A trained assistant on your site that answers real questions, qualifies enquiries and books calls at 3am — and knows when to stop and hand over to a person.
There’s nobody to ask, so they fill in a form and wait, or more often they don’t bother and go to the next result.
Meanwhile your team spends a real portion of every week answering the same twenty questions. Not complicated ones. The same ones.
Both problems have the same solution, and both are usually left alone because the available option was a chatbot that made things worse.
A good chatbot answers the question. Most of them are built to avoid answering it.
Worth reading before you commission one, because most of these are build decisions rather than limitations of the technology.
The classic failure: you ask a question, it offers three menu options that don’t match, then asks “did that help?” It’s built to reduce support tickets rather than to answer, and users learn within two exchanges to stop trying.
Language models will confidently state a price, a lead time or an availability that doesn’t exist. The fix isn’t hoping — it’s constraining the knowledge base so the bot answers from your actual content and says “let me get someone to confirm” when it doesn’t know. Ours are configured to prefer admitting uncertainty over guessing.
No visible way to reach a human is the fastest way to turn a warm enquiry into a bad review. Every conversation needs an obvious exit to a person, with the history carried across so nobody repeats themselves.
It doesn’t work, and being caught costs more trust than being upfront ever would. Ours say what they are.
A chatbot handles the routine. It shouldn’t be the only route to reaching you, and phone and email need to stay visible.
Complaints, pricing negotiations, anything emotionally loaded. These need a person, and a well-built bot recognises them and escalates immediately rather than trying.
We’d rather build something that answers eighty per cent of questions genuinely well and hands over cleanly on the rest, than something that appears to handle everything and frustrates people on a third of it.
Which questions your team answers repeatedly, and which are worth automating.
Trained on your actual content, scoped so it answers accurately or defers.
How it opens, qualifies, escalates and books — mapped before anything is built.
Connected to your CRM and calendar, so bookings and leads land properly.
Adversarial testing before launch, including the questions it should refuse.
Real conversations reviewed regularly, with gaps in the knowledge base closed.
Answers from your actual services, pages and documents, not generic knowledge.
Working out whether an enquiry is a fit before it reaches you.
Qualified prospects booking themselves straight into your diary.
Every conversation logged as a lead, with its source attached. to it.
A clear route to a person, with conversation history carried over.
Regular review of real chats, with knowledge gaps filled.
It probably isn’t right yet if your services aren’t clearly documented anywhere — the bot can only answer from what exists. And if you get five enquiries a month, a person answering them is cheaper and better. We’d say so.
This is usually the first phase-four project we’d recommend. It’s the smallest, the cheapest, and the one where the benefit is most immediately visible.
Swipe to compare
| Feature | Essential | Most Popular Growth | Advanced | Custom |
|---|---|---|---|---|
| Setup | £1,200 | £2,800 | £5,500 | From£10,000 |
| Scope | Website chatbot, FAQ, lead capture | + booking, CRM sync, qualification | + multi-step workflows, voice | Bespoke |
| Knowledge base | Up to 30 answers | Up to 100 | Unlimited | Unlimited |
Ongoing tuning and monitoring: £250–£750/month.
LLM usage: billed at cost, typically £10–£60/month depending on volume. Estimated before you commit.
The internal work behind the enquiries.
Where the leads and bookings need to land.
Only if built to deflect. Ours answer properly and hand over the moment they can't, with the conversation history attached so nobody has to repeat themselves.
It's constrained to your actual content and configured to defer rather than guess. We test adversarially before launch and review real conversations afterwards to close gaps.
Typically £10–£60 a month in LLM usage for a normal business volume, billed to you at cost with no markup. We estimate it before you commit.
Your data stays in your accounts. We'll tell you exactly which provider processes what before anything is built, so you can make an informed decision rather than a hopeful one.
Yes, from the Growth tier — connected to your calendar, with qualification first so you're not booking calls with people who aren't a fit.
Two to three weeks for most businesses. Most of that is building and testing the knowledge base, which is the part that determines whether it's any good.
Get a free automation audit. We’ll map which questions your team answers repeatedly, what a chatbot could genuinely handle, and what it shouldn’t — including when the honest answer is that you don’t need one.
We reply within one working day.
No spam, no pitch deck.
WhatsApp us