■ HIGH RISK ■ Education
The teaching-only lecturer is genuinely exposed. Delivering standard curriculum content to a hall of students is the single most substitutable activity in higher education, and casualized teaching staff have the least institutional protection when budgets tighten.
“The lecture part? AI nails it. The tenure politics and committee work? Unfortunately still human.”
Our AI replacement risk score — how we score jobs
This role differs from a tenured professorship in a way that matters here: it is weighted heavily or entirely toward teaching. The work is preparing and delivering modules, writing slides, running seminars, setting and marking coursework and exams, holding office hours, moderating grades, and handling the administrative churn of a cohort. Many people doing it are on fixed-term, hourly, or adjunct contracts with little bargaining power and no research portfolio to fall back on.
That profile aligns badly with what AI does well. Explaining established material is a model's strongest suit, and it does it on demand, one-to-one, at 2am before an exam — a service no lecture timetable can match. Slide production, module outlines, question banks, worked solutions, and feedback drafting all automate. Recorded and asynchronous delivery, already normalized after the pandemic, means one recording can serve many cohorts and many institutions. Grading of structured assessment is largely automatable, and the essay-based assessment that isn't has been destabilized from the student side. Meanwhile universities under financial strain look at teaching costs first, and casual staff are the cheapest line to cut. Our 55 reflects real, near-term employment risk rather than an abstract capability question.
The counterweight is that universities sell credentials and experience, not information — information has been free since the library, and freer since the internet. Seminar discussion, lab supervision, project mentoring, and the ability to notice a struggling student still require presence. Assessment integrity now demands supervised or oral formats that need humans in the room. Professional and accredited programmes require qualified instructors by regulation. Lecturers who anchor themselves in those formats are considerably safer than those whose observable contribution is a slide deck and a voice.
Automatability: our editorial assessment of current and near-term AI capability
Pressure is immediate for casualized teaching staff and will be acute by around 2030 as institutions combine financial strain with AI-assisted course production and asynchronous delivery. Expect fewer teaching-only posts, larger cohorts per lecturer, and consolidation of standard course content across institutions. Roles tied to labs, professional accreditation, and supervised assessment hold up considerably longer.
Because the role is weighted toward the most substitutable activity — delivering established content — and often lacks the research portfolio, tenure, and institutional standing that protect professorships. Many lecturers are also on casual or fixed-term contracts, which makes them the first line cut when university finances tighten. It's an employment-security gap as much as a capability gap.
It can explain the material, answer follow-up questions individually, and do it at any hour in any language — which is arguably better than a one-way lecture to three hundred people. What it can't do is run a seminar where disagreement produces insight, supervise a lab, examine a student orally, or notice that someone has stopped attending. Those formats are where lecturers should be.
It's the most exposed segment in higher education. Course content can increasingly be produced and delivered with AI support, recordings scale across cohorts, and casual contracts carry no protection. The practical advice is to specialize fast — lab-based teaching, accredited professional programmes, or supervision work — or to build a route into a permanent or research-linked position.
By changing what gets assessed rather than policing what gets generated. Detection tools are unreliable and adversarial. Oral defences, in-class writing, iterative work with visible drafts, and assessments requiring engagement with seminar-specific material all work better. Lecturers who lead this redesign become institutionally valuable, which is a useful side effect.