Sam's take
Every new survey on AI in higher ed tells the same story: usage climbing, and faculty consistently trailing their students. I don't read that as faculty being behind. I read it as faculty being careful, which is the right instinct when the tool is this new and the stakes (grading, integrity, your name on a syllabus) are this high.
But careful isn't the same as absent. There's a narrow band of work where I think the calculus is already clear: rewriting a muddled question stem, drafting a rubric you'll edit anyway, turning three paragraphs of lecture notes into a discussion prompt, summarizing what 80 students wrote in an open-response check-in. Low stakes, high friction, easy to verify. The kind of prep work that eats a Sunday afternoon and doesn't make your teaching any better for the eating.
The parts I'd still keep my hands on: anything that shapes what a student actually learns, anything that assigns a grade, anything a student will read and assume came from you. That line will move over time. It hasn't moved yet.
Below are the specific uses I've seen finance and real estate faculty get real time back from.
Teaching notes
How to use AI to actually save you time
Most faculty I talk to have tried ChatGPT once or twice, felt underwhelmed, and moved on. The folks getting real leverage are using it for a small number of specific, repeatable tasks. Here is a working inventory from finance and real estate faculty I have compared notes with over the last semester.
Class prep
Turn a reading or case into a pre-class quiz in one prompt. Paste the source, ask for 5–10 multiple-choice or short-answer questions plus an answer key at your target difficulty. Two minutes instead of thirty.
Draft a Socratic question ladder for any case. Ask for 8–12 discussion questions that escalate from recall to analysis to judgment. You get a ready-made cold-call sequence you can mark up in the margin.
Turn dense industry reports or 10-Ks into classroom-ready teaching notes. A one-page summary, three discussion prompts, one suggested in-class exercise. This is the one that changes behavior most: swapping in a current example stops being a weekend project.
Build first-pass simulations and role-plays. Negotiation scenarios, founder pitches to evaluate, customer personas for students to interview. The first draft is rarely the version you run, but it gets you to a usable draft in minutes rather than starting from a blank page.
Week-to-week management
Convert lecture notes or slides into a student-facing study guide. Key terms, worked examples, likely exam traps. Doubles as your own pre-class outline, which is the real reason to do it.
Generate three variants of the same problem set. Easy, medium, hard, or three different industry contexts. Useful for differentiation without writing three assignments from scratch. Also solves the make-up exam problem when a student misses a Tuesday.
Refining the class over time
Debrief your own teaching after class. Paste the class plan and a few sentences on how it went. Ask for a short "what worked, what to change" analysis. Two minutes at your desk before you leave the building. It is a small habit that compounds across a semester and gives you something concrete to look at when you prep the same session next year.
Admin writing
Draft the recurring class-management writing. Syllabus updates, assignment instructions, late-policy explanations, Canvas announcements, replies to the eight student emails you answer every semester. Keep a folder of past examples in a document and paste it in as a style reference so the output matches your tone rather than the default AI voice. This is the single biggest lift on "saving personal time" and the easiest place to start if you are skeptical.
Grading
Rubric drafting and rubric-based feedback. Describe the assignment, generate a rubric, then paste anonymized student work and get rubric-anchored feedback you can edit rather than write from scratch. Biggest lift on written memos, ethics analyses, and market-analysis assignments where the marginal cost of good feedback is what usually keeps you from assigning more of them.
A few notes on what does not seem to work well yet. Anything requiring current market data (the model will hallucinate a cap rate or a stock price with confidence). Anything requiring judgment about a specific student. Anything where the value is in you doing the thinking: your research, your op-eds, your recommendation letters for students you actually know.
The pattern across everything above: AI is useful when the output is a first draft you edit, not a final artifact you ship. Faculty who treat it as a junior TA who is fast, tireless, and occasionally wrong are getting hours back per week. Faculty who treat it as an oracle get burned once and give up.
Worth a look
The data is limited! Here's what we know so far.
New time saving features from Blended Teaching
Here's what we released in August to help save you time on managing your classes.
New assessment methods are replacing AI detectors
Questions about the reliability of these tools are steering more instructors to find new ways of assessing students.
Sam