Teaching with AI Feedback: An Experience Report on the Challenges and Takeaways from a Quasi-Experimental Study

This experience report presents qualitative findings from a quasi-experimental study at Towson University examining ChatGPT's role as an evaluative partner in a junior-level business communication course. Across 15 course sections taught by five instructors, the intervention embedded AI feedback into the drafting process using rubric-based heuristics, staged integration, and reflective prompts. Post-semester faculty interviews revealed that when scaffolded intentionally, AI feedback reinforced instructor guidance, functioned as a “second reader,” and supported deeper revision, rhetorical awareness, and metacognitive growth—particularly among engaged students. However, instructors also reported uneven engagement, limited feedback literacy, and instances of misuse when students treated AI as a generative shortcut rather than a revision partner. These findings underscore the need for pedagogical guardrails, early integration, and explicit instruction in evaluative prompting. The study positions AI not as a replacement for human feedback but as part of a multi-voiced feedback ecosystem, blending machine and human perspectives to foster clarity, audience awareness, and confidence in student writing.