She tried ChatGPT first. It missed the nuances.
Kamala runs formal evaluations for non-tenured teachers at a 1,500-student Philadelphia high school. She had already tested ChatGPT and Playlab on the job.
The notes were good. The second draft was the problem.
Kamala captures feedback in shorthand while she is in the room — rich in content, informal, unstructured, exactly what useful observation notes look like. The evaluation a teacher actually receives cannot look like that.
So every observation carried a second job: turning the shorthand into formal writing, balancing tone, accuracy and formatting. That regularly took more than an hour per teacher, on top of the coaching itself. At a high school where non-tenured teachers need several formal evaluations a year, that hour is not occasional.
“Without the AI, definitely you’re writing it down and then you’re going a second iteration as to, hey, the language also needs to be tweaked, formatted properly. Those things take time.”
Straight from the notes to something she can send.
Swiftscore takes her observation — timestamps, what the teacher did, what the students did, the instructional decisions behind both — and produces the formal write-up directly. The second draft stops existing.
It also gives her two registers to choose from: a conversational teacher summary, and a directive growth plan. Which one she reaches for depends on the teacher and the moment, and she picks per conversation rather than rewriting to change tone.
- Shorthand notes turned into professional language, written once rather than rewritten
- Instructional nuance kept — including whether learning was student-led or teacher-led
- Two tones on tap: a conversational teacher summary, or a directive growth plan
An hour became five minutes, and the writing got better.
This is the part that makes her story different from a time-saving story. The hour did come down to five minutes. But Kamala’s own assessment is that the output is better than what she would have produced with the hour.
She had a basis for comparison. She had tried ChatGPT and Playlab on the same work, and her verdict was specific: Swiftscore gave more detailed feedback, caught the nuances the others flattened, and structured it better.
“What used to take me an hour now takes five minutes — and the feedback is not only beautifully written, it’s actually better than what I’d write myself.”
“This helps a lot in language… I read it through. I say, yeah, is this what I mean? And most of the time, yeah, it’s accurate.”
“I’m liking the growth plan. I’m liking the teacher summary. I find teacher summary more informal… It looks conversational. So depending on what I need at what point, I pick and choose.”
It is giving me more detailed feedback. And it is catching the nuances…
Run your next evaluation cycle with Swiftscore.
A 30-minute call with our team. We’ll walk through your framework and show you what your first cycle would look like.