‍

By Guest Author Shoham Adizes

Will AI take away the human's job? Are we heading toward a future where AI makes every decision and humanity becomes a species with no agency, just following our superintelligent digital overlord?

That's a real fear, and it may be justified: a system with its own goals is worth worrying about. But it's not the danger we're most exposed to right now. Long before AI seizes anything, we're going to hand it over, because it's easier.

Two kinds of conflict

Management theory distinguishes two kinds of conflict. A convergent conflict resolves through discussion, like disputes over terminology or conflicting facts. Clarify the term, or do a little research, and people converge on a common understanding.

A divergent conflict can't be resolved that way: both sides are right, and mutually exclusive. For example, should we pursue increased market share, which would require us to increase our offering and decrease our prices, or should we pursue increased profit margins, which would require us to reduce our offering and increase our prices? Both increasing market share and increasing profit are legitimate goals, but we cannot pursue both goals at the same time. It's a structural trade-off, not a knowledge gap. Another example of a divergent conflict is conflicting interests. A win-win is a utopian expectation, and most of the time, at least in the short run, someone wins and someone loses. That is just the way it is, and no amount of talking will change that.

Where AI helps, and where it can't

AI will be excellent at convergent conflicts: finding the right word, surfacing missing information. That's a knowledge problem.

Divergent conflicts aren't. No amount of data will resolve conflicting, mutually exclusive goals or conflicting interests. The job isn't to optimize one side; it's to find the right balance, and today's balance won't be tomorrow's. It's like driving, constantly correcting the wheel. That correction is a judgment call, with no right answer, only risk, and it can't be outsourced to whoever is best at maximizing a single variable.

A recent example

Fauci was recently hauled before Congress, again, over pandemic-era decisions. Worth being precise: as Director of NIAID, he held real authority over research funding and institute priorities, but never the authority to set nationwide public health mandates. On the White House Coronavirus Task Force, chaired by Trump and led by HHS, he served as lead scientific advisor, translating epidemiology into policy options, not setting policy. On the decision that mattered, whether the country locked down, he was an advisor, not a decision-maker, exactly the role AI will play for us.

As an academic and government administrator, Fauci looked at COVID through one lens: minimizing the risk hospitals would be overrun. A legitimate goal, though optimizing it meant not weighing the social cost of lockdowns, the economic hit, or the cost to personal freedom. That wasn't Fauci's failure; weighing those trade-offs was never his job.

It was Trump's. Trump held the authority; Fauci only held influence. But Trump handed Fauci the podium and treated his recommendations as settled fact, not because Fauci overstepped, but because he was invited, and few people turn down real leadership when it's offered. It's flattering, it feeds the ego. The abdication is easy on both ends: the person with authority avoids a hard call, and the person handed the podium rarely hands it back.

The deeper mistake wasn't listening to Fauci. It was only listening to Fauci. Real balance comes from putting conflicting viewpoints in the same room: the epidemiologist arguing for lockdown, business leaders on the economic cost, psychologists on the toll of isolation, civil liberties advocates on personal freedom. That conversation is uncomfortable, but it's how balance actually gets found. None of that is a knock on Fauci, who did the job he was given. The failure was Trump's, for never doing the harder job of working through the conflict and making the necessary judgment call on what the right approach should be. Instead, he deferred to the one voice, and now blames that voice because he didn't like how things turned out.

The actual danger

This pattern predates AI. It's why companies pay McKinsey. A CEO facing a politically dangerous call, say layoffs, isn't really buying analysis; he's buying cover. Let the consultants "recommend" what he already suspected, and it becomes "McKinsey told us to," not "I decided." That's often the real product being sold.

AI is about to become the cheapest, fastest version of that same service. The risk here isn't that Fauci, McKinsey, or AI seizes authority. It's a person with authority quietly stepping back and letting the advisor drive, because the decision is hard, the advisor sounds certain, and if it goes wrong, someone else is there to blame.

AI will keep getting better at convergent conflicts, the ones caused by missing information. It will never be right for divergent conflicts: those are balance problems, and balance requires a human willing to sit with conflicting viewpoints and own a judgment call with no right answer. That's not a job AI will take from us. It's a job we'll hand over, one convenient decision at a time.