing, buying or stuck, and behaves differently for each.
- - Answers grounded in your approved sources only: website, documents, product data, policies. It declines rather than invents.
- - Lead capture that triggers on buying signals, not on every visitor, with the fields your sales team actually uses.
- - A clean handover to a human, with the conversation and the detected intent attached, for the cases the assistant should not handle.
- - A test set of real conversations run before every change, and a monthly report of what it answered, declined and escalated.
What it costs
There is no honest fixed price before discovery, and anyone who gives you one is guessing. What we can say in writing after two weeks: the build cost, the running cost per month at your expected volume, and what moves each. The build is driven by how many systems the work must touch and how strict the accuracy, latency and compliance bars are. The running cost is model calls plus infrastructure, and it can differ by ten times depending on the choices made in week one. How we price AI work.
How long it takes
Discovery, two weeks: we map the problem, the systems involved and the success criteria, and put the numbers in writing. Pilot, four to six weeks: one capability, live, on a slice of real traffic, measured against those criteria. Then production hardening and scale out. Each stage is a separate agreement with no minimum commitment, so you can stop after discovery with a plan you own. How we work.
What you own afterwards
Everything. The code, the prompts, the evaluation set, the infrastructure configuration and the documentation, in your accounts, under your name. We do not hold anything hostage and we do not license our own platform to you. The evaluation set matters most: it is the thing that lets your team, or any other team, change the system later without breaking it. The deliverables clause.
A chatbot that had to get intent right
A US automotive SaaS product had an assistant that needed to understand customer intent, identify leads from conversations and search a very large knowledge base. It struggled with intent accuracy and took 10 to 30 seconds to answer. We rebuilt it around dynamic intent detection, lead generation logic and authorization aware retrieval. First token in under three seconds in typical interactions and about 98 percent accuracy across intent recognition and retrieval in the client’s evaluated use cases.
Read how it was doneQuestions buyers ask
What does an AI chatbot development company do differently from a chatbot builder tool?
A builder tool gives you a flow or a general model and leaves the hard parts to you: what the assistant may say, where its answers come from, when it should capture a lead, and when it should stop and hand over. We build those parts. The result is an assistant that answers only from your approved knowledge, detects intent, and escalates correctly, running on your site and owned by you.
How much does custom AI chatbot development cost?
The build is driven by how many sources the assistant must answer from, how many intents it must handle, and whether it must act in other systems such as a CRM or booking tool. The running cost is model calls per conversation and is usually small for a well scoped assistant. After two weeks of discovery both go in writing.
How do you stop the chatbot making things up?
By grounding every answer in sources you have approved and requiring the assistant to decline when the answer is not there, rather than by filtering its output afterwards. Output filters are the weakest control; narrowing what the assistant may draw on is the strongest. We test it with the questions it should refuse as well as the ones it should answer.
Can the chatbot capture leads and hand over to a human?
Yes, and both are designed rather than bolted on. Lead capture triggers on buying signals the assistant detects, with the fields your sales team uses. Handover passes the conversation and the detected intent to a person, so the visitor does not repeat themselves. Which cases hand over is agreed with you before the build.
Before you decide
If you want the technical background first, read how conversational AI systems work. Then, from our engineers:
Tell us what your visitors ask
Two weeks of discovery puts the cost, the timeline and the plan in writing before you commit to anything. Or start with the free audit.