Trending
Written by: Anat Baron
AI doesn't need formal decision-making authority to determine an outcome. It only needs to produce a recommendation that shows up fast, sounds certain, and is easier to accept than to challenge. That's how advice quietly becomes the decision.
I call this authority drift: what happens when the work mix changes but the authority decision is not revisited.
Most organizations have already asked what AI can do. Far fewer have asked the harder question: what should it actually be allowed to decide? The first question gets plenty of attention. The second one often gets skipped.
Here's what makes it easy to miss. How much of a task AI performs, and how much authority it's actually been handed, are not the same variable. A system can carry out nearly all the visible work while a person still makes the real call. Or it can do very little of the visible work and still shape the outcome, quietly, by determining which options even look credible enough to consider. Treating those as the same thing is exactly how authority drifts without anyone deciding it should.
I've been caught by this myself. While rebuilding my own online presence, I ran four different AI models side by side and treated their agreement as proof. All four reached the same wrong conclusion about a technical issue on my own site. Agreement felt like confirmation. It wasn't.
I'd recognized this exact failure mode before, just not in myself. Years earlier, bringing my documentary Beer Wars to roughly 450 theaters for a one-night event, the chains handed back standard seat allocations built on their own track record. My own data from industry specific promotional partners showed something different. They trusted their history over the evidence in front of them, and the theaters in the cities with real demand sold out while the allocated suburban ones sat empty. I challenged that kind of misplaced confidence when it came from movie theater executives with years of industry experience. I didn't recognize it when it came from four AI systems agreeing with each other on a subject outside my own expertise.
A short exercise surfaces where this may already be happening in your own organization. Pull the ten most consequential decisions from last quarter. For each one, ask who actually made the call. Was AI used as evidence, as a recommendation, or as the default? Was a person named as the decision-owner before the recommendation arrived? Could that person explain, on their own, why the decision was accepted or rejected?
This is exactly the gap The Human + AI Equation exists to close. It asks leaders to decide, on purpose, what stays human-led, what should be AI-augmented, and what can be fully automated, rather than allowing the mix to drift on its own.
Closing that gap requires naming who owns the decision, ensuring that person has the authority to override the system when the specifics warrant it, and making clear who answers for the outcome when the system is wrong.