We’re not just delegating tasks to AI.

   We’re delegating accountability

It started with a coffee cup.

Starbucks Korea sells an oversized stainless-steel tumbler it calls the “Tank.” Late last spring, the marketing team decided to run a promotion around it. They used AI to help develop the campaign — including a launch name and a slogan. The AI suggested “Tank Day” for the event and “Thwack on the desk” as the tagline, meant to evoke the satisfying thump of setting a heavy cup down. Both suggestions were innocent on their face and nobody flagged either one. The campaign was approved and launched on May 18.

What nobody on the team had registered — or if they had, nobody said so — was that May 18 is one of the most sacred and painful dates in modern South Korean history: the anniversary of the 1980 Gwangju Uprising, in which the military used tanks to crush pro-democracy protesters, killing hundreds. And “Thwack on the desk” wasn’t just a catchy tagline. It echoed the exact phrase police used to cover up the 1987 torture death of student activist Park Jong-chul — the claim that he had simply collapsed after an officer “struck the desk with a thwack.”

In one simple product promotion, Starbucks Korea had stumbled into two of the country’s rawest historical wounds. Within a week, card payments at Starbucks stores fell 33%. Customers filmed themselves destroying tumblers. All 2,160 South Korean stores closed early for mandatory history training. The CEO was fired. The Korea Times reported that some managers who approved the campaign had never even opened the email attachments showing the actual materials.

None of this was intentional. That’s precisely what makes it such a clear illustration of the problem.

The Tool Didn’t Fail. The Oversight Did.

Nobody is arguing that AI isn’t useful. It is. But useful is not the same as sufficient.

What the Starbucks Korea team was missing wasn’t better technology. It was the human layer — the judgment, cultural fluency, and contextual awareness that no language model currently carries. AI generated the language. Humans were supposed to evaluate it. That step got skipped, or rushed, or assumed to be handled. It wasn’t.

This is the pattern I’m seeing play out in organizations regularly. Leaders adopt AI tools — legitimately powerful ones — and then, under pressure and moving fast, shift from “AI assists our judgment” to “AI replaces the need for it.” The autopilot gets turned on.

And It Turns Out, We’re Not Checking.

Here’s what makes this particularly striking. It’s not just that AI produces errors. It’s that we stop looking for them.

A recent study published in The New York Times examined what happens when companies treat AI as a formal “employee” — listing AI agents on org charts, giving them names, embedding them in teams. Researchers from Boston University and Boston Consulting Group surveyed more than 1,000 corporate managers and found something that should give every executive pause: managers at companies that had placed AI agents on their org charts caught significantly fewer errors when reviewing AI-produced work than when reviewing work produced by a human.

The reason, the researchers speculated, is a shift in felt accountability. When a human direct report makes a mistake, managers instinctively feel responsible (“that’s on me”). But when an AI “employee” makes a mistake, the psychological out is right there: that’s on the tech team, or on whoever decided to implement it. “But it’s not your problem,” as the lead researcher put it, describing the managers’ unspoken assumption.

The Starbucks Korea story is that study, made visible. Someone approved a campaign without really looking. And things went very wrong.

The Complexity Hasn’t Gone Anywhere.

The issue isn’t just in our handling of AI. The environment executives are operating in has never been more complex. Geopolitical shifts are reshaping supply chains overnight. Workforce expectations are evolving faster than most HR frameworks can track. AI itself is moving so quickly that the tool you implemented six months ago may already be outdated — or may be surfacing biases you haven’t accounted for.

The leaders who will navigate this well are not the ones who find the best automation. They’re the ones who bring clear judgment to the automation they deploy. They know what AI is optimizing for — and what it can’t see.

That kind of clarity doesn’t come from a dashboard. It comes from the inside out.

What Gets in the Way of Good Judgment

Most senior leaders I work with aren’t lacking information or intelligence. What they’re running up against is something more subtle — and more consequential.

Each of us carries  a Winning Strategy™: the set of behaviors and approaches that got us to where we are. It worked. That’s exactly why it’s so hard to see. It’s baked into how we make decisions, how we read a room, how we define what a good outcome looks like.

And in a world that’s shifting as fast as this one, a Winning Strategy built for a different era — or a different scale of complexity — can become the thing that limits you. You’re not making bad decisions. You’re making the same good decisions you’ve always made. The environment just changed around them.

The Work That Requires a Human

The work we do at Renascence is built around exactly this: helping executives see the operating system underneath their leadership — clearly enough to update it intentionally, rather than discovering its limits in a crisis.

That means developing the capacity to pause, even when everything is moving fast. To ask what does my judgment say — and do I trust it?” To recognize when you’re running on autopilot and then to make a deliberate choice to re-engage.

This is what I mean when I say AI isn’t a set-it-and-forget-it strategy. Neither is leadership. Both require ongoing, attentive human judgment at the center.

The Question Worth Sitting With

Before you move on to the next thing on your list: where in your organization is something running on autopilot that hasn’t had a real human gut-check in a while?

It might be a process. It might be an assumption. It might be you.

The best leaders aren’t the ones who never run on autopilot. They’re the ones who know when they are.

—

Suggested meta description (for WordPress):

AI is a powerful tool — until no one’s watching it. Here’s what the Starbucks Korea incident reveals about leadership judgment in a world that’s moving fast.