What keeps happening?
I am seeing a lot more of this in the leaders I work with, briefing papers to boards and executive teams are getting longer. There isn't more to say.
Saying more is now free. A briefing that took a day to write can be generated in minutes, and it arrives polished, structured and confident.
The research is catching up. In September 2025, Harvard Business Review published a study from BetterUp Labs and the Stanford Social Media Lab that named the pattern "workslop": AI-generated work that looks good but lacks the substance to advance the task. Of 1,150 US workers surveyed, 40% had received some in the past month, and they estimated about 15% of what lands in front of them fits the description. Some of it travels upward, from direct reports to their managers.
The researchers' central point is that it shifts the burden downstream. The receiver has to interpret, correct or redo it.
Only writing got cheap.
Reading, weighing and deciding did not. A busy leader still has one evening with the pack. They receive text the author may never have thought through, and are left to do the reflecting on a smaller time budget. If they reach for an AI summary, a machine condenses what a machine wrote, and the person who signs off meets a summary of a summary.
Why it happens
Jeff Sparrow's recent essay in The Conversation is about Australian universities, but the mechanism travels. His argument is that AI didn't break higher education. It landed in a system that already rewarded the appearance of work over the work itself. He quotes a historian describing AI as "a catalytic solvent, titrating out the incoherence that was always there."
He points to one academic journal where submissions rose 42% after ChatGPT arrived, with editors blaming AI combined with publish-or-perish incentives. He also notes that some universities encourage AI-assisted grading, provided a human being is in the loop.
Organisations have the same shape. Where effort is measured by documents produced, documents will multiply. And "a human signs off" is a weak control when no human could realistically read what is being signed. Approval without reading is ritual, not oversight.
This is a governance problem, not a technology one. It sits in decision rights and expectations, and nobody has made the decision.
What to do about it
Leaders often complain about the overload, and some quietly stop reading. Neither helps. The leader is the one person with the authority to set the terms of what reaches them. For example:
Two pages maximum, or a 15-minute slot, whichever suits the matter.
The decision needed, stated at the top.
A recommendation, not a menu of options left for the reader to sort out.
What the author checked themselves, and what they are unsure of.
A short live conversation where the author defends it.
The constraint does the real work. It forces the author to decide what matters, which is exactly the thinking AI lets people skip.
Some universities are heading the same way.
NYU Stern researchers describe students handing in polished project work they couldn't explain under follow-up questions, and have trialled voice-AI oral exams in response. Ireland's higher education authority has proposed an "oral assessment safeguard": inviting a student to explain and defend their work when authorship is in doubt. It is early and uneven. Many UK universities keep viva exams for suspected misconduct because live questioning is costly to run at scale. But the logic carries. A document can be generated. The understanding behind it can't.
The part that falls to the leader
A standard only exists while the person who set it honours it. Read what meets the rule. Engage with the recommendation. Give a decision. Send back what doesn't comply, unread. Send short things yourself. The first time a leader waves through a 20-page paper they didn't read, staff learn the real rule, and the stated one becomes decoration. The HBR researchers land in the same place: leaders should set clear norms and model purposeful use. Leading by example isn't a nice extra here. It is the mechanism.
Take this to your next meeting
Choose one: set a page limit and ask every author for a recommendation, or accept that you are approving papers you haven't read. And ask the room: could anyone defend the longest paper we accepted this quarter without opening it?
Sources
Jeff Sparrow, "Friday essay: AI is devastating higher education, but the system was already broken", The Conversation, 23 Oct 2026. https://theconversation.com/friday-essay-ai-is-devastating-higher-education-but-the-system-was-already-broken-292988
Kate Niederhoffer et al., "AI-Generated 'Workslop' Is Destroying Productivity", Harvard Business Review, 22 Sep 2025. https://hbr.org/2025/09/ai-generated-workslop-is-destroying-productivity
Panos Ipeirotis and Konstantinos Rizakos, "Scalable and Personalized Oral Assessments Using Voice AI", arXiv, 2026. https://arxiv.org/abs/2603.18221
"Oral assessment as a safeguard for academic integrity", International Center for Academic Integrity, Mar 2026, citing the Irish HEA policy framework (2025). https://academicintegrity.org/aws/ICAI/pt/sd/news_article/631829/_PARENT/layout_details/false
Brumby, Cox, Sarkar and Gould, "Sampled vivas are pivotal in combating AI cheating", Times Higher Education, 6 Mar 2025. https://www.timeshighereducation.com/node/735603