AI Canvas Studio — Draw, Annotate, Get AI Feedback

A teacher professional development program serving two school districts. The lead coach had a 90-minute video PD that teachers were watching once and forgetting. She wanted a tool where teachers actually practiced charting a real lesson, not just watched someone else do it.

The brief was a deliverable that fits a single 30-minute PD block, runs on the district's existing browser, and gives every teacher a concrete chart they can bring to their next peer review. The studio had to teach the technique, let the learner practice it on a real lesson clip, and ship them out with a worked example to compare against.

Learners watch a tightly edited demo of two real teachers running a co-taught lesson. Interactive cards walk them through the three chart types: key steps, vocabulary, and concepts. They then draw on an embedded canvas while an AI coach reads their work and gives feedback in real time on what is missing, what is unclear, and what to keep.

The AI coach reads the canvas against a six-chart-type rubric the lead coach authored. The rubric is stable across all six chart types, so future modules can drop in without rebuilding the coaching layer. The canvas itself is serialized into a node graph the LLM can reason about, not raw pixels.