How it works
One process. Thirty days. Your decision.
We pick one process that hurts, build the agents it needs against your own data, and run them on your infrastructure. Thirty days, a fixed price, and at the end you decide whether to keep it.
What does an agent team actually do?
Someone asks a question in plain Dutch, and the answer comes back without anyone walking to another desk.
Your knowledge is spread across systems that do not talk: the ERP, a spreadsheet, a drawing, someone's mailbox, and the head of the person who has been here twenty years. An agent team is a specialist for one of those workflows. It reads your sources, understands your terms, and answers the questions your people already ask each other all day. Not a chatbot on top of your website. A colleague who has read everything and never forgets.
Why your people first
Why do you start with our people and not our data?
“Je moet er wel iets verstand van hebben, anders bouw je iets wat toch niet werkt.”
Because the data does not say what it means. Your people do.
Before anything gets built, we sit with the people who actually run the workflow and learn how it really works, including the parts that are not written down anywhere. What “afgekeurd” means in your shop. Which field in the ERP is the one everybody actually trusts. Why that one order always goes through Jan. Domain expertise is not a nice-to-have at the start of this, it is the input. Skip it and you get a system that is technically correct and useless.
What happens in the first thirty days?
One process, one data source, and a report at the end that tells you whether it worked.
- Week 1
- We get access to one system and measure how the process runs today: how long an answer takes, how often it is needed, what it costs when it is wrong. If getting access takes longer than a week, that is itself the finding, and it is worth knowing.
- Week 2 and 3
- We build the agents against your data, on your infrastructure. Your process owner reviews twice.
- Week 4
- Your people use it. We stay out of the way and measure.
- Day 30
- One hour, with you in the room. The numbers, and your decision.
What do we need from you?
One person who owns the process, access to one system, and thirty minutes of your time on day 30. That is the whole list.
- One process owner
- Someone who actually does the work and can tell us when we have it wrong, not someone assigned to manage us. This matters most: if nobody on your side owns it, the experiment fails, and it fails in a way that looks like our fault and is not.
- Access to one system
- The one system that holds the process we agreed to work on. Read access is enough to start.
- Thirty minutes on day 30
- You, in the room, when we show you what we measured. That is the whole list.
What it costs
€5.000, fixed.
Not an hourly rate, not an estimate, not a range that becomes a bigger number later. One process, one data source, one report, for the thirty days. If it takes us longer than we thought, that is our problem, not your invoice. What happens after day 30 costs more and we will tell you exactly what before you decide anything, but you never have to have that conversation to find out whether this works for you.
What if it does not work?
We turn it off and take it out, and you have spent €5.000 to find out something true about your own company.
There is no phase two you get talked into. On day 30 you see what we measured against what we measured in week 1, and you say keep it or stop. If you stop, we remove it, and we would like thirty minutes to hear why, because that is worth more to us than the deal was.
And if it works?
The same team builds the next workflow, and the one after that, on the same foundation.
The agents from your experiment stay where they are and keep working. Everything after that clips onto the same lattice: the next data source, the next workflow, the next question your people are tired of asking each other. It runs on your infrastructure, it stays yours, and you can take over operating it yourself whenever you want.
Who actually builds this?
We do, using the same agents we run our own company on.
Cultivator AI runs on its own agent teams. They handle our research, our engineering, and our operations, and they are the same architecture we would deploy at your company. That is not a case study we bought. It is how we work, every day. We take one new customer at a time, so the person who sits with your team is the person who builds it.
When is this not for you?
If your processes are already documented, standard, and running well, you do not need us.
This works when your value lives in how your company does things, and when a lot of that lives in people rather than in systems. If you already have a software team that could build this, they should. If what you want is a tool everybody else also has, buy the tool, it will be cheaper. We are for the part of your business that is yours.
Waar zit het bij u vast?
Thirty minutes, at your site, no deck. We want to hear where the knowledge sits and what happens when it walks out the door.
Book a call