Your questions, our answers

With what costs you time, not with the technology. The method comes down to five steps:

• list, with your teams, the tasks that come back every week;
• put figures on them: how often, and how long each time;
• hand the heaviest and simplest one to an agent: that is Task Automation;
• measure the gain 30 days after go-live;
• expand only if the result justifies it.

Not sure which one to choose? The free Flash Diagnostic is there for that. If your teams already use AI without written rules, start by setting rules for it: that is what the AI Usage Audit is for.

A contact available two hours a week, real examples, and someone who can settle a question when one comes up: nothing is left hanging. The rest is on us.

Beforehand, the Flash Diagnostic asks you for a questionnaire of under 10 minutes and a one-hour interview. After go-live, depending on the offer, allow a 60 to 90-minute onboarding session, then a few hours a month for the person who keeps the agent alive in your company.

Not always. For one to three tasks, or for a clear process with known rules, the Flash Diagnostic is enough to set the price. If a process is complex only because of its connections to your tools, we check them during the diagnostic. A process that is complex because of its rules, special cases or data goes through the Process Audit first. Business Agents always start with the Domain Audit.

An audit can also conclude that nothing should be done: you keep the file, usable even with another provider.

In three steps, and it is free:

• an online questionnaire of under 10 minutes: your company, your tools, what takes up most of your time;
• an interview of 45 to 60 minutes, on site in the Annecy area and the surrounding region, remotely elsewhere: we talk about your work, not about technology;
• within 72 hours of the interview, a written report: 3 to 5 leads costed in time saved, an order of magnitude for the budget, and one recommendation, only one.

A person reads your answers: no score, no automatic result.

No. It is free and with no commitment: it is the way to judge us on the evidence, before spending anything. One hour is enough to know whether we can be useful to you, and to tell you when we cannot.

The written report is yours to keep, and you do what you like with it. It may conclude that an off-the-shelf tool is enough for you. And the diagnostic requires no access to your data or your tools.

No. We work with SMEs and mid-sized companies, and our questionnaire allows for small organisations, where one person wears several hats. An SME sometimes finds it even easier than a large group: its decision circuits are short, and a single person can lead the approach.

What matters is not size, but work that comes back often and takes up your time. You start with a task, measure, then expand.

Yes, it is even the safest route. Task Automation takes over a task that keeps coming back, in one to two weeks, with no audit, from €1,490 excl. VAT; the gain is measured 30 days later.

If the result justifies it, the Process Agent takes over the process the task belongs to. At the start, one or two processes are enough. When several belong to the same business domain (a department, a whole area of your activity), it is better to handle it as a whole, starting with the Domain Audit.

Work that comes back often and is easy to measure, not necessarily the most expensive: that one is sometimes the riskiest or the least documented. A clear first success on an average piece of work is better. Five criteria help you choose:

• it comes back at least several times a week;
• the information is in a tool, not in someone’s head;
• an occasional error can be put right;
• the gain can be measured, from a figure known today;
• it can be stopped tomorrow without damage.

Qualifying incoming requests, meeting minutes, sorting emails or routine customer requests often tick these boxes.

No. The technical skills are on our side. Above all, you need someone who knows the work being handed over, its exceptions and its rules: that knowledge is what makes a good agent.

For an SME, hiring an AI specialist is rarely useful at the start: the profile is scarce and expensive, and an isolated specialist soon burns out. Better a business leader who drives the approach and an internal lead, who knows the work and keeps the agents alive. If you have an IT department, its main help is opening access to your tools.

Almost nothing: an online questionnaire of under 10 minutes. It asks about your everyday tools, one to three activities that take up most of your time, how often and how long, and the one thing you would like to stop doing by hand.

No file to prepare, no access to your data: you describe your work to us, and that is enough. For the interview: one hour of your time, and a person who knows the processes well, if possible the one who does the work.

It is not a dead end: it is often your real first project. For a first task, your data is often ready enough. If the information is scattered, or only in one person’s head, it is better to start elsewhere, or first to gather it and write down the rules. If nothing is measured, the starting point has to be recorded.

We start from your data as it is: rebuilding or cleaning it up is not included. The Flash Diagnostic tells you where to start.

Not for a first step. A well-chosen task, measured before and after, teaches you more than a strategy written in advance. A shared AI platform is not a prerequisite either.

What matters is a few shared rules, applied to every agent: who approves what, which access, which records, where your data is processed. A shared memory, the Company Brain, becomes useful when several agents have to work together. And if your teams already use AI without written rules, the AI Usage Audit gives you an AI usage charter and a correction plan.

Six things, written into every proposal:

• access to the tools concerned, obtained before the start: it is the first cause of delay;
• a contact who knows the work, available two hours a week;
• real examples, difficult cases included: for a complete process, the agent is checked on 15 to 30 of your cases;
• the process described in writing, or a workshop to capture it with us;
• someone who can settle a question when one comes up;
• a test environment, if the agent writes into a production tool.

No need for detailed specifications: a description of the real work, exceptions included, and the indicator to measure are enough.

Ask yourself whether there is a decision to make along the way. If there is none (receive, extract, file), it is a task. If you have to choose between several next steps depending on the case, it is a process: three to six linked tasks, with special cases; beyond six tasks, it counts as two.

If you describe an entire department (“our customer service is swamped”), it is a business domain: at least two processes that share the same rules, the same vocabulary and the same person in charge. During the Flash Diagnostic, we count the tasks in each process with you.

From €1,490 excl. VAT for a task. Our prices are public, excluding VAT, and follow what you hand over:

• a task (one trigger, one action, one result): Task Automation, from €1,490 excl. VAT;
• a process (several steps, with decisions): the Process Agent, from €3,900 excl. VAT;
• an entire business domain (a department, such as customer service or purchasing): Business Agents, from €12,900 excl. VAT, after a Domain Audit.

For the Process Agent and Business Agents, these are base prices, depending on complexity, audit not included. Your price is agreed with you in the quote, before we start: it is guaranteed, with no surprises.

Often a few months, when the task takes several hours a week. A rare task never pays for itself: it is not a good candidate. To estimate it:

• put a figure on the time spent today: frequency, duration, and the fully loaded hourly cost of the person (gross salary and employer contributions, divided by the hours actually worked; if in doubt, a low estimate);
• take away what will stay human: review, special cases;
• compare the gain with the cost of setting up, running and, if you choose it, Monitoring.

Also count the gains that are not time: a quote sent the same day is signed more often, better-tracked stock avoids shortages.

You do not pay the last instalment. The price and the indicators (two at most, time spent or processing time for example) are agreed with you before signing, and their starting point is measured before we start. If, at final approval, they are not met, the last payment instalment is not invoiced, and we complete the work, up to two days at our expense.

Running costs, and the options you choose.

• Running costs: AI model usage, depending on the volume handled, hosting and third-party tool licences. They are priced in the same quote: you know them before signing, and you pay them at actual cost or as a flat fee, as you prefer.
• Monitoring, optional: from €190 excl. VAT per month.
• The evolution credit, optional, to make the agent evolve: €390 excl. VAT per month.
• With Sovereignty level 2 (France) or level 3 (In-house), an annual check: €390 excl. VAT per year.

Sometimes, yes, and we will tell you: it is one of the possible conclusions of the Flash Diagnostic. An assistant like ChatGPT or Copilot helps each person write, summarise or search, provided you write rules of use for it; it does the same thing for everyone. For recurring work to get done without anyone chasing it, you need an agent connected to your tools: we start from the way you work, with your rules, and everything we write for you is handed over to you. The two are often combined.

For a simple task, yes, and it is sometimes the right choice. The difference comes after delivery: an agent connected to your tools must be checked, monitored and adapted when a vendor changes its software. We deliver those checks, the documentation and, if you choose it, Monitoring, and everything is handed over to you: your solution depends on no one. To compare two quotes, look at the scope, the checks and what is handed over to you, not just the price: the same word “agent” covers very different work.

No, as long as the scope does not change. The price is agreed in the quote, before we start, once the scope has been counted during the Flash Diagnostic, or during the audit when one is needed: which tasks or processes, how many tools to connect, which human checks. It is guaranteed, including for offers shown with a “from” price.

A small deviation in scope is absorbed at no extra cost. If a new need appears along the way, we flag it to you and you decide whether to add it: nothing is invoiced beyond the order without your written agreement.

First, the extent: a task, a process or an entire business domain (a department of the company). Your quote starts from the offer’s base price, then follows these sliders:

• the tools to connect, with or without a documented connection;
• the state of your data, up to date or scattered, and access, obtained in advance or still to obtain;
• special cases, rare or many, and the human checks you want;
• the sensitivity of the data (health, finance, secrecy) and where it is processed: Europe is included, France as an option, your premises subject to a study;
• for the Company Brain, what is already written down in your company;
• your availability: two hours a week and quick decisions keep the price on track.

Because the price shown covers the basic scope described: your quote starts from there.

• For the audits, Task Automation and Monitoring, it is the first tier of a published scale: each additional process, team, task or business domain has its own price, also published.
• For the Process Agent and Business Agents, it is a base price, depending on complexity (known rules, or many exceptions), audit not included. This complexity is assessed with you during the Flash Diagnostic or the audit.
• For the Company Brain, the price of each business domain is added to that of the shared foundation.

In every case, your price is agreed in the quote, before we start: it is guaranteed, with no surprises.

From €1,900 excl. VAT, depending on the audit:

• Process Audit: €1,900 excl. VAT for one process, + €890 excl. VAT for a second;
• Domain Audit: €3,900 excl. VAT for two processes in the same business domain (a department of the company) and their rules, + €890 excl. VAT per additional process; the public funding application is included, when a scheme can fund the project;
• AI Usage Audit: €2,900 excl. VAT for one entity, up to two teams, + €390 excl. VAT per additional team.

An audit is paid separately: the base price of the Process Agent or Business Agents does not include it. The Flash Diagnostic, for its part, is free.

In several instalments; for a project that puts agents into service, the last one depends on the result.

• Such a project is paid at launch, at the approved demonstration, then at final approval, tied to your indicators.
• An audit is paid on order, then at the presentation of its findings.
• Monitoring and the evolution credit are paid monthly, in advance.

Invoices are payable within 30 days. The exact payment schedule, adapted to the amount, is set out in your quote: you know it before signing.

At the start, yes. A per-user subscription costs little, but guarantees no gain. A custom agent requires an initial investment, in proportion to the scope: a task costs far less than a process or an entire business domain (a department, such as purchasing), and Task Automation starts at €1,490 excl. VAT. In return, it does precise work, within your tools, with a measured gain. Some off-the-shelf tools let you build a simple agent; as soon as several tools must be connected, your rules applied and decisions kept in your hands, the design work remains the same.

No. You pay for a result, not a number of agents: a task, a process or a business domain (an entire department) taken over. How many agents are needed is a design choice, and it is ours to make: a process may need one agent, or one agent and two automations. Charging per agent would make you pay for our design choices; the price follows what you hand over.

No: you commit phase by phase. A large-scale project combines several offers, for example a Domain Audit, then Business Agents, then a Company Brain. A single quote gives the price of each phase, and each phase is a separate project, with its own scope, deliverable and approval. The next one starts with your agreement, once the previous one has been delivered and approved; you can stop between two phases without owing anything for those that have not started. The knowledge of a business domain (a department of the company) is only ever charged once, and each business domain has its own Business Agents, after its own Domain Audit.

Software that does a piece of work for you, within your tools, following your rules. It receives a goal, chooses the steps to reach it and acts: it reads an email, checks that an order exists, updates a customer record, drafts a reply. It applies your timeframes, your prices, your way of answering a customer, and hands over to a person as soon as a decision commits the company.

It is not a robot: it has no physical form. Depending on your need, you hand agents a task (Task Automation), a process (Process Agent) or an entire business domain, such as customer service (Business Agents).

Yes, and denying it would be a lie. It can misread a document, misapply a rule, or invent a plausible answer when information is missing (what is called a hallucination). We design it so that errors are rare, visible and can be put right:

• it is checked on your real cases before go-live (15 to 30 cases for a Process Agent);
• it answers from your sources, and hands over when it does not know;
• whatever commits the company is approved by a person;
• every decision leaves a record and, with Monitoring, how it works is followed over time.

The acceptable level of error depends on the cost of an error: each action is therefore classified by what is at stake.

No: an agent takes on tasks and processes, not a whole job. It absorbs the repetitive work; what requires judgement stays with people: deciding, handling exceptions, looking after your customer relationships. The person who used to do the work supervises the agent and approves what needs approving.

In an SME, the time freed up first goes to what had been put off for lack of time: following up customers, preparing quotes better.

No. The agents work within the tools you already use: email, management or accounting software, customer relationship software (CRM), shared documents. When a tool offers no connection, as is often the case with older software, we go through files or exports, and the quote allows for it. Overhauling your information system is a different project, and it is almost never a prerequisite.

The agent does the work itself, within your tools.

• An assistant such as ChatGPT, Copilot or Le Chat helps a person write or summarise; that person stays in charge.
• A chatbot answers in a chat window, often from a list of answers.
• Traditional automation always runs the same sequence of actions, without understanding free text.
• An agent understands a request, chooses the steps and acts, within a set framework.

Traditional automation is not obsolete: for work that never changes, it remains the simplest and cheapest solution, and we will tell you so. The two are often combined: the automation triggers the agent, the agent handles the case.

You do, depending on what is at stake in each action. The mode of each action is written down before go-live, and you approve it:

• it acts alone on what is reversible and low-stakes, such as the minutes of a meeting;
• it prepares, you approve what goes outside the company, such as the reply to a complaint;
• it prepares, you decide and carry out what cannot be undone, such as a payment.

By default, what goes to your customers is therefore approved by a person. This setting remains yours: for payment reminders, for example, you can let the first courteous ones go out on their own, or exclude certain customers from any automatic reminder.

No. No irreversible action goes out without a person’s approval: it is a design rule, not an option. An agent never does any of these alone:

• pay or make a transfer: it prepares, a person executes;
• sign, or set a price;
• publish without review;
• handle a sensitive dispute: for a delicate complaint, it reconstructs the history and prepares the reply, a person approves it;
• send a formal notice, or settle a strategic decision.

It can prepare, compare and calculate everything. The decision remains yours.

Your company remains responsible for what it does, with or without an agent. That is why actions that commit it remain approved by a person. Each agent also comes with a log of what it does and a trail of its decisions: you can understand an error, correct it and prevent it from happening again. The documentation handed over explains how to take back control in case of error.

It says so, and hands over. Instead of inventing, it passes the request on to the right person, with a summary of the exchange. The hand-over rules are written during scoping: for a given type of case, or beyond a threshold of doubt, the agent stops and passes it on.

So that it does not invent, it answers only from your sources (your terms, your products, your timeframes), never beyond them. These rules for handing over to a person come with every Process Agent and with Business Agents.

By involving them from the start. The person who does the work takes part in scoping: they know the exceptions, and an agent they helped build is more readily accepted. Then:

• a 60 to 90-minute onboarding session, on real cases, with a simple fact sheet: what to hand it, what not to hand it, what to do if it gets something wrong (included with the Process Agent and Business Agents);
• visible control rules: everyone knows who decides;
• a gain measured on their own work, before and after.

If someone refuses to use it, find out why: a badly handled case or a legitimate concern is valuable information for improving the agent.

By default, European or open AI models, hosted in the European Union, at no extra cost. We choose the right model for the task, without depending on a single supplier; this choice is documented, justified and open to review, and switching AI supplier is simple. If, for a specific task, the right tool is outside the Union, the exception is written down, justified and approved by you.

Each proposal names the providers involved, in its “where your data goes” page. Your data is never used to train an AI model. If you choose Monitoring, it adapts what is in service when a model changes.

Yes. Even if it is not in the catalogue, your process can have its agent: it is often a case specific to your business, spanning several departments, or written down nowhere. We build it with the same method and the same rule: no irreversible action without a human check.

The free Flash Diagnostic first looks at whether it lends itself to an agent. If it is clear, you then receive your quote for a Process Agent; if it is complex because of its rules, its edge cases or its data, the Process Audit examines it first. Several processes in the same business domain fall under Business Agents, always prepared by a Domain Audit.

Yes, provided they are organised from the second one onwards, if they deal with the same customers, the same data or the same rules. Otherwise, the quoting agent and the customer service agent end up announcing two different delivery times. One agent can coordinate others, but the essentials come down to five rules:

• a clear role for each agent;
• written hand-overs;
• shared rules, in a single place;
• decisions reserved for people;
• records that follow the case.

Within a business domain (an entire department), after the Domain Audit, Business Agents pass the work on to each other; when several business domains are involved, the Company Brain can hold the shared rules. You pay for a result, not a number of agents.

Yes: an email, a shared document, a scanned letter. It extracts what matters (an amount, a date, a reference) and files it in your tool. A scanned document calls for more precautions: if it is badly scanned or illegible, it is flagged to a person rather than guessed.

A document that arrives already structured, such as an electronic invoice, makes its work easier without making it unnecessary: it still has to be matched with the order, discrepancies checked and rejections handled. Extracting the data from an invoice received by email and filing it in your accounting tool is a task: Task Automation is enough.

Your company’s memory, written down in one place: your teams find the right answer there, your AI agents apply your rules from it. It includes:

• the shared foundation: your fundamentals, from the vision to the shared rules, including strategy, offers and prices, organisation and vocabulary;
• your business domains (a department, a whole area of your activity), at least one, added one by one: vocabulary, processes, rules, special cases;
• what you need to use it day to day: the arbitration rules and rules of use, your agents already in service connected to it, and an interface to question it in plain language.

No: it is never a prerequisite. Task Automation, the Process Agent and Business Agents can be bought without it, before as well as after. A first agent only needs the rules of its own work, written during scoping.

The Company Brain becomes useful when several agents or several teams have to apply the same rules, or when knowledge rests on a few people. It can come first, after a project, or never: it is up to you to decide, and the Flash Diagnostic helps you see clearly.

Neither: it is your company’s written memory, built with you. It is structured, dated and kept version after version, readable by your teams and usable by your agents. The Company Brain belongs to you: we are its architect, never its owner. To keep it alive with you, the “Company Brain maintenance” monitoring is offered monthly, on top of Monitoring for what is in service. It is the method we apply to ourselves, every day, set up in your company and tailored to your business.

The shared foundation from €9,890 excl. VAT, plus €3,900 excl. VAT per business domain (a department, a whole area of your activity), always with at least one business domain. What makes the price vary first: what is already written down in your company. A company where nothing is written down needs more work than one that is already documented; we assess this with you during the Flash Diagnostic.

A business domain already covered by your Business Agents is not charged twice. The public funding application is included, when a scheme can fund the project. Your price is agreed with you in the quote, before we start, and the running costs are priced in it: you know them before signing.

Answers that are right for you, and consistent with each other. Without your rules, an agent answers correctly for everyone else, and wrongly for you. With the Company Brain:

• your agents apply your timeframes, your prices, your tone and your exceptions;
• they stay consistent with each other: the quoting agent and the customer service agent announce the same timeframe;
• they invent less, because they answer from your sources;
• your next agents are set up faster: the knowledge is already there.

Your teams use it too: a newcomer finds the rules, everyone knows which version is authoritative. It is not yet another document base: each business domain is kept up to date by its lead.

No. It writes down your fundamentals and the knowledge of your business domains, so that your teams and your agents apply the same rules. Doing the work is the role of Task Automation, the Process Agent and Business Agents. Your agents already in service are connected to the Company Brain; those that come later will rely on it from the day they go into service.

No. Your strategy remains yours. We structure your fundamentals (vision, priorities, shared rules) and make them usable by your teams and your agents; we do not decide them for you. That is why a business leader available to approve the shared foundation is one of the two conditions for success: management approves it at each milestone, every two weeks.

No: it does not learn on its own. It lives because it is kept up to date by your business domain leads, under your governance, and each change is approved by the lead concerned. Three months after delivery, we measure whether it is still kept up to date: that test tells whether we delivered an asset, and not a document. To keep it alive with you, the “Company Brain maintenance” monitoring is offered, on top of Monitoring for what is in service: update workshops, the integration of your new decisions, a quarterly governance review.

Your business domain leads, one for each business domain (a department, a whole area of your activity): each keeps the knowledge of their own up to date, directly, without going through anyone else. Management approves the shared foundation; your teams use it. Someone can coordinate the whole if you wish, never as a mandatory gateway. You appoint these leads with us from the start, and each learns to keep theirs up to date during onboarding.

Little day to day, once the habit is formed: write down a decision when you take it, update a rule when it changes. The most expensive thing is not doing it, and paying for it later in searching and contradictions.

During set-up, six to ten weeks for a first scope, you need a business leader available to approve the shared foundation and, for each business domain, a lead who reviews theirs at each milestone, every two weeks.

Yes, and it is the right way to start. The shared foundation and one business domain (a department, a whole area of your activity) are enough to begin: the one that matters most to you, or the one your agents already cover. The others are added one by one, once the previous one has proved itself. You do not need to write everything down to start.

A first scope is set up in six to ten weeks. If this business domain has not yet been examined, a Domain Audit can prepare it, depending on the case.

In Europe by default, at no extra cost. In France if you choose, as an option (+ €690 excl. VAT), with a written commitment on where it is located, re-checked every year (€390 excl. VAT per year). Or at your premises, on dedicated infrastructure, after a study. You choose when you order.

You do not pay for it twice. Its knowledge is already written down: the Company Brain quote deducts the work already done, as a business domain’s knowledge is only ever charged once. The Company Brain always includes at least one business domain: this can be it, and your Business Agents are connected to it. Conversely, if your Company Brain already contains a business domain, its knowledge is not charged twice to your Business Agents.

No. Each agent is connected only to the scope it needs, with its rights, its approvals and its owners. The customer service agent reads the shared foundation and the customer relations knowledge; it has no business reading HR files. This access is written down, approved by you, and can be changed or withdrawn at any time.

By keeping a single version of each rule. Four principles see to it:

• a single source for each type of information: a rule lives in one place, other documents refer to it;
• everything is dated and kept version after version: you know which version was authoritative, and when;
• named people have the right to write and approve;
• the shared foundation and the business domains are kept separate: the foundation rarely changes and is approved by management; a business domain changes often and is kept up to date by its lead.

You keep it: it belongs to you, with its history, version after version. Your fundamentals, your rules and the knowledge of your business domains (sales, purchasing…) are yours; we are its architect, never its owner. Structured and dated, it remains readable by your teams and usable by your tools. Monitoring, if you had chosen it, can be stopped from one month to the next, after a three-month commitment.

Yes, and in writing. It is processed in Europe by default, in France if you choose, and is never used to train an AI model. Every proposal includes a page that says where it goes. The agent only accesses the data it needs for its work, with access rights that you approve and can withdraw at any time, and no irreversible action goes out without a person’s approval.

In France. The examples, documents and exports you give us during the project are kept on our server, hosted by OVHcloud in Strasbourg. None of them go into our office tools, and a document is anonymised before it goes through any tool located outside Europe, an AI assistant for example. Once the agent is in service, your data is processed where your sovereignty level provides: in Europe by default, in France or on your premises if you choose.

No, never. Your data is never reused to train an AI model: this is written into every one of our projects, and we only use the business offerings of providers that rule out this use. Check this point in writing with any provider: consumer tools, for their part, sometimes allow what your teams enter into them to be reused.

Not in what we put into service, unless you agree in writing. The documents you give us during the project stay in France, on our server. By default, our projects use European or open AI models, hosted in the European Union, at no extra cost. Many solutions go outside Europe by default; our default is Europe. If, for a specific task, the right tool is located outside the Union, the exception is written down, justified and approved by you: for this task, what leaves, where, and why.

Yes, with Sovereignty level 2 (France), offered as an option on Task Automation, the Process Agent, Business Agents and the Company Brain. The AI models and the hosting are in France. We check the whole chain end to end (logs, backups and monitoring tools included), then give you a written commitment on where your data is located, re-checked every year. If moving to a French model calls for adjustments, we redo the settings and the checks: the quality stays as agreed.

No. It only accesses the data and tools it needs for its work, read-only or with write access, as written down at the start. You approve the people, the roles and the access levels; they can be changed and revoked at any time. The documentation delivered with each agent says who has access to what and where the data flows, and the “Where your data goes” page of the proposal names the providers involved.

Three boundaries your data does not leave, which you choose when you order:

• Sovereignty level 1 (European), included as standard: Europe, with European or open AI models hosted in the European Union;
• Sovereignty level 2 (France), optional: France, with the whole chain checked and a written commitment, re-checked every year;
• Sovereignty level 3 (In-house), on request, after a study: at your premises, on dedicated infrastructure, at your expense.

The Flash Diagnostic and the audits do not process your data in a system: they are what determines the level needed for what follows.

Not necessarily: for most companies, it is not a legal obligation, but a choice or a client’s requirement. The GDPR provides for the free movement of data within the European Union. Some sectors have stricter rules: in France, a provider hosting health data must be HDS certified and store that data within the European Economic Area; state administrations that entrust sensitive data to a cloud provider must choose a SecNumCloud-qualified offer. Processing in France is mainly useful for:

• a business bound by professional secrecy, such as a law firm;
• a company whose client requires it, in a security questionnaire or a call for tenders;
• any business leader who makes it their choice.

An extra charge added once per project, depending on the offer:

• + €290 excl. VAT on a Task Automation;
• + €690 excl. VAT on a Process Agent;
• + €690 excl. VAT on Business Agents;
• + €690 excl. VAT on a Company Brain.

On top of this comes the annual sovereignty check, €390 excl. VAT per year. What you pay for is not the change of model, which is simple: it is the check of the whole chain and the written commitment. Europe, for its part, is included in the price of every project.

From a page attached to every proposal: “Where your data goes”. It says where each processing operation takes place, which providers are involved, how long the data is kept and what never leaves. You know before signing, and you can check it afterwards. If a provider, whoever it is, cannot give you this document, that is a warning sign.

No. The model is only one link: logs, backups, monitoring tools and other providers may be located elsewhere. Guaranteeing that nothing leaves the country requires checking the whole chain, end to end, then committing to it in writing. That is what Sovereignty level 2 (France) includes, re-checked every year.

Yes, after a study, with Sovereignty level 3 (In-house): your data does not leave your premises. It relies on dedicated infrastructure, at your expense, priced after an architecture study, and comes with the annual sovereignty check (€390 excl. VAT per year). It is offered with the Process Agent, Business Agents and the Company Brain; for standalone tasks, it goes through a Process Agent or Business Agents.

To prove, every year, that your data stays where we committed to keeping it. On the anniversary date, we re-check the whole chain that processes it and give you a written, dated certificate of location. It is mandatory with Sovereignty level 2 (France) or level 3 (In-house) and costs €390 excl. VAT per year. Monitoring, for its part, stays the same whatever the place of processing.

It is used to prepare the interview and the report, then to follow up on our exchange; you can opt out of our messages at any time. A person reads it: the questionnaire calculates neither a score nor a recommendation. It is neither sold nor used to train an AI model; without any follow-up from you, it is deleted three years after your last contact. The site sets no cookies. Its forms are hosted by Netlify, in the United States, a transfer covered by safeguards; we delete your submissions from Netlify at the latest 30 days after they are received. The details are in our privacy policy.

No. The GDPR governs the processing of personal data, whatever the tool: a well-designed AI agent complies with it like any other software. With us, every project that involves personal data (yours, your employees’ or your clients’) is designed from the outset to comply with it, at no extra cost. Always included:

• processing only the data that is necessary, with a legal basis for each processing operation;
• defined retention periods, and the rights of the people concerned respected;
• providers bound by contract, including the AI model supplier;
• audit trails, and an impact assessment when one is required.

Yes, as soon as your teams use an AI tool, ChatGPT included. The European AI regulation (the AI Act) mainly asks three things of an SME:

• competence: everyone is trained in proportion to their use;
• transparency: the people you deal with know when they are interacting with an AI, and the generated content you publish is flagged;
• caution: no screening of job applications or assessment of employees left to a tool alone, and no prohibited practice.

For most SMEs, the question will first come from a client asking for guarantees, not from an inspection.

No: a ban without an alternative moves usage to phones and personal accounts, which are even less controlled. The gains remain, the control disappears. It is better to allow chosen tools, with rules everyone understands, in four steps:

• take an inventory of uses, including those nobody has declared, without seeking to punish;
• write an AI use charter: the tools allowed, the data prohibited, who reviews what;
• identify the technical gaps and plan how to fix them;
• keep evidence to show your clients.

That is the work of the AI Usage Audit, delivered within five working days.

Yes. Since 2 August 2026, the European AI regulation has required people to be told when they are interacting with an AI, on your website, on the phone or by email. In everything we deliver that interacts with people, this notice is built in by default. It is also a matter of trust: a customer who finds out afterwards feels deceived.

Yes, everyone in proportion to their use: it has been an obligation under the European AI regulation since February 2025. For a specific agent, the onboarding session with its users contributes to this; it is included with the Process Agent, Business Agents and the Company Brain. For broad use of off-the-shelf tools, an AI use charter and awareness-raising for your teams are useful: the AI Usage Audit gives you this charter.

Yes. When we process personal data on your behalf, we are your processor within the meaning of Article 28 of the GDPR: a personal data processing agreement is signed with the contract, of which it forms part. It names the providers that process this data, including as soon as a third-party AI model is involved, and we remain responsible for them towards you.

In stages, from 2025 to 2027 (law as at 11 September 2026):

• since February 2025: the prohibited practices, and the obligation to train the people who use AI;
• since 2 August 2025: the obligations of providers of general-purpose AI models (the vendors, not the users);
• since 2 August 2026: transparency, that is, telling people they are interacting with an AI and flagging generated content;
• 2 December 2026: the technical marking of content, for systems already on the market;
• 2 December 2027: high-risk systems, including those used for recruitment.

Penalties, with high ceilings; for an SME, the lower of the two ceilings provided for applies. For most companies, the concrete risk will first come from a client: not being able to answer when they ask for guarantees on your uses of AI. To assess a risk of penalty, we point you towards your data protection officer or your lawyer.

Four, mainly:

• data that leaks out: a contract, a customer file or a payslip pasted into a consumer tool, whose terms of use sometimes allow it to be reused;
• content published without review, with an error or a made-up claim;
• decisions made on a wrong answer, given with confidence;
• being unable to answer a client who asks which AI tools you use, and with what rules.

The AI Usage Audit identifies them and proposes how to fix them.

Show them evidence: which AI tools you use, with what rules, and what you have fixed. That is often how the question arrives, in a call for tenders or a questionnaire, well before any inspection. The AI Usage Audit gives you a reusable file of evidence, with the inventory of your uses, the register of your tools, an AI use charter and a correction plan. If this client also requires your data to stay in France, Sovereignty level 2 (France) meets that requirement, for the agents we put into service, with a written commitment.

A short document, one or two pages, written for your teams: the tools allowed and for what uses, the data that must never be put into them (personal, confidential, strategic), who reviews what before it goes out, and when to flag generated content. It is based on a real inventory of uses. The AI Usage Audit gives it to you together with the register of your tools; Monitoring can then keep the register and the rules up to date.

No. Our field is technical compliance: what your tools do, what they keep, who approves, how the data is processed and protected. A legal assessment, a certificate of compliance or an opinion on a risk of penalty are for your data protection officer or your lawyer, and we point you towards them. The AI Usage Audit does, however, give you a file of evidence, reusable to answer your clients. And with Sovereignty level 2 (France) or level 3 (In-house), we certify every year, in writing, where your data is located.

Only when the processing presents a high risk for people: sensitive data, surveillance, large-scale automated decisions. The impact assessment then examines these risks and the measures to reduce them. The question is raised during scoping; whether one is required is for your data protection officer or your lawyer to say. When it is, it is part of the project and is not priced separately.

No, but it is tightly regulated. Systems used for recruitment are classified as high-risk by the European AI regulation, with stricter obligations from 2 December 2027. Never leave the screening of applications or the assessment of your employees to a tool alone: a person must decide. That is why, as things stand, we do not offer application screening; an agent can, however, acknowledge receipt of applications.

Yes, for companies based in France, but it mainly funds awareness, diagnostics, consulting and training, more rarely the building of the project itself. Three places to look:

• Bpifrance, the French public investment bank: free tools to take stock of AI, and co-funded support for SMEs and mid-sized companies;
• your region: support for consulting and transformation projects, often reserved for certain sectors;
• your chamber of commerce and industry: free digital support.

Schemes change often: we check which ones are open at the time of your project.

Yes, when your project can be funded, and it is included in the price of the Domain Audit and of the Company Brain. We identify the schemes open to your company, depending on its size, region, sector and project; we check their conditions and timetable; then we prepare the application with you: the description of the project, the budget, the timetable, the result indicators and the documents required. The application is submitted in your company’s name, and we stay by your side to answer the funder’s questions.

The Domain Audit always comes before Business Agents: its roadmap prices the funding available and what remains for you to pay. For the other offers, ask us during the Flash Diagnostic: if a scheme can apply, we will tell you.

Ask us during the Flash Diagnostic, free and with no commitment: if a scheme can apply to your project, we will tell you. The answer depends on your size, region, sector and project, then on the conditions and timetable of the schemes open at that time. If none applies, we tell you that too.

No. The decision belongs to the funder. Not every company is eligible either: depending on size, sector or project, sometimes no scheme applies, and we would rather tell you than prepare a pointless application. The application is made in your company’s name, and your company remains responsible for the information it provides.

Often, yes. Many schemes must be applied for before the project starts: for those, work that has already begun can no longer be funded. Some schemes also only accept providers approved by the funder. That is why we check the conditions and the timetable before any commitment, and tell you from the outset.

That is rarer. Most schemes fund awareness, diagnostics or consulting; some regional grants also cover a study or a proof of concept (a first version, tested on a small scale). This is checked case by case, at the time of the project: for a Task Automation or a Process Agent, ask us during the Flash Diagnostic; for Business Agents, the Domain Audit’s roadmap prices the funding available.

Sometimes, and it is checked before you order it. Diagnostics and consulting are what funding most often pays for, notably some regional grants, each with its own conditions: many must be applied for before the work starts, and some require a provider approved by the funder. Ask during the Flash Diagnostic: if a scheme can apply, we will tell you.

Some, yes. Other schemes require a minimum turnover or headcount. Bpifrance’s free tools and the support offered by chambers of commerce and industry remain open to the smallest companies; for a first step, that is sometimes all you need, and we will tell you.

It depends on the scheme: we do not quote any rate. Rates and ceilings change several times a year; that of the Diag Data IA, a co-funded support programme from Bpifrance, changed at least three times in 2026. For up-to-date figures, contact Bpifrance or your regional advisor, and do not build your budget on a percentage you read six months ago. For your project, the Domain Audit’s roadmap prices the funding available and what remains for you to pay.

The French state’s national plan to spread AI across companies (its name means “Dare to use AI”), launched on 1 July 2025 and stepped up on 18 June 2026. It offers free awareness tools and a platform of use cases; Bpifrance adds co-funded support for SMEs and mid-sized companies. Its goal: 80% of SMEs and mid-sized companies, and half of very small businesses, using AI by 2030.

No. What we deliver belongs to you: the code, the instructions given to the AI, the connections to your tools and the documentation are all documented and handed over to you (except third-party tool licences). Your team, or the provider of your choice, can take it over and develop it. Our methods and generic tools remain ours: where they are part of it, you keep the right to use them with no time limit. Your data and what is specific to your company (your rules, your procedures) remain yours. Our choices of AI models and providers are documented, justified and open to review, and switching AI provider is simple. And Monitoring remains optional.

From €190 excl. VAT per month. With us, maintenance is called Monitoring, and its price follows what is in service:

• a task: €190 excl. VAT per month, + €90 excl. VAT per additional task, up to three;
• a process: €390 excl. VAT per month, + €290 excl. VAT per additional process;
• a business domain (an entire department): €890 excl. VAT per month, + €590 excl. VAT per additional business domain.

The business domain tier covers its processes and their tasks: they are not charged on top. What is in service separately adds up, for example a task and a process that are unrelated. Company Brain maintenance, offered in addition, adds €590 excl. VAT per month. Three months’ commitment, then monthly: you stop whenever you want.

You keep everything: the code, the instructions given to the AI, the connections to your tools and the documentation (except third-party tool licences), and you can develop them further with the provider of your choice. Your data remains yours, and the access you had given us can be revoked. Monitoring can be stopped at the end of any month, after the first three months of commitment.

No. It is offered with every project, without you having to ask for it, and remains optional. If you take it, it starts at go-live, with a three-month commitment, then monthly: you stop whenever you want. Without it, plan who, in your company, watches the agent and adapts it when a tool changes: an agent connected to your tools lives with them, and without monitoring, what used to work ends up missing the mark.

Checking that everything works properly, fixes, adaptation to your tools and to AI models, and a value report every quarter, with you. It expands with what is in service:

• from one process: keeping running costs under control, and a monthly check on your cases;
• from two business domains: a check every week, a value report every month and a direct contact.

It maintains what is in service: changes fall under the evolution credit. It is neither on-call support nor a committed response time.

Your teams take back control, then the agent is fixed. Every agent is delivered with a log of what it does, where errors are reported, and documentation that explains how to take back control in case of error. With Monitoring, we check that it works properly and fix it, at the agreed pace. It is neither on-call support nor a committed response time. Without Monitoring, the person you appoint takes care of it, with the same documentation.

Yes. Adding a case, a rule or a tool is separate work: Monitoring maintains what is in service. For changes as you go, the optional evolution credit gives you half a day per month, which can be carried over within the quarter, for €390 excl. VAT per month. Beyond that, work is invoiced, with your agreement, at the rate announced at signing. For a broader need, a new process for example, it is a new project, with its own quote.

We adapt the agent, if you have chosen Monitoring. When a vendor changes its software (a field that is renamed, an export that changes format, a connection that evolves), the adaptation is included. If you replace one tool with another, a new customer management tool (CRM) for example, or add one, that is a change: it falls under the evolution credit, or a new project with its own quote. Without Monitoring, plan who, in your company, keeps track of these changes.

Yes, with the Process Agent, Business Agents and the Company Brain. After go-live, a 60 to 90-minute group session, held remotely: a demonstration in your environment, then exercises on your cases. For an agent, your teams see what to ask it, what not to entrust to it and what to do if it gets it wrong. You keep a simple onboarding fact sheet. Task Automation includes the documentation and the handover. An extra session, for other teams, is priced separately; general AI training is not included.

Through the value report, included in Monitoring. Every quarter, with you (every month from two business domains), it sets what has been handled against the indicators agreed at the start: volume handled, cases passed on to a person, time saved. Gaps are visible, and the decision to continue, correct or stop remains yours, made on the facts. Without Monitoring, track these indicators yourself: their starting point was measured before go-live.

It is replaced: switching AI provider is simple, and our choices of models and providers are documented, justified and open to review. With Monitoring, adaptation to AI models is included; from one process, it also includes keeping running costs under control: the right model for each task, and an alert if costs overrun. A new model respects the agreed place of processing: in Europe by default, in France if you have chosen it; any exception is written down, justified and approved by you.

Yes, provided you appoint a person, give them the time and the access, and write down what they monitor. Every agent comes with what you need to follow it: the log of what it does, the trail of its decisions, access rights you can change or revoke, and documentation that explains how to take back control in case of error. The internal lead is often the right starting point, with outside support for technical adjustments.

You keep it, and it remains usable, including with another provider.

• Process Audit: the current state, the business case and the scoping document, ready to build from.
• Domain Audit: the current state, the ranking of processes, the roadmap and the indicators.
• AI Usage Audit: the inventory of uses, the register of AI tools, the AI use charter and the correction plan.

An audit can conclude that nothing should be done, or that the project is not feasible: we stop, and you keep the file you have paid for.

In seven steps, from the free diagnostic to follow-up:

• the Flash Diagnostic: a one-hour interview, then a written report within 72 hours;
• an audit, only if your case calls for it;
• your quote, within 7 days: the price, the indicators and the page that says where your data goes;
• the build: for a process or a business domain, in two-week stages, each ending with a demonstration;
• the check on your real cases, go-live and onboarding;
• final approval, linked to your indicators;
• Monitoring, if you choose it: a three-month commitment, then monthly.

You only go further if the result justifies it.

One to two weeks for a task, with Task Automation. For a complete process, allow three weeks with the Process Agent, four to five for two processes. An entire business domain (a department, a part of your business) takes four to six weeks with Business Agents, after the Domain Audit; the Company Brain, six to ten weeks for a first scope.

These timeframes assume access has been obtained before the start. Long delays rarely come from technology: they come from access, trade-offs and approvals.

Six, written into every proposal, before signing:

• a guaranteed price, agreed with you before we start, and the running costs estimated in the same quote;
• indicators agreed before signing, measured from the start: if they are not met, the last instalment is not invoiced, and we complete the work, up to two days at our expense;
• a page that says where your data goes, and safeguards for personal data, including when a third-party AI model is involved;
• no irreversible action without a person’s approval;
• the code, the instructions given to the AI, the connections and the documentation, which are handed over to you (except third-party tool licences);
• Monitoring, offered for what comes next.

To one offer at a time. The Flash Diagnostic is free and commits you to nothing. Each offer can be bought on its own, at the price agreed in the quote, and you only go further if the result justifies it. After an audit, you can stop: you keep the file.

Monitoring, offered with every project, is optional: a three-month commitment, then monthly, and you stop whenever you want. The evolution credit (half a day per month to develop what is in service) is optional too.

Synergetik, an AI-native agency: an AI agent integrator for SMEs and mid-sized companies. We put AI agents into service within your current tools, taking over your time-consuming tasks and processes, following your rules, under your control, with measured gains. We are based in the Annecy area, in Haute-Savoie, on the doorstep of Geneva, and work all over France and in Switzerland.

“Native” means that we apply our method to ourselves, every day: our rules are written down, and our own agents rely on them. Our name is inspired by synergetics, the science that studies how scattered elements come into phase. With us, you talk directly to the people who design your agents.

Both. We are based in the Annecy area, in Haute-Savoie, on the doorstep of Geneva: when your company is in the region, we come on site, for the Flash Diagnostic, the Domain Audit interviews or progress reviews, with your teams.

Everywhere else in France, and in Switzerland, we work by video, with the same method and at the same pace. The onboarding session takes place remotely. And our prices are the same for all our clients, in France and in Switzerland: they change neither by region nor by country.

Most of the time, yes: what matters is the work to be handed over, more than the sector. Answering requests, keeping a schedule, processing orders: these exist almost everywhere. Our AI agent catalogue also shows what changes from one sector to another, for example:

• construction, civil engineering, trades;
• accounting firms and the professions;
• e-commerce and distribution;
• industry and manufacturing;
• real estate and property management;
• health and social care.

The Flash Diagnostic asks for your sector: one hour is enough to know whether we can be useful to you, and to tell you when that is not the case.

Seven days. The quote is based on what we have seen together during the Flash Diagnostic, or during the audit when there is one.

It states what is counted (tasks, processes or business domains), the deliverables, the timetable, the indicators to measure, the price, the payment stages, what we expect from you and, for a build, the estimated running costs. A page says where your data goes. Your price is agreed with you in the quote, before we start: it is guaranteed, with no surprises.

At our end, you talk directly to the people who design your agents: we handle the scoping, the build, the checks and the documentation.

At your end, three roles: a contact who knows the work, two hours a week during the build; someone who can settle a question when one comes up; then, after go-live, an internal lead who reports badly handled cases and gets the rules updated. In an SME, one lead is often enough; a mid-sized company tends to appoint one per business domain, often its manager, with one person coordinating the shared rules.

The starting point, and the proof. Many trials prove that something is possible, not that it is useful: chosen because they could be demonstrated, with no agreed indicator, tested on hand-picked data, with nobody to take things forward. We do the opposite:

• work chosen for what it costs you today;
• indicators agreed before signing, measured from the start;
• a check on your real cases, exceptions included;
• who decides what, written down before go-live;
• what comes next organised: an internal lead at your company, and Monitoring offered.

The rules are written in advance, depending on what is blocking:

• an access right or a planned piece of information is missing: the build is paused until it arrives. That is why access is obtained before the start;
• an audit concludes that the build is not feasible: we stop there, and you keep the file you have paid for;
• the scoping of a Process Agent shows that it needs the knowledge of its entire business domain: we tell you before going further, and propose Business Agents. You decide.

One to two weeks for the Process Audit. The Domain Audit takes two to three weeks for two processes, three weeks for three, three to four for four, and four weeks for five; beyond that, the timeframe is set out in the proposal. The AI Usage Audit is delivered within five working days.

The processes examined are chosen with you during the Flash Diagnostic. The Process Audit and the Domain Audit end with a presentation of the findings.

Yes, every day. Our own Company Brain brings together our decisions, offers, prices, rules and procedures in a single reference base, dated, kept version after version. Our in-house agents rely on it, one per role: a CEO advisor who prepares our decisions and challenges them, a business analyst, a solution architect, and others for documents, prospecting, finance or legal matters. A person approves what they produce.

This site is drawn from it: every time it is built, it checks that every price shown appears in our price list.

Through two forms, or by phone. For a diagnostic, go straight to the Flash Diagnostic questionnaire: it prepares the interview. For any other question (an offer, a partnership, the press, a job application), use the contact form: we read it and reply by email. Our phone number is on the Contact page and at the bottom of every page; you can also find us on LinkedIn.