On the occasion of the publication of the study “ The French, agentic AI and customer relations: state of play in 2026″, we spoke to Laurent Nicolas-Guennoc, Marketing & Communications Director at Converteo. He looks back at the key findings of this survey, between the growing adoption of AI tools, strong expectations for humanization in customer relations and the necessary conditions for deploying agentic AI that is useful, reliable and creates value for brands and their customers alike.
Question 1: Agentic AI is on everyone’s lips. How can we define it simply? And above all, how does it change the relationship between brands and their customers?
Agentic AI is the next stage in the maturation of AI: we’re moving from systems that respond to systems that act.
For brands, this means a change in strategic posture. Generative AI operates in a reactive, transactional logic: one question, one answer. Agentic AI introduces a proactive and continuous logic, where it’s the system that takes the initiative: a maintenance slot booked before the customer detects the breakdown, a dispute handled and resolved without the customer having to call again, a complex purchasing path managed from A to Z, without friction…
Our advice to the brands we work with at Converteo is not to approach this subject solely from a technological angle. Agentic AI redesigns processes, responsibilities and the service promise. It’s first and foremost a strategic and data-driven project, and that’s where value is really built.
“Agentic AI redesigns processes, responsibilities and service promises. It’s first and foremost a strategic and data-driven project, and that’s where value is really built.”
Laurent Nicolas-Guennoc
Marketing & Communications Director at Converteo
Question 2: Uses are progressing fast: 64% of French people are already using AI tools. And yet, in the event of a problem with a brand, 52% still prefer to have a human on line, and only 27% say they’re ready to exchange with an AI. How do you explain this discrepancy between tool adoption and reluctance to delegate the relationship?
Massive adoption concerns AI “tools” that increase personal productivity for tasks with low emotional stakes (research, writing, translation). We remain masters of the final decision.
Customer relations are a completely different matter.
It’s a high-stakes moment, often fraught with stress and frustration, where 3 profound expectations come into play simultaneously:
- empathy: the need to feel recognized in one’s uniqueness ;
- confidence: the perceived ability to handle a complex, non-standard situation that doesn’t fit into a script;
- habit, or rather its negative corollary: repeated experiences with inflexible, disappointing chatbots have led to a lasting distrust of automated conversational interfaces.
What these 52% are measuring, then, is not a rejection of AI as a technology.
It’s a resistance to the risk of dehumanization at a critical moment. The nuance is important: it means that the lever is not AI education, but the quality of the experience delivered. Brands that deploy AI capable of recognizing its limitations and intelligently transferring them to a human will change this perception much faster than those that try to convince their customers to trust it a priori.
Question 3: In light of these results, what should be the next step for brands: testing, accelerating or reassuring consumers about the use of AI?
Our conviction at Converteo is that these three actions are not mutually exclusive: they must be sequenced. Reassure to test, test to accelerate.
Reassurance first. Trust cannot be decreed; it must be built through concrete commitments: clearly indicating to customers when they are interacting with an AI, and always giving them the option of simply switching to a human. Without these two conditions, any deployment exposes itself to mistrust.
Then, test on the right use cases. Not the most ambitious, but the most demonstrative: real-time order tracking, appointment scheduling, post-purchase information. Scenarios where AI brings real immediate value, and where satisfaction can be measured quickly.
Accelerate, finally, by moving to a logic of industrialization. Our conviction is that scaling up is not a question of technology, but of method and foundations. Accelerating means deploying a robust technical framework before multiplying use cases, to avoid the trap of PoCs that don’t hold up. It means replacing the slowness of traditional project cycles with hyper-agility, enabling us to adapt continuously to changing business models. So, acceleration is not about making “more AI”, but about capitalizing on sound foundations to transform local successes into a structuring and sustainable competitive advantage.
