Reminders, emails, quotes: an AI agent prepares the work that comes back every day. You decide what it does on its own, and you approve the rest.
In short: it reads, it prepares, your team approves.
An AI agent is a programme. It reads what arrives at your company: an email, an invoice, a quote request. It prepares what comes next and suggests it to you. It can connect to your tools, such as your email or your invoicing software, through their official access points (APIs) where they exist.
It does not replace your team: it prepares the work for them.
A few figures
In short: we look at your tools and your repetitive tasks. You leave with a one-page report and a quote.
In short: you approve what the agent will do. Nothing is built before you agree.
The rulebook answers simple questions:
The rulebook also prepares your paperwork. It provides the elements of your record of processing activities: the document required by Article 30 of the GDPR, which lists what you do with personal data. It states who is the “provider” and who is the “deployer” under the EU AI Act (Regulation (EU) 2024/1689), to be checked with your adviser. The provider develops the system, or has it developed, and places it on the market or puts it into service under its own name. The deployer uses it under its own authority.
Each agent will also have a diagram of its journey, showing where your team approves.
In short: we will connect it to your software, with your rules, where you choose.
Each agent will start from a common template: reminders and invoices, emails and support, quotes and prospects, accounting documents. These templates are being prepared. They will be shared between the companies we work with and tested automatically at every change: a fix made to a template will benefit every company that uses it.
What will be done for you:
Before showing you anything, we test on your anonymised data, then we carry out a security review.
Where the agent runs is your choice: your company's cloud account, in a European region; your own servers; or a dedicated server with us. Only indicators and alerts will go up to our supervision tool. The AI model itself may be called at a provider: the agent's rulebook says where (see the Data and security page).
In short: for a few weeks, the agent suggests but sends nothing. You judge on figures, then you give the go-ahead, or not.
The length is agreed with you. Nothing goes out to your customers or suppliers.
We measure:
Second decision: you give the go-ahead to go live, or not. You decide with the figures in hand.
In short: at first, a person approves every action. The agent only gains freedom with good figures and your written agreement.
The agent starts at level 1, explained below: a person approves every action before it goes out.
We support your team in getting to grips with it: reading a suggestion, approving it, correcting it, reporting a problem. The EU AI Act, by the way, requires companies that use an AI system to take measures to support the AI literacy of their staff (Article 4).
In short: a follow-up by subscription, being prepared: you will see what each agent does, and we will be alerted if there is a problem.
Supervision will be a subscription, on quote. It is being prepared and will include:
You decide twice:
After that, the agent only gains freedom with your written agreement. The rest (technical choices, order of work, tests) is our job: we decide it and we write it down.
Hello, Unless we are mistaken, invoice no. 2041 fell due 12 days ago. Could you let us know the planned payment date? If payment has already been made, please disregard this message. This message was written by an AI agent on behalf of the company.
Correct the tone of the message:
Example: the suggestion is approved. With JavaScript, you can approve it, correct it or refuse it.
Its freedom is set with you, action by action. The same agent can send an acknowledgement of receipt on its own, and submit to you every reply that commits your company.
For example: the agent writes a draft reply to each customer request. Nobody sends it. It is compared with the reply that actually went out.
For example: the agent prepares the reminder for an overdue invoice, and your accountant checks it before it is sent.
For example: acknowledging a request, giving your opening hours, chasing a quote left unanswered. The recipient is told that the message comes from an AI system.
For example: sorting incoming emails by type and priority.
Whatever the figures:
Each agent's rulebook lists these actions in black and white.
Each rulebook lists the cases where the agent stops rather than guess.
For example
In short: the agent only acts alone if the figures prove it and you agree.
Three rules follow:
An agent does not gain freedom because “it seems to work”. It gains it when the measurements show it.
These measurements are in its rulebook from the start: the agreement rate, errors, time saved, and indicators specific to its mission. For the emails agent, it is the first-response time. For the payment reminders agent, it is the average payment time, before and after.
You always know why an agent acts alone, and which figures that trust rests on.
In short: same principle: you suggest your budget, we tell you what can be built with it, and you approve.
You approve the specification before the first development. You receive the software after acceptance testing, that is, a point-by-point check. As for the price, you suggest your budget, and we present two or three scopes that fit that amount. The price chosen becomes a fixed quote, signed before any work starts.
Everything starts with the audit. You leave with:
The audit costs €490 excl. VAT.
Accessed on 13 October 2026.
Next step
Everything starts with an audit: your tools, your ways of working, your volumes. You then receive a quote.