FAQ
The questions that decide the choice
Not “what is AI”. The questions that usually come up at the second meeting, answered in advance.
What happens when the agent makes a mistake?
At the start the agent does not write anything into your systems. It only suggests a result, a person confirms it, and we measure how often their decisions differ. The right to act independently is granted per type of operation, only when the numbers justify it, and it can be withdrawn in the admin panel in one click.
Does our data leave our network?
Our standard option uses cloud AI models through a private gateway: personal data is hidden before a request is sent and restored after the answer, and every request is recorded. If no data may leave your network, a closed network with a local model is possible as an additional option that we assess individually: it needs your own hardware, and local models are usually less capable than the best cloud ones.
Who owns the code and what happens if we stop working with you?
The code, the instructions for the agents, the settings and the logs belong to you. The system runs in your environment. If we stop working together, it keeps running: we hand over the repository and the operating instructions, and your team or another contractor can continue.
Do we need to change our ERP?
No. We connect through the interface your system already has: web services, a copy of the database or file exchange, and we write results back through the same route your integrations already use. If the audit shows that your data needs to be cleaned or brought to a single standard, we agree that step with you separately.
Why do you start with an audit and data, not with AI?
An agent is only as good as the data it receives and the place where its result is recorded. So we first study the systems, set up automatic data exchange and bring the data to a single standard, and only then connect AI. This takes more work at the start, but the result is reliable.
How long until we see something real?
Usually one to two months for a working pilot on one process, longer if you need a closed network or your source system has no open interface. The configurator on this site gives you the estimate for your specific setup before you talk to anyone.
How is this different from buying a ready-made AI assistant?
A ready-made assistant answers questions about the documents you give it. What we build works with your live business data, checks the result against your rules and writes it back into your system, and that is the part that really removes work from people. Ready-made assistants have their place for other tasks.
What does it cost?
A pilot is a project with a fixed scope, priced from its estimate: the systems involved, the requirements and the volume of the process. Instead of a price list, we publish the logic of the estimate in the configurator, because the same process costs very differently with a system that has an open interface and with an old closed one.
Your question isn't here
Ask the AI architect, which answers from our cases and articles, or write to a person.