AI-Supported Analysis of Student Work Saves Teachers Time Without Replacing Their Judgment
5 Teachers
Middle School Math Teachers
6 Months
of Classroom Implementation
3 Hours
Saved Per Class Per Week (on average)
Finding 1
EdLight Cuts Student Work Review Time Nearly in Half
Teachers went from spending roughly 45 minutes per class reviewing exit tickets to 15-30 minutes – without sacrificing instructional quality. Efficiency gains were greatest during high-pressure weeks when time was most limited.
Instead of reviewing one paper at a time, teachers used EdLight to quickly surface shared misconceptions across a whole class – making precision pairing and small-group instruction faster and more targeted.
Finding 3
AI Feedback Works Best When Teachers Are in the Loop
Alignment between AI feedback and teacher judgment was strongest for numeric and symbolic tasks. For reasoning-heavy work like diagrams and graphs, teachers mediated AI insights – and students learned more when feedback was embedded in instructional routines rather than delivered directly.
Dive Deeper Into Our Research
Explore the full findings from our AIMS Collaboratory and Gates Foundation-supported pilot.
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
