DIRIGEO: Jean-Bernard Thonus wants to make AI the strategic right hand of SME leaders

SME leaders often face the same legal, financial, social, or strategic issues as large organizations, without having the same internal resources to address them. With DIRIGEO, Jean-Bernard Thonus wants to use artificial intelligence to reduce this asymmetry. His ambition: to create a “strategic right hand” capable of helping leaders understand a situation, identify risks, and prepare their decisions, without replacing them. A proposition that raises a central question: how far can AI enter a company’s decision-making process?

For several years, artificial intelligence in business has primarily been viewed as a productivity tool. It writes, summarizes, translates, analyzes data, automates tasks, or produces code.

Another stage is now opening. AI is gradually beginning to intervene upstream of action, at the moment when it’s necessary to understand a situation, compare several scenarios, and prepare a decision.

This is precisely the ground that Jean-Bernard Thonus wants to occupy with DIRIGEO. The entrepreneur is developing his project around one expression: making artificial intelligence the leader’s “strategic right hand.”

In the definition proposed by DIRIGEO, a governance AI is a system designed to help a business leader analyze a situation, identify its risks, understand its environment, and structure several options before deciding.

This ambition starts, however, from a much less technological reality: the loneliness of the leader.

SME leaders facing the loneliness of decision-making

Jean-Bernard Thonus comes from entrepreneurship. During our interview, he reflects on his background, his business studies in Paris and Oxford, his experiences in wealth management, then in technology, as well as his responsibilities as a local elected official. But it’s primarily his experience as a business leader that structures his thinking.

With his partners Patrick Anschutz and Yann KERROS, he says he started from a simple observation:

“Every time we ran companies of different sizes, when we were in small structures, systematically, we were confronted with the problems of large ones. But we didn’t have their resources. DIRIGEO is the AI we would have liked to have in those moments”

This difference in resources is fundamental.

A ten-employee company can encounter an HR problem, a commercial conflict, a tax question, or a regulatory constraint whose complexity is not necessarily proportional to its size.

Jean-Bernard Thonus summarizes:

“The business leader of a ten-person SME or a 500-person mid-sized company will be confronted with identical HR, regulatory, and legal problems.”

The difference lies in what happens next.

In a large organization, the leader can transfer the file to a functional department.

“The boss, who encounters this situation in a 500-person mid-sized company, has his general manager who transmits to his legal department. He tells them: in three hours, I need your analysis.”

In a small company, the leader turns around. He often has neither a legal department, nor a financial director, nor an HR director, nor a team of consultants permanently available. And he still must decide.

The “evening inbox,” symbol of the leader’s invisible burden

Jean-Bernard Thonus uses a particularly revealing expression: the “evening inbox.”

The leader starts his day with what makes his company run. He meets his clients, supports his employees, responds to suppliers, develops his business, manages his markets.

The other problems accumulate, whether it’s a contract, a dispute, a tax question, an HR issue, a new regulation, or a financing decision: they don’t disappear, they wait.

“The boss of an SME often experiences a high level of ‘business stress’ related to his company management and not to his trade,” explains Jean-Bernard Thonus.

The files are then postponed.

“He piles them up in the evening or weekend inbox,” he says.

Then comes this very concrete formula:

“When evening or the weekend arrives, he knows he has the inbox of ‘hassles’ to deal with, without having anyone to help him.”

The vocabulary is direct. But the image is powerful.

It describes an invisible part of the leader’s job: a cognitive load consisting of peripheral issues to the core business, but whose consequences can be considerable. It’s this “evening inbox” that DIRIGEO wants to tackle.

DIRIGEO wants to develop a governance AI for SMEs

From this observation, Jean-Bernard Thonus and his partners asked themselves a question. What if artificial intelligence could provide a small business leader with part of the analytical capacity that large organizations have?

He thus recounts the genesis of the project:

“We said to ourselves: what if we created the first French governance AI, that is, one that will be the leader’s strategic right hand?”

The expression “first French governance AI” here reflects the positioning claimed by DIRIGEO. In a market where specialized assistants, decision copilots, and agentic systems are multiplying rapidly, such primacy remains difficult to establish independently.

But the term “strategic right hand” allows for a better understanding of the proposition.

The objective is not only to answer a question, it’s to structure a problem.

“Instead of saying: ‘how am I going to manage when I’m alone facing this?’ the leader will be able to submit his problem, understand what’s happening to him, have an action plan, have the context, the risks, the deadlines in which he needs to make his decisions.”

This sentence is probably the heart of DIRIGEO.

AI no longer seeks only to execute. It intervenes in the preparatory phase of judgment.

An AI to understand before consulting an expert. Jean-Bernard Thonus, however, insists on an essential distinction. DIRIGEO should not, according to him, replace a lawyer, an accountant, or a specialized professional. The AI intervenes upstream.

It must enable the leader to better understand his problem before requesting human expertise.

“When he goes to see his advisor because he needs to, he will need that person’s added value immediately to find solutions for him and not to explain the situation he’s in.”

The economic proposition is interesting. Part of the time devoted to an external consultation indeed consists of qualifying the problem, collecting useful information, and putting the situation in context.

DIRIGEO intends to reduce this phase. Jean-Bernard Thonus thus estimates that the user can “save the first 300 euros of the consultation” because he arrives before his advisor with a first understanding of the file. He adds:

“As soon as you’ve asked a question regarding a governance issue for which you have a structured answer, you’ve already saved the first external consultation you would have needed.”

This statement constitutes a value promise carried by DIRIGEO. Its economic effect will naturally need to be evaluated according to the complexity of situations and the actual quality of responses.

On the most sensitive legal, tax, or social issues, the final responsibility will remain that of the leader and the professionals concerned.

Why a specialized AI rather than ChatGPT?

The existence of extremely powerful generalist tools, however, raises an obvious question. Why would a leader use a specialized AI?

Jean-Bernard Thonus identifies two main differences: context and data.

A generalist AI can produce a relevant answer to a generic question. But a business decision often depends on information specific to the organization.

“If you want to have relevant analysis, you need to put your company’s strategic data in it,” he emphasizes.

Contracts, bylaws, financial information, HR elements, or tax documents may become necessary.

DIRIGEO therefore wants to build a contextualized memory of the company.

This evolution constitutes a major challenge for professional AI systems. A generic AI knows a lot about companies. A contextualized AI begins to know something about this specific company.

And it’s this change that can transform the nature of decision support.

Data, sovereignty, and contextualization of artificial intelligence

Jean-Bernard Thonus indicates that DIRIGEO relies on Mistral and on an infrastructure deployed notably at OVHcloud.

“We based it on Mistral,” he explains.

The solution also relies, according to him, on a proprietary professional corpus and on system prompts developed specifically for governance use cases.

Client documents are hosted in individualized spaces.

“They store their documents in digital vaults that we open in their name.”

The objective is to allow the system to use this information as context.

Jean-Bernard Thonus particularly insists on confidentiality:

“The strategic data you give me won’t go to your neighbors, to your competitors.”

DIRIGEO thus claims a sovereign and secure approach.

This dimension, however, deserves to be treated with precision.

Digital sovereignty is never reduced to the nationality of a model or the location of a server. It also depends on subcontractors, contracts, data flows, technological dependencies, and jurisdictions likely to apply.

For an AI designed to handle strategic information, trust will therefore need to be demonstrated as much as claimed.

What use cases for an AI designed for leaders?

The examples presented by Jean-Bernard Thonus show the scope of the ambition. He notably mentions companies faced with social, regulatory, commercial, or international issues.

One case involved professionals who import wood from Vietnam.

“We have use cases about people who import wood from Vietnam who asked us how to deal with freight forwarders, with the European Union, to bring this wood in.”

Another involved a Korean company wishing to introduce its cosmetic products in France.

“A Korean company came to see us saying: ‘I want to establish my cosmetics range in France, but I don’t know how to do it.'”

According to Jean-Bernard Thonus, DIRIGEO then worked on several dimensions: product, distribution, commercial establishment, subsidiary, and tax scenarios.

He summarizes the scope he wants to give his tool in an ambitious formula:

“We have a tool today that is a real general management for someone who is alone in his company.”

This is also one of the points that raises the most questions. The more fields a tool claims to cover, the more complex reliability control becomes.

The challenge will therefore not only be to provide an answer. It will be to know when the answer is robust enough to inform a decision, and when the system must explicitly acknowledge its limits.

An agentic AI that doesn’t want to replace the business leader

DIRIGEO uses agentic mechanisms in its architecture.

Jean-Bernard Thonus notably mentions “an agentic system prompt” designed to cover different areas of governance.

But he refuses to position his company as a simple provider of autonomous agents.

“We don’t develop the agent. We develop an AI that doesn’t replace the business leader.”

This distinction deserves attention.

A large part of the current market seeks to automate execution: responding to emails, handling calls, prospecting, generating documents, or triggering processes.

DIRIGEO claims to want to intervene elsewhere, not primarily in execution, but in the reasoning that precedes it. This is where the subject becomes much more sensitive. Because an AI that executes can be controlled by its result.

An AI that influences reasoning acts earlier, sometimes even before the leader has actually formulated his own analysis.

The 4S Circle: putting human intelligence back around artificial intelligence

The project has another dimension which, in my opinion, constitutes one of its most interesting elements. DIRIGEO wants to associate its AI with a network of business leaders. Jean-Bernard Thonus explains that subscribers should be able to be connected with each other to share experiences, difficulties, and solutions.

He summarizes this philosophy with a particularly strong sentence:

“The idea is to share human intelligence, in addition to sharing artificial intelligence.”

This is probably where the proposition goes beyond the simple digital product.

In a world where the same large models are gradually becoming accessible to all companies, differentiation could depend less on the AI itself than on what surrounds it: expertise, proprietary data, community, trust, and methods.

Jean-Bernard Thonus insists:

“We keep humans at the center of all this.”

Then he formulates DIRIGEO’s ambition:

“People who, today, lead alone will no longer lead alone tomorrow. They will be augmented and accompanied.”

This sentence touches on a much broader question than that of technology. It concerns the isolation of the leader.

A business model based on trust and peer-to-peer

This desire to preserve a human relationship also appears in the distribution model.

Jean-Bernard Thonus doesn’t want to sell DIRIGEO exclusively via the Internet.

“We market DIRIGEO through exclusive agents, territory by territory, who are networked business leaders, who irrigate peer-to-peer.”

He compares this organization to certain networks developed in real estate.

The paradox is interesting.

A company that markets an artificial intelligence chooses to rely on a human network to distribute it.

But this choice corresponds to the very nature of the product.

A solution that intervenes in strategic issues must create trust.

The issue is no longer simply buying software.

It’s about deciding whether one accepts that a digital system enters an intimate part of one’s company’s functioning.

DIRIGEO also wants to work on AI sobriety

Jean-Bernard Thonus also mentions a future technical evolution of DIRIGEO.

He announces a “self-hosted architecture in more than 80% of responses.” According to him:

“In 80% of responses, it won’t use tokens.”

The stated objective is twofold: controlling costs and reducing energy consumption. Jean-Bernard Thonus believes that not all problems require systematically querying heavy models. The intuition is relevant. An architecture capable of directing each query to the level of computation actually necessary probably constitutes one of the major paths to improving the economic and environmental efficiency of AI systems.

The figure of 80% nevertheless remains to be considered as an objective announced by DIRIGEO. It will need to be demonstrated in real conditions, as well as the associated energy savings.

Can a governance AI be internationalized?

Jean-Bernard Thonus finally envisions expansion outside France. But he doesn’t simply want to translate the French solution.

“It’s not about asking people in Spain to take a French AI, it’s about developing a sovereign Spanish AI for them.”

The project would therefore consist of gradually building local implementations adapted to each ecosystem. This approach is consistent with a governance AI. Business decisions are deeply linked to law, taxation, social practices, and the economic environment of the country concerned.

Jean-Bernard Thonus is moreover cautious about the pace of this internationalization. He explains wanting to choose sufficiently mature markets, find local partners, and gradually deploy the model rather than seeking an international presence for its own sake. This caution will be necessary. Because while technology can be replicated relatively quickly, the institutional and cultural context cannot.

From AI that produces to AI that influences decision-making

DIRIGEO remains a young company. Its business model will need to be tested, the quality of responses will need to be evaluated, and several technical promises will need to be confronted with actual usage.

But Jean-Bernard Thonus’s proposition deserves attention because it reveals a more general change.

We are gradually moving from AI that produces to AI that advises.

And this transition changes the nature of the problem. A tool that writes an email can be corrected. A system that helps a leader determine possible options intervenes much more deeply in his reasoning.

This is where the real challenge of decision-making AI lies.

Can a leader truly remain master of his judgment when a system progressively knows his contracts, his data, his organization, his problems, and his history?

An algorithmic recommendation presented in a structured and convincing manner can create an automation bias. The leader may attribute more credibility to it simply because it appears coherent, documented, and fast.

The quality of a governance AI should therefore not be measured only by its ability to provide answers. It must also be evaluated on its ability to signal uncertainty, present contradictory options, make its limits visible, and encourage contestation of its own conclusions.

Because governance doesn’t consist of having a single right answer. It often consists of arbitrating between several imperfect decisions. This is where human responsibility becomes central again.

Jean-Bernard Thonus clearly states:

“We develop an AI that doesn’t replace the business leader.”

This sentence will need to constitute more than a communication principle. It will need to become a design principle. The challenge of tomorrow will probably not be to determine whether leaders will use artificial intelligence to decide. It will probably be more only to determine whether leaders will use artificial intelligence to inform their decisions, but how far they will accept integrating it into their reasoning process.

The real question will then be to know how they will preserve their intellectual autonomy when they work daily with systems capable of structuring part of their analysis.

DIRIGEO starts from an ambition: ensuring that those who “lead alone” are no longer alone.

The proposition is attractive. But it immediately opens a more demanding question.

When artificial intelligence becomes the leader’s right hand, how can we ensure that it never becomes, gradually and almost imperceptibly, the one that guides his hand?