Generic chatbots damage trust because they guess. The moment a bot invents a price or promises a service you do not offer, you are worse off than having no bot at all.
We build assistants grounded in your real content: your service pages, pricing rules, FAQs, policies and past enquiries. When the answer is not in that material, the assistant says so and offers a person instead of improvising.
The other half of the job is commercial. An assistant should be capturing a name and a number, qualifying the enquiry, and putting it somewhere you will act on it. A conversation that ends without a contact detail is a missed lead.
What is included
Knowledge base build
We collect and structure the material the assistant is allowed to answer from, including the things it must never say.
Retrieval-grounded answers
Responses are drawn from your content rather than the model’s general knowledge, which is what keeps the answers factual.
Lead capture and routing
Name, contact and qualified detail captured mid-conversation and delivered to your inbox, CRM or admin dashboard.
Human handover
A clear escalation path with the conversation history attached, so the customer never repeats themselves.
Brand voice and interface
Styled to your site, written in your tone, and built to be accessible on mobile rather than a floating box that covers the content.
Transcript review
Every conversation logged so you can see what customers actually ask, which is frequently the most valuable output of the whole project.
How we work
- 01
Gather the source material
Service detail, pricing rules, coverage area, policies and the twenty questions you answer every week.
- 02
Define the boundaries
We agree in writing what the assistant may quote, what it must escalate, and what it must never say.
- 03
Build and test
We run the assistant against a written test set of real questions, including the awkward ones, before it goes near a customer.
- 04
Launch and tune
After launch we review transcripts weekly for the first month and close the gaps the real questions expose.
Where an assistant earns its keep
- After-hours enquiries that would otherwise wait until morning and go to a competitor
- Repetitive pre-sales questions about coverage, pricing bands, lead times and inclusions
- Booking and quote intake, where the assistant collects the detail a quote actually needs
- Internal use, answering staff questions about process and policy from your own documentation
Accuracy is a design decision
Grounding, refusal behaviour and escalation are not settings you switch on at the end. They shape how the assistant is built, and they are the difference between a tool your staff trust and one they quietly ask you to remove.
Frequently asked questions
Can it quote prices?
Only where the pricing is rule-based and you have signed off on the rules. Where pricing depends on a site visit or judgement, the assistant explains that and captures the enquiry instead.
What stops it making things up?
Answers are retrieved from your approved content, and the assistant is instructed and tested to say it does not know rather than fill a gap. We validate that behaviour with a written test set before launch.
Which channels can it run on?
The website is the usual home. We can also connect it to email intake and messaging platforms where there is a supported API.
Does it work with our existing website?
Yes. It is a script and an API, so it can sit on a Next.js site, WordPress, Squarespace, Webflow or a custom build.
How much does it cost to run?
Hosting plus model usage, typically modest for small-business conversation volumes. We show you the per-conversation cost during the build so you can forecast it.
Related work
Private Client
A bespoke AI feature built into a removals company’s quote form: the customer uploads a few photos of their rooms, and a vision-capable AI model estimates the total moving volume in cubic metres, itemises the goods it can see, and instantly sizes the right truck, replacing manual inventory counts and in-home surveys with a quote that takes seconds.
Read the case study →AI SolutionsAuction House
A bespoke AI-assisted description generator for a busy auction house, staff upload item photos, a vision-language model drafts a professional listing in seconds, and experienced staff review before publication, combining AI speed with human quality control.
Read the case study →Often paired with
Or see every service we offer.