French companies adopt AI but make mistakes: starting from tech instead of the problem, neglect the quality of the data and forget about successful metrics.
Artificial intelligence is now part of the daily life of French companies. But behind the ambient enthusiasm hide the pitfalls revealed by the feedback from the first adopters.
Forget the AI to focus on the problem
“Forget about Ai, What is your biggest problem?” This question of Rick Rioboli, CTO of Comcast, should be engraved on the pediment of each computer direction. Too many French companies start with technology – “we have to do AI” – rather than real need. Result: gadget projects that impress in demonstration but do not bring any business value and above all do not produce any king.
The most logical approach would first consist of identifying a problem with high impact business, then determining what solution IA can actually solve it. An inversion of perspective which seems obvious but remains a minority in fact.
Data, nerve of war
The message is clear: it all starts with an orderly database before dreaming of AI, start by auditing your data: “Without unified and accessible data, even the most advanced generative initiatives will be able to deliver real value.”
Generative AI models, trained on the Internet, do not know your owner data. Hence the crucial importance of the RAG (Retrieval Augmented Generation) to contextualize AI with your business information. But beware of the security and governance challenges that this implies, often underestimated by the project teams.
Metrics and safeguards: the forgotten of AI
Define from the start, it is necessary to define quantifiable success indicators: NPS, response time, resolution rate … without clear metric, impossible to distinguish the true success of technological mirage. And continuously monitoring: 74 % of companies using generative AI find a return on investment, but only among those following their indicators closely.
Governance, risk management, regulatory compliance … with the gradual entry into force of AI European Act, it is better to anticipate than undergo. Otherwise, organizations risk multiple compliance problems if they do not understand how AI data has been generated.
The French challenge of AI
These lessons resonate particularly in France, where our companies often accuse a delay in technological adoption (10% of companies located in France use AI-related technology against 13% on average in Europe, and even 28% in the Netherlands, according to an INSE study). But this delay can become an advantage: that of learning from the errors of the pioneers.
Companies that adopt “intelligently” AI claim 1.5 times more growth. But “intelligently” involves going beyond the fashion effect to build a solid, ethical and lasting IA strategy.
AI is not a race, it is a marathon. And in a marathon, it is better to leave with the right strategy than the fastest. Our French companies have this opportunity: that of creating scalable, transparent and adapted to their specific needs.
It is time to go from the AI show to the useful AI. Will our leaders know how to seize this chance?
