A few years ago, if something went wrong with a client outreach, a missed deadline, or a message sent at the wrong time, you could usually trace it back to a decision someone made. Today, that decision might belong to a system. It chose the timing. It drafted the message. It decided who to reach and when. And yet your name is still the one attached to the outcome.
That gap — between who’s responsible and who actually did the work — is becoming one of the most overlooked sources of stress in modern workplaces. Most leaders haven’t named it yet, let alone built anything to close it. A recent IBM study found two-thirds of CIOs and CTOs are now held accountable for AI systems they don’t fully control, and 77% say governance is already falling behind adoption.
I’ve talked with enough people managing automated systems to know there’s a discomfort here that goes deeper than most conversations about AI adoption acknowledge. Stefan Dunigan, VP of Customer Experience at Gryphon, a contact governance solution, framed it this way: “The first question is some version of ‘if this goes wrong, whose name is on it?’ Agents aren’t worried the system will replace them. They’re worried they’ll be accountable for a decision they didn’t make and probably can’t explain.”
That’s a different kind of anxiety than the one most automation conversations are built to address. We talk endlessly about efficiency and adoption curves. We don’t talk enough about what it feels like to stand behind a decision you didn’t make and can’t fully reconstruct.
The fix isn’t less automation. It’s more visibility.
It’s tempting to think the answer here is to slow down or pull back, to automate less and keep more decisions in human hands. What actually helps is visibility: people can stand behind a decision a system made, as long as they can see what it did and why. Take that visibility away, and you’ve handed someone accountability without giving them any of the agency that used to come with it.
So, what does “visibility” actually mean in practice? Dunigan breaks it down to three things: “What the system saw, which rule it applied, and what would have produced a different outcome.” He’s honest about what happens without that. “If you can’t reconstruct a specific interaction six months, a year, two years later and show why it was permitted, you don’t have true oversight. You have optimism.”
That line has stuck with me. Optimism might get you through the early stages of adopting new technology, but it won’t hold up the first time someone asks you to explain a system’s decision. It turns out the research agrees. A 2025 study found that increasing AI system transparency consistently improved people’s trust, confidence, and understanding of the AI’s choices.
Settle it before you deploy, not after something breaks
If you’re a leader about to hand more decisions to a system, this is the conversation to have first: Who’s accountable when the system gets it wrong, and what does that person actually get to see? Dunigan has seen how often that conversation gets skipped. “Most teams try to answer that after an incident surfaces, which is the worst possible moment to do so.” Answering it after the fact means someone is left holding responsibility they never agreed to, for a decision they had no hand in shaping.
This isn’t only a leadership exercise, either. If you’re the one whose name ends up on outcomes you didn’t personally shape, it’s worth asking your own version of these questions before you’re the one standing in front of a decision you had no part in making.
Why this matters beyond compliance
It’s easy to file all of this under “governance” and move on, but the stakes are more personal than that. Carrying responsibility with no real ability to influence the outcome is its own driver of burnout. Research on the job demands-control model has long shown that low control combined with high demands does more damage than heavy workload alone — and that erosion is much harder to reverse once it sets in.
The organizations getting this right are the ones who decided, early and out loud, what accountability actually means once a system is part of the decision, then built the visibility to back it up. Everyone else is finding out the hard way, usually after something’s already gone wrong.
If you’re building or leading a team where AI is starting to make real decisions, don’t wait for the first mistake to have this conversation. Have it now, while you can still shape what accountability looks like instead of assigning it after the fact.