Tufts Built an AI Tool That Captures the Feedback Veterinary Students Never Knew They Were Getting

Anyone who has been through clinical rotations knows the feeling. Your instructor says something in passing over the surgical table, a quick correction, a piece of guidance delivered mid-procedure, and by the time you are scrubbing out you cannot remember the exact words. Neither, sometimes, can they.

A team at Cummings School of Veterinary Medicine at Tufts University decided that problem was worth solving, and they built something to fix it.

The tool is called VetFeedback.ai, and it is exactly what it sounds like: an AI-powered platform designed to capture, transcribe, and contextualize the real-time feedback that instructors deliver during labs, surgeries, and clinical rotations. The kind of feedback that happens in the moment, in passing, between tasks, and that research suggests students are absorbing far less of than their instructors believe.

That last point is backed by data. A study published in Frontiers in Education surveyed students and instructors across 33 universities and found that students reported receiving feedback only 28.5 percent of the time, while instructors reported giving it 44.5 percent of the time. That gap is not a matter of effort. It is a matter of how feedback lands when someone is in the middle of learning a high-stakes skill under pressure. As assistant teaching professor Ariana Hinckley-Boltax puts it, interpreting and internalizing feedback is an emotional and subjective process, and in a busy clinical environment, students can easily lose the forest for the trees.

The platform works simply. Instructors record from a phone in their pocket throughout a lab or clinical session. VetFeedback.ai picks up the audio, transcribes it, and then maps what was said against evidence-based feedback guidelines and customizable learning outcomes. The AI does not rewrite or editorialize. The team was deliberate about training it to be minimally creative. The instructor's words stay the instructor's words. The platform just organizes them, surfaces their relevance to the curriculum, and delivers them to students in a format they can actually process after the fact.

There is another layer worth noting. If instructors use the platform consistently enough, it will eventually begin analyzing their own communication patterns and offer feedback on how they give feedback. That is a level of professional development that most clinical educators never get access to.

The tool is still in early testing, with a workable model expected by fall. But the problem it is solving has been sitting in plain sight for years. Clinical education is dense, fast, and emotionally loaded. The most useful teaching often happens informally and disappears just as fast. VetFeedback.ai is an attempt to stop letting that go to waste.

Students always want more feedback. Now there may finally be a way to make sure they actually get it.