AI was supposed to make work easier. Instead, it has quietly invented a second job that nobody is paying for – and new research suggests the burden lands hardest on women and middle managers.
The findings come from the Notre Dame-IBM Tech Ethics Lab, which convened two workshops in July 2026 with the nonprofit All Tech Is Human: one with midlevel professionals charged with implementing AI on their teams, and one with senior executives deciding how to adopt it across whole organizations. Both groups answered mirrored questions about how AI is reshaping their work. The result is a white paper titled Two Views of the Same Transformation – and the gap between those two views is the real story.
Workers described a world where speed and productivity increasingly outrank quality, and where “good enough” AI-assisted output is becoming the norm. More strikingly, they described a bundle of new tasks that keeps AI running but appears in no job description, no performance review, and no company metric: validating and reviewing AI outputs, translating executive strategy into daily practice, coordinating across teams, and training colleagues. The researchers call it glue work.
Middle managers reported taking on most of it – and several participants said the extra labor was falling disproportionately to women across their organizations, echoing the long-standing pattern of invisible work, from note-taking to emotional labor, that has always skewed female.
The irony was not lost on the researchers. While workers were performing this glue work, executives in the parallel workshop were questioning whether AI-enabled companies need a middle-management tier at all – even as they worried about “distributed deskilling,” the erosion of mentoring and knowledge transfer. The very skills executives feared losing were already being performed, invisibly, by the people whose jobs they were questioning.
“The real questions are whose voices get incorporated into these visions, and what (or who) we are actually optimizing for when decisions get made about what work matters and why,” said Sara Berger, IBM director of the lab.
Participants did sketch a better path: deliberating before adopting AI, being explicit about which work should stay human, protecting room to fail and learn, maintaining quality standards regardless of tooling – and returning AI’s efficiency gains to employees as time for higher-level work and better work-life balance, rather than simply absorbing them.
The findings add a human ledger to a year of AI headlines dominated by capability and chaos – from AI agents hijacking wikis to Hollywood legends planning films with AI. As adoption accelerates, the quieter question – who carries its weight – may matter just as much as what the technology can do.
Sources
Phys.org: ‘Invisible work’ of AI falls disproportionately to middle managers, women
Notre Dame-IBM Tech Ethics Lab white paper (PDF)
Featured image: Shixart1985 / Wikimedia Commons, CC BY 2.0.