Every AI discussion eventually lands on the same anxious question: what part of the job is still defensible?
In a new Harvard Business Review article, Goldman Sachs CIO Marco Argenti frames that question through a senior banker wondering which portion of his work AI will never be able to do. It is a reasonable fear. It is also the wrong starting point.
The real advantage is not finding a tiny protected corner of work. It is becoming the kind of person who can keep redesigning the work as the tools change.
The skills list will keep expiring
For years, the standard career advice was to learn the next hard skill: coding, analytics, automation, prompt engineering, data visualization, whatever sat closest to the current wave of leverage. That advice was not wrong. Skills matter. But in the AI era, any narrow skill can become less scarce quickly.
That does not make people less valuable. It changes where value lives.
If a task can be clearly defined, repeated, measured, and improved through feedback, AI will keep moving deeper into it. The worker who survives by guarding a fixed task list is standing on weak ground. The worker who learns how to ask better questions, test assumptions, connect domains, and make judgment calls in messy situations is building on something stronger.
AI compresses the shelf life of technical competence. Mindset determines how fast someone can renew it.
Curiosity becomes operational leverage
The most useful people in an AI-powered company will not be the ones who memorize every new tool. They will be the ones who keep asking what the tool changes.
What should no longer take a week? What customer problem can now be solved earlier? What internal handoff should disappear? What decision used to require a specialist but can now be prototyped by a generalist with good judgment?
Those are not software questions. They are business questions.
This is where Argenti’s mindset argument lands. The strongest employees will treat AI less like a replacement threat and more like a force that exposes how work actually works. They will be willing to experiment, unlearn, reframe, and move before the org chart catches up.
The leadership problem is cultural, not technical
Companies still talk about AI adoption as if the central issue is access: buy the platform, train the team, publish the policy, and wait for productivity to rise. That is too shallow.
The harder question is whether the organization rewards the behaviors AI makes necessary. Do people get credit for rebuilding workflows, or only for completing inherited tasks? Are teams allowed to admit that a process is obsolete? Are managers comfortable when junior employees use AI to challenge old assumptions? Is experimentation treated as real work, or as a side activity after the calendar is already full?
A company cannot skill-train its way out of a culture that punishes adaptation.
This matters because AI does not merely automate work. It changes the boundary between thinking and execution. A person with a clear problem, strong taste, and a willingness to iterate can now produce drafts, analyses, prototypes, and strategic options at a speed that used to require a team. That makes mindset visible. Fast.
The real takeaway
The question is not, “What can I do that AI can never do?” That framing is defensive. It assumes the future is a shrinking circle of human-only tasks.
A better question is: “What can I become better at because AI now handles more of the mechanical load?”
For individuals, that means developing judgment, taste, curiosity, resilience, and the ability to learn in public. For leaders, it means designing teams where people are expected to challenge processes instead of simply perform them.
The winners will not be the people with the longest list of AI tools in their resume. They will be the people who can look at a changing system and calmly ask: what is possible now that was not possible before?
That is not a skillset. It is a posture.
Source: Harvard Business Review