Here's a result that should worry every team lead. In a study published in May 2026, the Boston Consulting Group gave people the same document to review. One group was told a human wrote it; the other was told it came from a named AI "employee." The second group identified fewer errors and "reported less accountability, blaming the AI agent, rather than themselves, for a mistake." BCG's Matthew Kropp called it "people sort of passing the buck."

That is not a story about lazy employees or evil AI. It is a textbook case of moral hazard—and it is spreading through workplaces faster than anyone is tracking. The term sounds like an accusation, like someone did something immoral. It isn't. It's one of the most useful and most misunderstood ideas in economics, and once you see it, you can't unsee it on your own team. This piece is about what moral hazard really is, why it's a wiring problem and not a character problem, why distributed teams have it worse, and the one structural fix that actually works.

What moral hazard actually means (and why "moral" is a misnomer)

Start with the definition, because almost everyone gets it wrong. Wikipedia puts it cleanly: moral hazard is when an economic actor "has an incentive to increase its exposure to risk because it will not bear the full costs associated with that risk." The Corporate Finance Institute is blunter—it's taking advantage of a situation "knowing that all the risks and fallout will land on another party."

Notice what's missing: morality. The word "moral" is a 200-year-old accident. As Wikipedia notes, "prominent mathematicians who studied decision-making in the 18th century used 'moral' to mean 'subjective'"—not ethical. When economists revived the concept in the 1960s, the usage "did not imply immoral behavior or fraud." It describes "inefficiencies that can occur when risks are displaced or cannot be fully evaluated." It's a property of a system, not a flaw in a person.

The concept comes from insurance, where the term "dates back to the 17th century and was widely used by English insurance companies by the late 19th century." Insurers noticed something inconvenient: a person with fire insurance is, on average, a little less careful about fire. Not because they're bad—because the downside is now someone else's problem. Insure the phone and you handle it more roughly. Cap the loss and you raise the risk.

The cleanest large-scale example is 2008. Banks made enormous bets on bad loans because, as CFI describes it, "the US Federal Government stepped in and bailed them out." Once you believe a rescue is coming, "banks are then more likely to continue making risky loans if doing so offers temporary gains." Privatize the upside, socialize the downside, and rational people take wild risks. The bankers weren't cartoon villains. They were responding, exactly as the model predicts, to a structure where the decider didn't pay.

Moral hazard at work: the decider doesn't pay

Now look at your own org with that lens, and the pattern is everywhere. It works like this: the person who makes a decision is insulated from its cost. Someone downstream pays instead.

Sales promises a feature to close a deal; engineering eats the all-nighters to build it. One developer said it perfectly on Hacker News: they "overpromise to the client and then expect engineering to save their ass. It's never sales that has to sacrifice their time to meet deadlines." The commission lands on the rep. The cost lands on someone else.

It runs the other way too. A developer merges a risky change on Friday afternoon and logs off for the weekend; the on-call engineer eats the 2 a.m. page. As another practitioner put it, "It really harms morale to have the developers all enjoying wonderful weekends while ops is on red alert because app changes they don't understand broke everything in production." The person who created the risk is asleep. The person who absorbs it never had a vote.

And the meeting nobody needed? Someone schedules ten people for an hour without ever counting the ten hours they just spent—because their hour is the only one on their calendar. The scheduler decides; the room pays. It's the quiet engine behind meeting bloat that no agenda fixes.

AI just put this on rocket fuel. A 2025 study by BetterUp Labs and Stanford, published in Harvard Business Review, named the new artifact "workslop"—"AI generated work content that masquerades as good work, but lacks the substance to meaningfully advance a given task." Of 1,150 U.S. employees, "40% report having received workslop in the last month," each instance costing "one hour and 56 minutes" to untangle—an "invisible tax of $186 per month" per worker, "over $9 million per year" for a 10,000-person org. The mechanism is pure moral hazard: workslop "shifts the burden of the work downstream... it transfers the effort from creator to receiver." Generating it is cheap for you and expensive for someone else. So more of it gets made.

Moral hazard is a system problem, not a people problem

Here's where most leaders go wrong, and it's worth steelmanning their instinct, because it's a good one. The standard response to all of the above is cultural: hire better people, align everyone on shared values, set clear OKRs, and trust your team to do the right thing. That's not naive. Trust beats surveillance, shared goals beat silos, and a team that needs a rule for everything is already broken.

But here is the hidden condition that breaks it. The effect is structural, which means good people respond to insulation from consequences exactly the way bad people do. You cannot values-statement your way out of an incentive. The 2008 bankers had codes of conduct. The Friday-deploy developer isn't a sociopath. The salesperson genuinely wants the company to win. Drop any of them into a structure where the cost lands on someone else, and behavior drifts—not because they're weak, but because that's what the wiring rewards. The counter-intuitive part: the more you lean on trust while leaving the cost misplaced, the worse the decisions get, because trust removes the one check the structure was missing.

This is also why moral hazard is distinct from its famous cousin, the principal-agent problem—and the difference matters for the fix. The principal-agent problem is about visibility: can you see what your people are actually doing? Moral hazard is about consequence: who pays when the decision goes wrong? You can have perfect visibility and still have a brutal version of it, because watching someone make a call doesn't mean they feel its cost. (Its other sibling, adverse selection, is about hidden information before a deal—"caused by hidden information, rather than hidden actions"—where moral hazard is the hidden action after.) Monitoring fixes the first problem. It does nothing for the second.

Why distributed teams get moral hazard worse

If the whole problem is that the cost lands on someone else, then distance is its best friend. On a co-located team, the cost is at least visible. You see the support engineer with their head in their hands. You watch the designer redo the slop. The person who created the mess has to walk past the person cleaning it up. That visibility is a tax on bad decisions, and it's free.

Spread the team across time zones and that tax disappears. The decider sees the shipped artifact—the closed deal, the merged PR, the generated doc—and ships off to the next thing. They never see the 2 a.m. page, because it happens in a channel they've muted, in a region that's asleep when they're awake. Async communication is wonderful for a lot of things, but it is exceptional at hiding downstream cost. What travels across the distance is the win. What stays invisible is who paid for it. It's the same blind spot that lets busy teams mistake motion for progress and lets the real signal get buried under everything everyone ships—and it compounds with every context switch the cost forces on the receiver.

So the first move isn't a new policy. It's making the cost visible and shared—putting the decision and its downstream consequence in the same place, in front of the same people, at the same time. That's the gap a tool like Coommit is built to close: by keeping the conversation, a live shared canvas, and contextual AI in one workspace, the person making the call and the person who'll absorb the cost are in the same room, looking at the same thing—so the decider actually sees who pays before the bill comes due.

The fix: put skin in the game

Visibility is the start. The structural fix has a name, and it's Nassim Taleb's: skin in the game). Give the decider the downside, not just the upside. Wire the consequence back to the person who chose. Concretely:

This isn't theory. When you actually give people the downside of their decisions—real ownership—behavior changes measurably. Employee-owned companies, where workers share the risk and not just the reward, report "voluntary quit rates of their employees at roughly one-third of the national average," are "only half as likely as non-ESOP firms to go bankrupt or close," and their people are "significantly less likely to be laid off (1.9% vs. 5.1%)." Skin in the game isn't a poster. It's a structure, and structures move numbers.

The bottom line

Moral hazard is not a morality problem, and treating it like one is why so many "accountability" initiatives fail. It's a wiring problem. The question to ask about any decision on your team isn't "are my people accountable enough?"—it's "does the person making this call actually feel its cost?" If the answer is no, no amount of trust, values, or alignment will save you, because you're fighting the incentive with a slogan. And if your team is distributed, assume the cost is invisible by default, because it is.

The work is to make the bill visible, route it back to the desk that ordered it, and give people genuine skin in the game. Do that, and you don't need to demand accountability—the structure produces it. The decider has to pay. Build the team so they do.