Frequently asked questions about AI agents.
What people ask us before a first call. If your question isn't here, ask it on the call.
Understanding AI agents
What is an AI agent?
A program that receives a written mission, reads the information it needs from your tools, prepares or carries out the work, and stops where a person has to decide. It runs on its own, on a schedule or whenever something new happens.
What does an AI agent look like?
You talk to your AI agent on the channel of your choice: Telegram, Slack, Discord, WhatsApp, email or something else. We decide it together.
How is it different from a chatbot?
A chatbot answers questions in a chat window. An agent acts in your tools. It updates a record, drafts a quote, follows up with a client, files a document, chaining several steps together.
Isn't an assistant like ChatGPT in a browser tab enough?
It answers when someone opens the tab and pastes the text. An agent takes on a defined mission in your tools, even when nobody is at the screen. It knows your company's context and doesn't need to be prompted again.
How is it different from classic automation?
Automation follows a fixed script and breaks as soon as a request falls outside it. An agent reads freely written messages and documents, handles variations, and passes cases it can't classify to a person.
Does an agent replace an employee?
It takes the repetitive part of a process. Sorting, data entry, follow-ups, preparation. People keep the relationships, the decisions and the tricky cases, with more time for them.
Can the agent make mistakes?
Yes, like anyone starting out. That's why stopping rules are written before it acts, and sensitive messages wait for approval. When it gets a case wrong, the correction is written down and it rereads it next time.
Data and security
Where is the data hosted?
The agent runs on our infrastructure, hosted in Europe. It reads from your tools only what its mission needs.
What does the agent have access to?
Only the tools and files listed in its job description, with its own credentials, which you can revoke at any time. Login codes, passwords and bank alerts stay outside what it reads.
What about GDPR?
The agent only processes the data its mission needs. Processing is documented with you, and the AI providers used are bound by contract. Sensitive areas (HR, health, credit) get their own review before anything is installed.
What does the EU AI Act change?
People interacting with an agent must know they're talking to an AI, and important decisions must stay under human control. That's how we design agents from the start.
Who approves what goes to a client?
You do. Sending, pricing, commitments and disputes wait for a person, unless a written rule says otherwise for a simple case you chose.
Deployment
Do we need an AI project already?
No. A task that comes back every week is enough to start the conversation.
Do we need to change tools?
No. We start from what you already have: email, CRM, documents, invoicing. The agent connects to them when access allows. No extra business software.
Is this an audit?
No. The call describes a possible mission and the engagement that goes with it. There's no separate paid assessment.
How long until an agent is running?
The timeline is set with the scope, on the call. A simple mission on accessible tools moves faster than a process that spans several systems.
What size of company do you work with?
Small businesses, SMEs and large groups. The starting point is the same: a specific process, the tools, and a person who stays in control.
Costs and support
How much does it cost?
The engagement is priced on the call, for the written mission. There's no price list on the site.
Who pays for AI model usage?
You do. The agent runs on your own API keys, so AI usage is billed to your account and you stay in control of your data. Connecting the keys is sorted out when the project starts.
What does maintenance include?
Hosting the agent, monitoring how it runs, updates, and adjusting its rules when the way you work changes.
Can we stop?
Yes. The agent's access is revoked, and your data stays in your tools.
Training and logistics
Is training mandatory?
Onboarding is part of every deployment. The AI training, in 4 hours, 8 hours or 3 days, is a separate service, if you want it.
Can training be funded?
It's looked at case by case, on the call.
Where does the work happen?
By video. On site if the work requires it, decided on the call.
Discovery call by video, no commitment.