There is a category of work that never makes it onto a job description but consumes enormous amounts of energy: the work of managing other people's feelings. Soothing the frustrated client, absorbing the anxious manager's tension, performing enthusiasm you don't feel, staying calm when someone is being unreasonable. This is emotional labor, and it's been quietly redistributed in the age of AI tools. Marcus here — and the shift is more complicated than the optimists suggest.

What Emotional Labor Actually Is

The term was coined by sociologist Arlie Hochschild in her 1983 study of flight attendants and bill collectors. Emotional labor is the management of feeling to create a publicly observable facial and bodily display — the requirement to feel, or at least appear to feel, particular emotions as part of doing your job. It's exhausting in a specific way because it involves not just doing things but regulating your inner experience while you do them. In most workplaces, this labor is unequally distributed. Research from the American Psychological Association consistently shows it falls disproportionately on women, on people in service roles, and on anyone in a subordinate position. The assistant absorbs the executive's frustration. The customer service rep absorbs the caller's anger. This isn't incidental. It's structural.

What AI Is Actually Absorbing

AI tools are now handling a meaningful portion of first-contact interactions in many industries — initial customer inquiries, scheduling, basic support requests, intake forms. What this means in emotional terms is that the first wave of frustration, confusion, and low-grade hostility that users bring to a system gets absorbed by something that doesn't have a nervous system. The AI doesn't go home depleted. It doesn't lie awake replaying a difficult interaction. This is genuinely significant. If a customer service representative handles forty interactions a day and twenty of them involve some degree of emotional management, and an AI tool filters out fifteen of those twenty, the cumulative benefit to that person's wellbeing over a year is real. It's not glamorous, but it's a form of load reduction that matters.

Where the Load Goes Instead

Here's the part worth watching carefully. When AI handles the easy emotional labor, the human interactions that remain tend to be the harder ones — escalations, edge cases, situations where someone is genuinely distressed and needs actual human presence. The emotional labor doesn't disappear. It concentrates. The representative who used to handle forty mixed interactions now handles twenty difficult ones, plus twenty routine ones offloaded from the AI queue. The average emotional weight per interaction increases. A study from MIT's Work of the Future task force found that AI augmentation in service roles often intensifies the complexity of remaining human tasks without proportionally reducing workload. This is a pattern worth naming. Tools that promise relief can, under the wrong implementation conditions, increase pressure on exactly the workers they were meant to help.

The Part That Doesn't Get Discussed Enough

There's a subtler redistribution happening at the management level. When AI tools handle customer-facing emotional labor, the emotional work of configuring, monitoring, and humanizing those systems falls to someone — usually a middle manager or team lead who now has to ensure that the AI's tone is appropriate, that escalations are handled with care, and that the system doesn't damage customer relationships. This is new emotional labor created by the tools themselves. It's invisible, difficult to quantify, and usually uncompensated.

What a Healthier Shift Looks Like

The question isn't whether AI should be involved in emotionally inflected work. It clearly already is. The question is whether the redistribution is being designed thoughtfully or just happening by default. Organizations that are serious about worker wellbeing are starting to map emotional labor explicitly — who carries it, how heavy it is, what AI is absorbing versus concentrating versus creating. That kind of intentional audit is rare but valuable. The goal should be genuine load reduction, not just displacement. If AI is going to take on emotional work, the savings need to actually reach the people who were doing that work before. Otherwise, the tools that promised relief end up just moving the weight around.