AI Resilience Report forTeaching Assistants, Postsecondary

Right now, AI is mostly augmenting the work of postsecondary teaching assistants rather than fully replacing them — but the line is starting to shift, especially for routine tasks. Universities are running real pilots where chatbot "AI TAs" handle the kinds of questions a human TA would normally answer. At Fort Hays State University, for example, a professor uploaded her syllabus and assignments to a generative AI model ^[1] so students could ask things like "when are the article reviews due?" at any hour, and the University of Michigan's Ross School of Business is doubling a virtual-TA pilot built on Google Gemini that already covers 20 courses.

Researchers running these programs say the AI tools deliver faster responses, higher student grades, and less time spent by instructors answering routine questions ^[1].

Grading — the second-biggest TA task — is also being partially automated. Ohio State's distance-education office notes that platforms like Gradescope, Crowdmark, and Akindi are now widely adopted at Cornell, Purdue, UC San Diego, Florida, Rutgers, and Indiana University ^[2], and that large language models can now evaluate open-ended essays and give detailed written feedback — work that used to be a graduate TA's job. However, the same review warns that AI grading still struggles with bias, transparency, and "black box" decisions, so a human usually has to review the results.

Tasks that depend on relationships — leading discussion sections, holding office hours, mentoring — remain firmly human, partly because evidence on AI tutors still suggests caution ^[3].

Adoption is happening fast at the institutional level, but unevenly at the classroom level. The biggest accelerator is sheer availability and cost pressure: the Chronicle of Higher Education reports that instructors are increasingly warming to AI in 2026 ^[4], and an AAUP survey summarized by Inside Higher Ed found that 90 percent of responding faculty said their institutions are integrating AI into teaching and research ^[5]. With budgets tight and "not everybody can have a teaching assistant," AI TAs look attractive as a cheap supplement.

But several things are slowing full replacement of human TAs. First, the tech doesn't always work. CalMatters found that California community college districts are spending heavily on AI chatbots — Los Angeles Community College District alone has approved roughly $3.8 million in contracts through 2029 ^[6] — yet students report the bots give outdated or wrong answers and they end up using Reddit instead.

Second, labor, ethics, and governance push back. The same AAUP survey reported that 71 percent of faculty say administrators introduce AI with "little meaningful input" from professors ^[5], and unions are framing AI as an academic-labor issue. Third, Brookings warns that the risks of AI in education currently outweigh the benefits when tools are deployed without strong guardrails, as outlined in their global task force on AI in education ^[7].

The honest takeaway: routine grading and FAQ-style support are clearly being automated, but the human parts of TA work — running discussions, mentoring nervous students, making judgment calls on tricky essays — are exactly the skills colleges still need real people for. If you're heading toward this role, leaning into facilitation, feedback, and mentoring (and learning to use these AI tools well) is a strong, future-friendly bet.