How to Use AI as a Professor: Uses, Limits, Risks

A professor's week splits into preparation, teaching, and original scholarship — and only the first of those is where a general-purpose AI tool is genuinely useful. A first-pass lecture outline, a slide deck draft, a bank of discussion questions, a routine email to a student or a committee: a tool drafts a reasonable first version of all of that quickly. It has no access to the second half — reading a room of confused undergraduates in real time — and it cannot do the third at all, because original research depends on judgement about what is actually true and what is actually new, which is exactly the property these tools are weakest on. Knowing which third of the week is safe to hand over is most of what makes this useful rather than risky.

Will AI replace teachers covers that same shrinking-versus-not split at the level of a whole role, for the K-12 classroom version of this job. This is the practical version for a lecture course: what to actually hand over this week, and where a fluent-sounding draft still needs a check before it reaches a lecture hall or a journal. How to use AI as a teacher covers the same practical ground one level down, for classroom prep rather than a lecture course.

Where AI genuinely helps

1. 01Drafting a first-pass lecture outline or a slide deck structure from a syllabus topic and a target length, so you are editing and sequencing a draft rather than building the whole thing from a blank page.
2. 02Turning a dense source — a paper, a technical report — into a plain-language summary you then check against the original and adapt for the level of the class actually in front of you.
3. 03Generating a bank of discussion questions or exam questions at a stated difficulty, which you then vet for accuracy and relevance before anyone sees them.
4. 04Drafting the first version of a routine email — a syllabus clarification, a committee update, a recommendation-letter template — so you are reviewing tone and detail rather than composing from nothing.
5. 05Producing a first-pass summary of a long batch of student feedback or evaluation comments, as a starting point you verify against the original responses.

A worked example

Weak: "Make slides about photosynthesis."

Better: "Draft a 12-slide outline introducing photosynthesis for a first-year undergraduate biology course. One core concept per slide, building from light-dependent to light-independent reactions. Flag any slide where a diagram would help more than text. No speaker notes yet." The second version states the audience level, the slide count, the structural logic and the output format, rather than leaving the model to guess at depth and shape — how to write a prompt that works on the first try covers that same discipline more generally.

Check every factual claim and every labelled diagram before it reaches a slide a student will study from. A lecture slide that states a mechanism incorrectly does more damage than the time saved drafting it — checking an AI answer when you are not the expert is the general version of that habit, and it matters doubly here because students often treat a professor's slide as the authoritative version of a fact.

Where it stays your call

The techniques with the strongest evidence behind them — spaced practice and retrieval practice — work because an instructor notices which specific concept a specific cohort has not grasped and times the right question for them, not because a slide deck exists in the abstract. Deans for Impact's synthesis of the learning-science evidence makes the same point: the evidence describes what works when a person applies it to a particular class, and that application is not a step a general-purpose tool can do for you.

A model can draft the slide deck faster than you can build it from scratch. It cannot tell you, reliably, which concept this specific cohort has not actually grasped yet — or whether the claim on slide nine is true.

Original scholarship is where the line is firmest. A widely cited survey of hallucination in large language models covers why a fluent, confident claim and an accurate one are not the same property in these systems, and a fabricated citation or a misreported result in a research context is a different order of problem than a wrong fact on a practice worksheet — it is the kind of error that damages a body of work others will build on.

Where it goes wrong

• —A claim, a citation or a reported result stated with total confidence and wrong, copied into lecture material or a draft paper without being checked against the primary source.
• —Student work or identifying details pasted into a tool with no institutional data agreement behind it — check what your department and your institution's research-ethics office have already decided before putting any student-specific or unpublished data into a general consumer tool.
• —Grading feedback generated and sent without being read against the actual submission, which tends to produce comments that are generic or address a mistake the student did not actually make.
• —A generated lecture outline or slide deck treated as finished rather than as a draft pitched at an average student who does not exist in your actual cohort.

What to do this week

1. 01Find out what your institution has already said about AI tools, student data and research use — many now have a written policy, worth reading before assuming anything is fine to use.
2. 02Pick one purely preparatory task — a lecture outline, a question bank, a routine email — and try it for a week, checking every output against your own knowledge of the subject and the course.
3. 03Keep the time actually saved in mind rather than a general impression that it helped, and spend that time on the part of the job a tool cannot do: reading how a specific cohort is actually responding, and the original analysis behind your own research.
4. 04Find the repetitive part is a short test for which other prep tasks are worth handing over once the first one is routine.

The short version

Use AI for first-draft lecture outlines, slide structures, discussion questions and routine correspondence. Check every fact, every citation and every piece of feedback before a student or a journal sees it, and never put unpublished research data or identifying student information into a tool your institution has not cleared. The prep work gets faster. Knowing this cohort, and the original judgement behind your own scholarship, does not move to software.

That split between faster prep and protected judgement is also what the wider labour market has started pricing in — PwC has measured a 62% wage premium for workers with demonstrable AI skills, up from 57% the year before. Coursium teaches that practical layer directly: writing a specific, levelled request and the habit of checking the result before it reaches anyone else. Stay ahead of AI by learning the tools on your phone.