How it works
At most companies that takes three to six weeks. You only pay for building once it is clear what we are building and what it returns.
An hour at your table or online. We walk through your week and look for the tasks that come back often, cost time and need little thinking. Afterwards you get a short note from us: what we would tackle, roughly what it returns and what it costs. If it costs more than it returns, we say so.
On your process, in your tone, with access to exactly the systems it needs and nothing more. Halfway through we show you what is there, so you can steer before it is finished. In this phase the agent touches nothing that is live.
The agent does the work, but a colleague of yours approves every action before it goes through. That way you see exactly what it does and where it gets things wrong. We adjust based on what goes wrong in those weeks, and something always does.
The agent settles what you have released and escalates the rest. You keep sight of it through the log and a monthly overview: how many tasks, how much time, where it went wrong. We keep improving and expand as soon as you are ready.
An agent with access to your mail and your books calls for clear boundaries. We set those up front, not afterwards.
Payments, contracts and anything irreversible always wait for approval from an employee. The agent prepares, you decide.
You can see what was done, when and on what basis. Useful during an audit and essential if you want to undo something.
Your data is processed on European servers. We sign a data processing agreement and record which data the agent may see.
What passes through the agent is not used to train language models. Not by us and not by our suppliers.
An agent for invoicing cannot reach your sales inbox. Each agent gets only the permissions its job requires.
If we stop working together you take the process descriptions and the settings with you, so another party can pick it up.
That depends on how many tasks you hand over and how complicated your systems are. We work with a one-off fee for building and a fixed monthly fee for running and improving. After the first conversation we know enough to quote a concrete price, and you get it on paper before we start.
No. The agent works inside the packages you already use. If an integration is available we use it. If it isn't, we operate the package the way an employee does.
That doesn't happen at our customers. What does happen: the vacancy that has been open for six months no longer needs filling, and the people who are retyping today start calling customers instead. Who you need and who you don't remains your decision.
That can happen, as it can with people. So we decide per task what the agent may settle itself and what goes past a colleague. If it is unsure it does nothing and escalates. Every action is in the log and can therefore be undone.
The first agent is usually shadowing within two to four weeks. In the weeks after that someone on your side approves its work, so you see for yourself whether it is right before it goes solo.
Quite the opposite. At twenty people one person often does five things at once, and there is a lot of repetition in that. Small companies frequently notice the difference sooner than large ones, because less consultation is needed to change anything.
We are, for how the agent works; you are, for the rules you give it. We record that split together with the data processing agreement before anything goes live.
The monthly contract can be cancelled per month. The build fee is paid once. If we stop working together you take the process descriptions and settings with you.
Send us a few sentences on the work that comes back every week. You'll hear from a person within one working day.