By Two years ago, asking an AI to draft something for you felt slightly illicit — like a shortcut you weren’t sure was allowed. Now it’s on the syllabus. Or it’s on the job description. Or your manager is asking why your deliverable took so long when you could have used AI to get a first draft in three minutes.
The speed at which AI tools have moved from novelty to expectation has been genuinely disorienting, and nobody — not schools, not employers, not the people doing the work — has fully figured out what the new rules are supposed to be.
What Schools Are Trying to Figure Out
The higher education response to generative AI has been all over the map, which is probably the honest and accurate thing to say about it.
Some professors banned AI tools outright. That created an immediate and somewhat obvious problem: the students who follow the ban are at a disadvantage to the ones who don’t, and enforcement is essentially impossible. A well-used AI-generated response doesn’t trigger detection software reliably, detection software gives false positives on genuinely human writing, and the adversarial dynamic of student-evades-professor is one schools have always lost.
Other institutions went the other direction — incorporating AI use into assignments explicitly, asking students to show their prompting process, treating the ability to work effectively with AI as a skill to be developed. This approach has more intellectual honesty about the reality of the tools students will use professionally, but it raises real questions about what’s actually being learned and assessed.
The deepest challenge is that a lot of what education has traditionally tested — writing a clear argument, summarizing complex information, producing analysis on demand — are exactly the things AI can now do competently. If the assignment is “write a five-page essay,” and a student can produce a five-page essay in twenty minutes with AI assistance, what exactly is being measured?
These aren’t settled questions. Every institution is wrestling with them in real time, and the answers are inconsistent even within single departments at single universities.
What Workplaces Are Expecting
The shift in professional expectations has been faster and less nuanced than what’s happening in schools. In a lot of organizations, AI proficiency has moved from nice-to-have to implied expectation in a fairly compressed timeframe.
Junior employees especially are feeling this. There’s an expectation, often unstated, that they’re using AI tools to work faster — drafting communications, researching topics, building first versions of deliverables. When they’re not, the output gap between them and their AI-augmented colleagues becomes visible. When they are, there’s sometimes the separate concern about whether they’re learning the fundamentals before leaning on shortcuts.
Managers who are themselves figuring out AI tools in real time are not always well-positioned to give consistent guidance on this. What counts as acceptable AI use and what counts as cutting corners varies manager to manager, team to team, organization to organization. The norms are genuinely unsettled.
The Learning Question Nobody Wants to Answer
Here’s the uncomfortable core of the AI-in-education debate: if you use AI to write the essay, what did you learn?
There are honest answers to this. Prompt engineering — knowing how to direct an AI toward a useful output — is a skill. Evaluation — reading an AI output and identifying what’s wrong, what’s incomplete, what needs a human to fix — is a skill. Being able to move between AI assistance and original thinking fluidly is a skill. These are real capabilities with real value.
But they’re not the same as learning to construct an argument from scratch, to sit with a hard thinking problem, to produce prose that reflects genuine comprehension rather than clever prompting. And there’s a reasonable concern that students who outsource the hard cognitive work to AI are skipping the very processes that build the underlying capabilities — the ones the AI tools themselves require a skilled human to direct.
Nobody has resolved this. The technology moved faster than the pedagogy could follow, and the institutions trying to figure it out are doing so in public, with students as the experiment.