Most of the advice faculty get about AI in the classroom falls into one of two camps: enthusiastic claims that it will transform everything, or blanket warnings to keep it out of the room entirely. Neither is especially useful day to day. What actually helps is a simple way to decide, task by task, whether AI belongs in a given piece of classroom work — and if so, how.

Start with what the task is actually testing

Before deciding whether AI is appropriate for an assignment, it’s worth asking what the assignment is meant to measure. Two very different answers require two very different approaches:

  • If the task is testing whether a student can produce correct output (a calculation, a well-structured paragraph, a formatted report), AI assistance can blur the signal you’re trying to get.
  • If the task is testing whether a student can reason, decide, or apply judgment, AI can often be a legitimate part of the process — a thinking partner, not a replacement for the thinking.

A lot of the anxiety around AI in education comes from applying the same blanket rule to both categories, when they call for different policies.

A simple three-tier approach

Tier 1 — AI as forbidden. Reserved for assessments specifically designed to measure a student’s unaided ability: closed-book exams, timed writing samples, foundational skill checks. Be explicit about this, and explain why, rather than leaving it as an unstated assumption.

Tier 2 — AI as a drafting tool, disclosed. For assignments like research summaries or first-draft reports, allow AI use but require students to disclose it and explain what they changed or verified. This turns AI use into a skill being assessed in its own right — can the student tell when the output is wrong, incomplete, or generic?

Tier 3 — AI as a built-in part of the exercise. For simulations, role-play scenarios, and decision exercises, AI can be the mechanism itself — the “counterpart” a student negotiates with, the entity generating a case scenario, or the tool that gives immediate feedback on a decision. Here the goal isn’t to restrict AI use but to design the exercise so the decision-making is still the student’s own.

Where faculty tend to get the most value quickly

In practice, the highest-value early use of AI for most faculty isn’t replacing their own teaching — it’s offloading the repetitive prep work that eats time without requiring much judgment. Generating a first draft of differentiated worksheet versions, producing a bank of discussion questions at varying difficulty levels, or building a role-play scenario around a concept already being taught are all places where AI can save real hours, provided the output is treated as a draft to review, not a finished product.

The judgment that doesn’t go away

Whatever role AI plays in a given assignment, one thing stays constant: the faculty member is still the one deciding what “good” looks like, and whether a student’s output — AI-assisted or not — demonstrates real understanding. AI can generate a first draft of almost anything. It can’t tell you whether a fourteen-year-old actually grasps supply and demand, or whether an MBA student can defend a pricing decision under pushback. That judgment call remains squarely a human one, and probably always will.