Customer service
Handles the questions that repeat — order status, lead times, returns, invoices — by querying your systems, not by making things up. What it cannot answer goes to a person with the context already gathered.
AI agents
We automate your company, area by area
An AI agent is not a chat window: it is a worker that reads, looks things up, decides and acts inside your tools. We start with one area, leave it running on its own and move to the next. Once there are several, the effect compounds.
By area
Where the bottleneck usually sits
Sales
Qualifies every form that comes in, researches the company, scores it against your criteria and leaves it in the CRM with the summary written. Your rep starts the day with the list already sorted.
Administration
Reads invoices and delivery notes, pulls the data out, matches it against the order and flags what does not add up. Somebody stops typing and starts reviewing exceptions.
Operations
Watches stock, lead times and incidents and raises the alert before the customer notices. Puts together the reports somebody builds by hand every Monday.
Marketing
Drafts in your voice and with your data, checks that campaigns stay inside budget and builds the monthly report without anybody exporting anything.
Management
Pulls together sales, cost and campaign numbers every week, compares them against target and flags what has drifted. You stop asking for reports and start reading them.
Why it compounds
One area saves hours. Several change the company
Automating customer service gives you hours back. Automating sales too means those hours go into selling. And when operations raises its own alerts, the sales team stops firefighting. That is why we do not sell a single agent: we sell an order to do it in.
- 01The area that hurts most, first
The one eating the most hours, or the one losing you customers. Measured before and after.
- 02Then the one that feeds it
Every automated process leaves clean data that makes the next agent more accurate.
- 03And the team moves up
Nobody becomes redundant: people stop doing the mechanical part and start deciding, selling and serving better.
How we build them
With limits, with a log and with an owner
- It works with your data. It connects to your tools and only reaches what its task needs. Nothing is trained on your company's information unless you decide so.
- It knows what it may do alone. We write down what it executes without asking and what needs a person to approve. Refunding €40 is not the same as refunding €4,000.
- It leaves a trail. Every action is logged: what it read, what it decided and why. So it can be audited and corrected instead of trusted blindly.
- Somebody owns it. You get the panel to watch it run and the documentation for your team to keep it. It is not tied to us.
- It gets measured. Before we start we record how long the process takes today and what it costs. Three months in, we measure again. If it has not improved, it changes or it goes.
- It starts small. One area, in production, genuinely working. The AI projects that fail are usually the ones that tried to start with everything.
Pricing
AI agent pricing
From
2,900 €
Reference price, excluding VAT. It is not a fixed rate: every project is quoted in its own proposal. It covers analysing the process, building it, testing and going live. Later agents cost less because the connections are already in place.
FAQ
AI agents
Is this going to put somebody out of a job?
It is not what we are after, and not what happens on the projects we run. What gets automated is the mechanical part: copying data, looking up an order, writing the same reply again. The person stays, handling what actually needs judgement. If your goal is to cut headcount, we are probably not the right agency.
What if the agent gets it wrong?
That is why we define up front what it may do alone. Anything with financial or legal consequences goes through a person. And because everything is logged, when something fails you can see exactly where and fix it.
Do I need my data tidy first?
Not entirely, but you do need to know where it is. Part of the first few weeks is exactly that. If your data lives in a spreadsheet and in two people's heads, you can still start — the first agent will simply be a simpler one.
Do you use ChatGPT or something of your own?
We use the models on the market, whichever fits each task best, and build on top of them the connections to your systems, the rules and the controls. The model is not the valuable part: what surrounds it is.
What if I only want to try one process?
That is exactly how we recommend starting. One area, measured before and after. If it does not work, you will know in weeks rather than in a year.
Getting started
What does your team repeat every week?
Tell us, and we will say whether automating it is worth it, what it would cost and how much time it would give back. If it is not worth it, we will say that too.