At the Rase conference in Paris, two business leaders asked why AIs are always lagging behind pilots. Their maintenance starts with people, not models.
Many businesses go bankrupt. Their AI agents are working on pilots and not reaching production. That was an indirect diagnosis from Daniel Dines, UiPath’s CEO. He spoke at the RAISE CxO Conference in Paris this week.
Dines aimed for industry popularity. He said that the promise of a “world of intellectuals” that behaves like a human worker is really harmful. He called it “the biggest inefficiency in the entire AI industry”. You can’t hire an AI, give it a laptop, plug it into your systems, and expect it to replace everyone else.
He gave a sobering prediction. “AI will not be like a human,” he said. It may reach AGI form, but it won’t think or learn the way humans do. So businesses need a new class of workers.
AI suggests, humans decide
💜 for EU tech
The latest talk from the EU tech scene, a story from our genius founder Boris, and some incredible AI art. It’s free, every week, in your inbox. Register now!
Dines has it as a simple triangle: AI suggests, humans decide, and automation delivers. The catch is the suggestion. Small wastes people’s time. The buggy takes longer to fix than the job itself. That gap, he said, is why the pilots stopped.
The deeper problem is knowledge. People learn a job by connecting to a wide range of leads, from emails and tickets to a quick word with a colleague. No company keeps all that in one place, and AI can’t handle it the same way.
So Dines floated a new role: “cartographer”. Their job is to draw a “career map” for AI. That map shows business entities, existing actions, and results. Who is responsible when an agent makes a payment? Can anyone reverse it? Today, he said, no one writes that information down.
Organization is the hard part
Guy-Laurent Arpino, chief information officer of the Louis Dreyfus Company, agreed, but added a caveat from experience. The hardest part of any transformation is the organization, not the technology.
The stock trading giant has spent years cleaning up its data and writing information into the heads of its traders. That context now resides in “hands-on details” and knowledge drives, and Arpino calls it “our company’s true IP”. Only above it can agents work reliably. “The worst enemy of AI agents is dishonesty,” he said.
LDC is now rebuilding its basic “contract-to-cash” end-to-end flow, from commercial acquisition to settlement to settlement. Some of it depends on the old automation that the company built with UiPath. It also explains where agents qualify and where people should enter.
Fewer passengers, more drivers
Both men kept coming back to the people. Dines thinks one human characteristic will outlast the automatic wave: will. “I don’t really see AI coming with its own will,” he said. Businesses run on countless human systems, so the future is “fewer passengers and more drivers”.
Danger lies in the middle. Senior leaders push back, and younger employees take to it quickly, but middle managers can be frozen between the two. LDC runs a “championship” program that equips its junior staff with new tools, driving innovation from the ground up.
Why is it important
The two have maintained their sharp recruitment warning. Youth employment has fallen dramatically over the past 18 months, and Dines calls that a mistake. Cut the pipe, he said, and starve the business of its future leaders. His answer echoes his earlier warning about cutting too quickly: senior staff must be mentors.
Arpino agrees, and says LDC continues to hire young people despite the pressure on entry-level roles. The trick, he admits, is to train them without slowing down the AI. Technology is racing ahead. The slow, hard work is teaching the company, and its people, to adapt. Phakamisa explained that tension this week.
Europe’s AI ambitions share the stage with their doubts, including whether AI will deliver the growth promised by its developers.
