Jean-Roch Houllier (SKEMA 2018)

“THE EXPERT IS NO LONGER OMNISCIENT;

KNOWLEDGE HAS BECOME

COLLECTIVE”

SKEMA graduate Jean-Roch Houllier is Head of Operations, Learning & Digital at Safran University, the Safran Group’s corporate university. This flagship French high-tech company, which employs 110,000 people, operates in critical sectors like aerospace and defence. At a time when artificial intelligence (AI) is revolutionising specialist roles, Houllier fosters an approach based on critical thinking, cognitive autonomy and “invariants” shaped by work experience. Between “substitute” and “confrontational” AI, he advocates for enhanced expertise that can structure analysis, resist the siren call of the easy option, and mobilise the community to tackle crises. We talk to him.

Interview by Antoine Boitez

What does it mean to be an expert at Safran now?

At Safran, expertise remains closely connected with technical fields, particularly in R&D. An expert is someone whose skills are recognised within and outside the Group through tangible contributions like publications, patents and seminal work.

Safran has four levels of technical expertise, which are organised and recognised in the same way as a management career path, right up to the highest levels of the Group’s expertise. This shows how important this approach is in a technology group like ours.

Is artificial intelligence disrupting this definition?

Above all, it requires us to clarify what involves competence and what involves judgement. AI can speed things up, generate summaries and automate certain tasks. Meanwhile, human expertise is the ability to put things into context, prioritise and make decisions under pressure. In an industrial world, responsibility cannot be delegated to a machine.

How does Safran regulate the use of AI?

We are moving forward in a very structured manner. Safran adheres to a code of conduct and makes a clear distinction between market tools that can only be used with low-sensitivity data, and secure environments that enable a more in-depth approach. Confidentiality is a key issue.

There are two stages in the roll-out: firstly cultural integration, then the phased introduction of use cases. We have launched internal programmes, worked on prompt libraries and organised “promptathon” initiatives, starting off from business requirements and associated use cases rather than tools. Our approach is based on progressiveness and complementarity: we are not interested in the idea of replacement.

In this context, what remains the same?

There are certain invariants even in changing professions. In project management, for instance, being able to read a schedule, identify the critical path, manage a risk matrix or take a step back to assess a situation are all fundamental skills. Alongside this, changes are bound to happen, of course. AI is transforming certain roles, starting with those of the Project Management Office (PMO). AI is set to become a powerful training partner for project managers, as it will be able to handle various support, analysis and structuring functions.

You differentiate between two forms of AI: “substitute” and “confrontational”. Why is that?

At Safran, we approach AI in terms of complementarity and augmentation. The key point is that we are not pursuing a policy of replacement, nor are we solely focused on performance. “Substitute AI” is based on the principle of delegation: it acts on your behalf, produces results, and speeds things up. It may be useful, but it carries a risk. If we push the “delegation slider” too far – towards the illusory easy option – it inevitably leads to intellectual impoverishment coupled with a form of dispossession that raises the question of recognition for the work carried out.

Conversely, “confrontational AI” – the approach we favour – aims to encourage critical thinking. It is designed as a training ground, not a crutch. It’s an AI that forces us to justify our arguments, rephrase our thoughts and challenge what we think we’ve understood. This is a learning approach designed to preserve cognitive autonomy: whereas “substitute AI” empties the mind, “confrontational AI” exercises it.

In an industrial world, responsibility cannot be delegated to a machine.

What does this “confrontational AI” actually look like, and how does it work in practice?

At Safran, we are working on training chambers and scenario-based exercises designed to achieve specific learning objectives. In this particular case, the learner interacts with an AI agent – for example, a character called “Lucas” – who presents a problem about priority management. The task is first to make the learner aware of the problem, then to convince them of the benefits of the Eisenhower Matrix (which the learner has already studied), while ensuring they understand how it works. The agent is not generic: we programme it so we can assign it specific behaviours, including a degree of resistance or bad faith.

To question what AI proposes, you must have travelled the path yourself.

At the end of the exchange, a review of the conversation helps to assess the skills used. What makes this a learning experience is its customised nature. What one person experiences with Lucas won’t be exactly the same as a colleague; every interaction is different. The advantage is that it creates a shift towards the real world: the exercise prepares you for real-life work situations, as if Lucas were going to challenge you in person the next day.

You often talk about critical thinking and cognitive autonomy. Just how indispensable are they? Are they the “safeguards of judgement”?

Critical thinking is a central invariant; in the age of generative AI, that’s a fact. Since the technological “Big Bang”, the focus has been very much on tools and performance, while the societal and human aspects are taking a while to pick up the pace. Now we are starting to put things into perspective and give more thought to the cognitive and organisational impacts. Critical thinking is not something that can be improvised; it can take years to develop because it relies on knowledge that has been internalised, assimilated and reworked, enabling us to form an opinion, make judgements and take a stand.

It is precisely at this point that AI poses a major challenge: if we succumb to the illusory easy option and the instant, thoughtless click, we remain helpless. We no longer have any control over what the “machine” is capable of producing. To be able to question what AI offers, you have to have travelled the path yourself: to have coded, written and experimented. All of this helps, over time, to gradually and slowly develop the “discerning palate” that enables us to scrutinise, challenge and above all understand.

Is AI changing the role of experts across the different generations?

We can see that the best ways of using AI often come from those who have already “knocked about a bit”. Older people have a “discerning palate”: they know what they are delegating, and above all what they need to check. A junior employee might be tempted to click without thinking, due to lack of experience.

But I don’t believe there is a generational divide. Every generation brings something to the table: skilfulness, experience or depth. The key lies in complementarity and a shared ability to produce meaning. Training is not just about building skills: it is also a source of meaning for all the generations working together today.

In a crisis, the first thing to mobilise is an analytical approach.

In a world saturated with tools, data and AI, what still sets an expert apart when a crisis strikes?

In a crisis, the first thing to mobilise, by exercising critical thinking, is an analytical approach. You have to tackle the root cause to sort out the problem. If you misinterpret the causality (for instance, by making a simple correlation), you apply superficial fixes and the problem will come right back in your face. A project manager who can identify causes and correlations and structure an analysis immediately puts a client at ease, because they give the impression of being in control of the situation. Another aspect is the art of discernment, which is crucial for gaining perspective and taking a step back from situations.

But there is also a third, decisive factor: humility. It is no longer possible to be omniscient within organisations. Knowledge has become collective. Experts or managers who acknowledge that they don’t know everything, who rely on others and delegate, are also helping themselves: those who try to do everything end up shouldering all the pressure as well, and are often the first to burn out.