AI chatbots for business and WhatsApp
A chatbot that answers from your documents, not from the internet. It qualifies enquiries, books meetings and hands over to a person the moment it gets serious.
✓ Answers grounded in your documents · Human handover · GDPR-compliant by design
Why most chatbots fail
And what we do differently.
The classic decision-tree bot asks you to "choose an option" and annoys everyone whose question isn't one of the four. The opposite extreme is a language model with no boundaries, politely inventing things your company never promised.
We build in between: the model may only answer from your own content — price list, service descriptions, delivery terms, frequently asked questions. When it can't find the answer there, it doesn't invent one. It says so and brings in a human.
A chatbot that gets it wrong one time in ten is worse than no chatbot at all. The boundaries matter more than the eloquence.
What it can take off your plate
On your website, in WhatsApp, or both.
- Questions about services, prices, lead times and terms — from your documents, not guessed.
- Qualifying enquiries: asking the right follow-up questions and writing structured answers into the CRM.
- Booking meetings, checking the real calendar.
- Answering order and job status once it's connected to your system.
- Catching out-of-hours enquiries that would otherwise sit until the next morning.
- Handing over to a person — with the conversation so far, so the customer doesn't repeat themselves.
WhatsApp: what to know before you start
So there are no surprises on the invoice.
WhatsApp Business isn't a free channel. Meta bills per conversation window, and for messages you start yourself you need pre-approved templates. Anyone offering you a WhatsApp bot without mentioning those costs either hasn't understood them or isn't telling you.
For inbound enquiries — the usual case — the cost stays modest and predictable. We give you the order of magnitude before quoting.
How the chatbot gets your knowledge
The technical part in plain terms, because it's what answer quality depends on.
We start by gathering whatever usable material you have: website, service descriptions, price lists, delivery terms, FAQs, internal manuals — sometimes simply the last five hundred answered customer emails. That becomes a searchable knowledge base.
When a question comes in, the system first finds the relevant passages in your material and hands the model only those extracts, with instructions to answer from them alone. If the search finds nothing suitable, there's no answer — there's a handover.
The most demanding part is usually not the technology. It's discovering that your documents contradict each other: the website quotes one lead time, the price list another, and sales says a third on the phone. The chatbot exposes that, because it has to pick one.
Several clients have got more out of that clean-up than out of the chatbot itself. It isn't a benefit we promise upfront, but it happens regularly.
How to tell whether it's working
Three numbers. You don't need more.
Resolution rate. The share of conversations that ended without a handover and where the person didn't then phone anyway. The second half of that sentence is the important one: a conversation the bot "resolved" that comes back as a call an hour later wasn't a success.
Qualified enquiries. How many contacts arrived in the CRM with complete details. That's the number that pays for the chatbot — minutes saved in support are pleasant, new enquiries are revenue.
Handovers with a reason. Not just how often it handed over, but for what. That list is next month's to-do: anything handed over three times belongs in the knowledge base.
We set this reporting up and review it with you during the first weeks. A chatbot nobody touches after launch doesn't improve.
Frequently asked questions
How does the chatbot know what to say?
From the material you give us: website, price list, service descriptions, FAQs, internal manuals. Only from that. Anything not in there doesn't get answered.
Can it promise something we can't deliver?
The risk can't be reduced to zero, so we restrict the sources, forbid commitments on price and dates without backing, and make it hand over when unsure. And we test it against your real enquiries before it goes live.
What about data protection?
We design the flow to be GDPR-compliant: a notice before the first exchange, defined retention periods and the option to keep personal data out of the model entirely. The final legal review belongs with your data protection officer, not with us.
Which questions do you get every day?
If the same five questions keep coming in, that's the best place to start. Tell us what they are.
Book a free call