There’s a number haunting the artificial intelligence space: 95%. As in, 95% of generative AI pilots are failing, according to MIT’s influential, arguably overblown research study in August 2025. When Fortune’s Diane Brady spoke with PwC global chairman Mohamed Kande roughly six months later, in Davos, Switzerland, that number was stubbornly high: 56% of CEOs PwC surveyed were getting “nothing” from their AI adoption efforts.
The solution is peace, love, understanding, and good parenting skills, according to Nvidia CEO Jensen Huang. The $4 trillion market-cap man arrived at the Cisco AI Summit with a message that sounded less like Wall Street rigor and more like a blend of 1960s counterculture and modern parenting: “Let a thousand flowers bloom.”
Sitting down with Cisco CEO Chuck Robbins, Huang addressed the tension facing enterprise leaders who feel the pressure to adopt AI but fear the lack of immediate, quantifiable results. When Robbins asked for the first steps an enterprise should take, Huang dismissed the immediate fixation on spreadsheets.
“I get questions like … ROI,” Huang said. “I wouldn’t go there.”
Instead, he advocated for a philosophy of abundance and messy experimentation, explicitly comparing corporate innovation to raising children. He argued that demanding proof of financial success before allowing an engineer to try a new AI tool is as stifling as asking a child to justify a hobby with a business plan.
“I want the same thing for my company that I want for my kids: Go explore life,” Huang explained. When your kids tell you they want to try something, he added, you should say yes. We never ask questions at home like: What is the return on investment here? How is this going to lead to financial success? How can you prove to me that it’s worthwhile? “We never do that at home. But we do it at work.”
Innovation needs therapy, not control
This approach requires executives to relinquish a degree of command that might feel uncomfortable, Huang admitted, arguing that creativity and innovation make it worthwhile. “The number of different AI projects in our company is—it’s out of control, and it’s great,” he said, remarking that innovation doesn’t always happen when you’re in control. “If you want to be in control, first of all, you’ve got to seek therapy. But second, it’s an illusion. You’re not in control. If you want your company to succeed, you can’t control it.”
Huang argued that to succeed, leaders must seek to influence their companies rather than control them. The logic behind letting “a thousand flowers bloom” is risk management through diversification. While this method “makes for a messy garden,” he said, it prevents the error of committing resources—“putting all your wood behind one arrow”—too early in a technological shift where the “winning” tools are not yet obvious.
Lift the hood
While advocating for a relaxed approach to ROI, Huang was adamant about the necessity of “tactile understanding.” He urged leaders not to rely solely on cloud rentals or finished products.
Computers are everywhere these days, he said, but if you built one yourself, you’d still get a better understanding, just like a good car owner wouldn’t take Uber all the time but would look closely at their engine. “Lift the hood, change the oil, understand all the components,” he said. “Build something. You might discover you’re actually insanely good at it. You might discover that you need that skill.”
He stressed that because AI technology is vital to the future, companies must build some infrastructure on-premise to truly understand how the “components” work. This relates to data privacy and what Huang calls the most valuable intellectual property: the questions. “The most valuable IP to me is not my answers… they’re my questions,” Huang said, remarking that answers are a commodity, but smart questions are irreplaceable.
Fortune recently visited KPMG’s Lakehouse in Orlando, where the firm was rolling out its AI training framework, first with interns and then firmwide. “Think, prompt, check” was how they were training employees to work with AI, stressing the first and last points as something not to take for granted.
From explicit to implicit
The urgency for this experimentation, according to Huang, stems from a fundamental “reinvention of computing.” The industry is moving from “explicit programming”—writing specific lines of code—to “implicit programming,” where users state their intent, and the AI figures out the solution.
In this new world, “typing is a commodity,” Huang noted. The true value lies in the domain expertise required to guide the AI. “You now tell the computer what your intent is, and it goes off and figures out how to solve your problem.”
Huang closed by flipping the popular ethical narrative of “humans in the loop” on its head. The goal, he stated, should be “AI in the loop.” By integrating AI into every process, companies can capture the “life experience” of their employees, turning daily work into permanent corporate intellectual property. In other words, they’ll be letting a thousand flowers bloom, but only if they have the right curiosity, the right questions, and the right support from above to think freely.
For this story, Fortune journalists used generative AI as a research tool. An editor verified the accuracy of the information before publishing.
