Not a chatbot for the sake of it. The two or three jobs where it saves real hours.
Answering the same questions at night, drafting quotes from a form, sorting the inbox. We find the repetitive work in your business, build the assistant for it, and make sure a person is always one step away.
Six jobs where this actually pays for itself.
If none of these sound like your week, you do not need AI yet, and we will tell you that.
"We answer the same five questions all day."
Opening hours, availability, prices, parking, dogs. That is an assistant job."Enquiries come in at night and go cold."
An answer within a minute at 22:00 is worth more than a perfect one on Tuesday."Making a quote takes an evening."
If the inputs are in a form, the first draft can be waiting for you in the morning."The inbox is the bottleneck."
Sorting, tagging and drafting replies is the least creative work in the company."We tried ChatGPT and it made things up."
Because it was guessing. An assistant on your own data does not have to."We are not sure this is even allowed."
It is, with disclosure and documentation. That is part of the build, not an afterthought.Included, and not included. Written down.
Every build comes with the paperwork the AI Act asks for.
Always included
- A half-day mapping where the repetitive hours actually go
- One assistant built on your own content: prices, availability, policies, FAQs
- Escalation rules, so a person takes over the moment it matters
- EU hosting and a data-processing agreement you can show a client
- Disclosure text on the site, in your tone
- AI Act documentation: purpose, data, limits, human oversight
- A month of tuning after go-live, based on real conversations
Not included, quoted separately
- Model usage after the first month, billed at cost
- Training a model of your own, which almost no small or medium-sized business needs
- Automating a process nobody has written down yet
- Anything that decides about people: hiring, credit, insurance
- Content generation at scale, which we would rather not build
Six weeks from half-day to live.
Most of the work is in your data and your escalation rules, not the model.
Half-day map
Where the hours go, and which two or three jobs are worth automating at all.
Data and rules
What the assistant may see, what it must never answer, and when it hands over.
Build
The assistant, the connections to your calendar or price list, the escalation path.
Test with real questions
We replay your actual inbox at it and fix what it gets wrong.
Live, with the paperwork
Disclosure on the site, AI Act documentation in your file, a month of tuning.
Three starting points. Your quote is one number.
Every job is quoted as one fixed amount for a defined result, written down before we start. Tell us which of these sounds like you and we put a number on it.
One assistant
One job: answering questions on your site from your own content.
- Everything in "always included"
- Live in 4 weeks
Assistant plus systems
The assistant reads your calendar or price list and writes into your CRM.
- Everything in One assistant
- Connected to Holded or your calendar
- Live in 6 weeks
Internal automation
Inbox triage, quote drafting or document work across your own systems.
- Scoped in the half-day
- Quoted as one number
A person is always one step away
Every assistant we build knows when to stop and hand the conversation to you, with the full thread.
Hosted in the EU
Your data stays in Europe and you get the processing agreement to prove it.
Documented under the AI Act
Purpose, data, limits and oversight written down, so a client or an auditor can read it.
Three recent ones.
Read up before you decide.

The cookie banner is on. Your site still isn't compliant.
Most consent banners block nothing. The Network tab in your browser tells you the truth in under two minutes, and the fix is usually one line in Tag Manager.

The online customer journey, and where people actually drop out
Six stages between never having heard of you and buying twice. Naming them is easy. Finding the one where you are losing people is the useful part.

Google Tag Manager, and when it is worth the trouble
One container on your site, and every tracking script managed from a dashboard instead of a developer ticket. Useful, as long as somebody keeps it tidy.
Before you ask
Will it make things up?
It answers from your own content and says it does not know when the answer is not there. That is a setup decision, and we test it with your real questions before it goes live.
Do we have to tell people it is AI?
Yes, and we write that line for you. Under the EU AI Act a chat assistant has to be recognisable as one; the disclosure is part of the build.
What does it cost to run?
Model usage is billed at cost, usually under 50 euro a month for a small business. We show you the actual numbers in the first month.
What if it is not worth it for us?
Then the half-day ends with us saying so. We would rather lose the build than sell an assistant that answers four questions a month.
AI is only as good as what it can see.
Want to know if AI would actually save you anything?
The free call is a straight answer on whether there is a job here worth automating, and what it would cost.

