8
min

AI in business: France's problem isn't technology, it's delegation

Ecris par
Publié le
19/11/2026

France has never talked so much about artificial intelligence. Its leaders believe in it, its employees use it, its companies invest in it. Yet several studies published in 2026 show that it struggles more than some of its neighbours to turn ambition into results. Our view: the missing link lies not in the IT department, but in the chief executive's office. Agentic AI raises a question no tool can answer for you: do you know how to delegate?

Plenty of ambition, execution lagging behind

The figures point the same way. According to ServiceNow's Enterprise AI Maturity Index 2026, as reported by Maddyness, France scores 51/100 for AI maturity, in line with the EMEA average, but 59/100 for ambition. With a 20-point gap between what it aims for and what it delivers, it has the second-widest gap in the region. Italy and Spain each score 54 for maturity.

On the employee side, the 2026 edition of BCG's AI at Work report ranks France last among the countries surveyed, with 62% regular users. Of those users, only one in five says AI saves them at least a day a week, compared with 42% across the whole sample. Insee, France's national statistics office, puts the share of companies with ten or more employees using AI at 18% in 2025, against 20% across the European Union. The gap widens among smaller firms: according to LIFT (HEC Paris), 15% of French companies with 10 to 49 employees use AI, compared with 23% in Germany.

It is not for want of investment: LIFT notes that French IT spending is close to that of the United States. The sticking point comes when it is time to scale.

The real issue: an AI agent is a colleague you delegate to

One figure from the ServiceNow study sums it up: in France, 61% of the companies surveyed already use AI agents, yet only 9% let them run sequences of tasks autonomously. AI is present, but rarely free to act.

What sets an agent apart from a mere tool is precisely its autonomy. Handing a process over to an agent means defining a scope, a degree of freedom, checkpoints and someone accountable when things go wrong. These are exactly the ingredients of successful delegation to a person.

The three priorities the study identifies are reliable data, clearly assigned responsibilities, and testing and audit procedures. Taken together, they read almost like a specification for good delegation. Yet only 17% of the French organisations surveyed have formalised these procedures, compared with 25% in Italy, 21% in Spain and an EMEA average of 19%.

Hence the hypothesis we put forward, which reflects our own reading rather than a research finding: a leader who struggles to delegate to their team will not delegate any more readily to a machine. Researchers at NEOMA Business School describe a French management style marked by many layers of hierarchy and predominantly top-down communication. That model sits uneasily with the logic of agentic AI. AI does not create this brake; it exposes it. We call it delegation debt: a managerial liability built up over the years, which suddenly becomes expensive.

What AI will do to the leader's role

In the same BCG AI at Work report, nearly two in three managers and leaders (65%) believe that within three years AI agents could take on half or more of their tasks. For a leader, this is not just about productivity. It is a question of identity: if half of what I do can be handed to a machine, what is left of my role?

What remains is what matters most: making trade-offs, giving meaning, asking the right questions, carrying responsibility. The leader shifts from "the one who knows and signs off" to architect of trust: they design the framework within which people and agents operate, then stand behind it.

French leaders have partly grasped this. In BCG's AI Radar 2026, 40% of French CEOs link their own success to how well they lead their company's AI transformation, compared with half of CEOs worldwide. They spend just over six hours a week learning about it, while the study average exceeds eight. But technical upskilling is not enough. What is required is a change of stance, and that is the very purpose of developmental coaching.

Why this is about performance, not comfort

AI governance is not a nice-to-have. According to the ServiceNow study as reported by Maddyness, the organisations most advanced in this area achieve an average return on investment of 164%; their productivity gains are five times greater and they are 2.6 times better placed to roll AI out at scale. The article itself stresses that these are correlations, not proof of causation. The signal is nonetheless clear.

Then there is the time freed up: hours saved but never reinvested create no value. Deciding where to put them (innovation, customers, teams) is a leadership decision, not a tool's.

Four questions for your executive committee before your next AI project

  1. What do I currently refuse to delegate to my team, and why? The answer often reveals what you will refuse to entrust to an agent tomorrow.
  2. Who is accountable when an agent gets it wrong? Without a named owner, no large-scale deployment will hold.
  3. Where will the freed-up time go? Decide before the project starts, not after.
  4. How much room for error do we allow? An organisation that punishes human mistakes will also curb the autonomy of machines.

The role of executive coaching in AI transformation

Succeeding with agentic AI means working on what technology roadmaps overlook: our relationship with control, the trust we extend, clarity of responsibilities. It is individual work for the leader and collective work for the executive committee. At VISCONTI Partners, our leader-coaches support this shift from a leadership of control to a leadership of orchestration.

France lacks neither ambition nor investment. What it lacks is organised trust. The AI race will not be won by the best model, but by leaders who can loosen their grip without letting go of the wheel.

FAQ

Why are French companies lagging behind on AI? Less for lack of tools than because they struggle to scale: studies published in 2026 point to governance, data and clarity of responsibilities.

What is delegation debt? The backlog built up by an organisation that delegates little to its teams. With agentic AI, which relies on autonomy, that backlog becomes a direct brake on performance.

How does executive coaching help AI transformation succeed? It helps leaders and their executive committee evolve their stance: clarifying what they delegate, to whom and with what level of oversight, and redefining their added value alongside AI agents.

Sources: ServiceNow, Enterprise AI Maturity Index 2026, French figures as reported by Maddyness on 16 November 2026; BCG, AI at Work 2026 (June 2026) and AI Radar 2026 (January 2026); Insee Première, "Les technologies de l'information et de la communication dans les entreprises en 2025" (in French); LIFT, HEC Paris, "Adopting AI at work isn't enough to boost productivity" (2026); NEOMA Business School, "French management style" (2021).

Table of contents

3
min
Business strategy

AI in business: France's problem isn't technology, it's delegation

Publié le
18/11/2026

Other Recommended Articles

Visconti Partners presents its advice, inspiration, and case studies to help you unlock your potential and that of your business.

View all articles

Other suggested podcasts

View all podcasts