Jason Wild: "We are very good at solving problems, and terrible at choosing which ones"

In Stockholm, a bar attempted to entrust management to an AI agent, complete with the title of CEO. The experiment, Jason Wild recounts (met at WOBI ON - AI & Business Transformation in Milan on June 9), produced an instructive result: the agent negotiated energy contracts with the coolness of a CEO, then ordered senseless quantities of canned tomatoes and forgot the bread. The balance sheet, in the numbers cited by Wild: 10,000 dollars in sales and about 25,000 dollars lost. The diagnosis: the agent had plenty of intelligence, but what it lacked was context, a commodity much harder to procure.

Wild has accumulated quite a bit of context. Thirty years leading innovation for IBM, Microsoft, and Salesforce, projects in 40 countries, clients ranging from NATO to Disney to the mayor of Rio de Janeiro, dealing with World Cups and Olympics. Today, he collaborates with Spotify and is co-author of Genius at Scale, released in March, a result of ten years of research on leadership. He had already been to Italy twenty years ago for an IBM project on Autogrill: his boss at the time told him that to innovate here, they would need three PhDs: in psychology, in transformation, and in luck. He tells me this and still laughs.

The Eliza Effect and Harari’s Candle

On stage, Wild dismantled with a certain satisfaction the articles, including those from Harvard Business Review, that describe AI as a "teammate", a digital colleague. The risk even has a historical name: the Eliza Effect, named after the chatbot built by Joseph Weizenbaum at MIT in the 1960s, referring to the human tendency to project human traits onto systems. When the system becomes a colleague, the responsibility for decisions begins to slide off, and that’s where organizations get hurt.

Reducing it to a simple "tool" seems equally shortsighted to him because it denies the technology's autonomous potential. The frame he proposes is that of infrastructure, and the metaphor he uses comes from futurist Oren Harari: electric light did not arrive through the continuous improvement of the candle. An infrastructure opens the door to entirely new products, and those who use AI to fine-tune existing processes are essentially perfecting candles.

From Doing to Deciding (and the middle management thanks)

The most contrarian part of his talk concerns middle managers. Jack Dorsey, co-founder of Twitter, has hypothesized their elimination. Wild argues the exact opposite: with execution becoming increasingly automated, the focus of managerial work shifts from doing to deciding, and decisions move from a monthly to a daily cadence. Someone has to absorb that decision-making weight, and it will fall right on intermediate levels.

On costs, Wild presented numbers that cool the enthusiasm from slides. Based on Goldman Sachs data, outside of programming (where Dario Amodei predicts real breakthroughs with autonomous agents), costs between AI agents and humans in roles like data entry and call centers are almost on par. A Deloitte finding captures the real problem: 93% of IT budgets go to technology, while only 7% goes to people. For Wild, the competitive advantage has shifted precisely to that overlooked 7%, because using AI solely to cut costs means playing the wrong game.

Architect, Bridger, Catalyst

The framework of the book identifies three roles that enable innovation. The Architect builds culture: an example is a middle manager at Pfizer who delivered the vaccine supply chain in 266 days, against the industry norm of 8-10 years. Or Ajay Banga, who took Mastercard where innovation was the last priority and multiplied its capitalization by twelve by focusing on the unbanked.

The Bridger is the one Wild is most fond of: "the unsung hero of innovation," he tells me in the interview. It’s someone who connects the dots between sales, technology, and market—the role that in tech today is called Forward Deployed Engineer. And it’s the reason why, in his words from the stage, "most innovation doesn’t die in the ideation phase. It dies in the integration phase, human and systemic." Companies that lay off bridgers to meet a quarterly target, he adds, are sawing off the branch they’re sitting on.

The Catalyst, finally, transforms a company into a movement and the movement into an ecosystem. The examples he cites are Nvidia and Salesforce. Over all this is a shift in leadership paradigm: the old "pathfinding," charting a vision and making it follow, no longer holds because assumptions change too quickly. What is needed is what he calls wayfinding: "how do you navigate when you’re surrounded by fog, when destination and path aren’t clear?"

Human Judgment as a Bottleneck

In the interview, the tone shifts; it becomes more personal. Wild starts from a technical observation: language models are trained on publicly available data generated by humans, following the normal intelligence curve, making them "fundamentally average." This leads to an obsession with prompt engineering (which he amusingly compares to the Matrix pill) and the rush for proprietary data as a differentiator. The "collective genius" referenced in his book emerges when AI powers an entire team, whereas hunting for isolated single talent yields much less.

Then comes the phrase that became the title of the interview: "we humans are really good at solving problems but terrible at selecting them." The danger of AI, he says, is its accommodating nature: a system that caters to you can convince you that you’re solving the right problems even when you're not. This is why human judgment and the ability to contextualize remain irreplaceable, to the point that Spotify has just included "human judgment" among its core corporate values. A tech company that writes something like this in 2026 is saying something about the market, even before saying something about itself.

There is also a Wild as a father, concerned about the trend of eliminating entry-level jobs: cutting ramps of access means forfeiting the next generation of leaders and impoverishing diversity of thought. And there’s a curious discovery made during the research for the book: he, his co-authors, and several of the leaders studied, including Ajay Banga, attended Montessori schools. Curiosity, independent thinking, and the idea that failure is a means to learn, he says, produce better leaders than any MBA.

Advice for Italian Companies

To a medium-sized Italian company, Wild suggests leveraging a local talent he appreciates: the propensity for debate. The practical tool is a single question: "what would you never say about your business?" He cites an American supermarket chain that, by answering ("no one would buy a last-minute gift from us"), launched a gift card business that today is worth three times the core business. This is what Wild, echoing Steve Jobs, calls vuja de (the opposite of deja vu): looking at the same information as everyone else and seeing it differently (the original coinage belongs to comedian George Carlin, but the essence holds just the same).

The other warning concerns the "theater" of innovation, that is innovation done for show, which does not bring real change: workshops, hackathons, and various activities that do not touch operational models. True innovation produces measurable results, and the right metrics according to Wild measure learning and collaboration, not the hundreds of financial indicators that large organizations adore. From startups, he steals a precise practice: an "obstacle dashboard," which puts problems on the table and makes difficult conversations a matter of data rather than blame.

I close with the anecdote that opened his talk. At the first projection of the Lumière brothers, in front of the oncoming train, the audience fled the room, or so the legend says, which film historians have debunked for decades without managing to chip away at it. Wild argues that that terror was a healthy reaction, and that it should be today in the face of AI: "if we focus on how frightening it is, perhaps it will motivate us to do the right things regarding governance."

Between those who minimize and those who catastrophize, a bit of well-placed fear seems to me to be the most concrete advice that emerged from this conversation.