Surfacing friction across Pearson's AI grading and assignment tools.

The experience is organized around tools, not tasks.

Different backend services each assert themselves visually, producing competing headers, redundant containers, and fragmented attention. Instructors don't think in terms of products or systems; they think "I need to create an assignment" or "I need to finish grading". The experience often failed to support that mental model.

Wayfinding and state break down when confidence matters most.

Many core workflows are multi-step and consequential, yet lack clear progress indicators, visible completion states, awareness of what's next, and safe ways to revise earlier decisions.

This forces instructors to commit early, especially around AI extraction and grading, without enough context to feel confident. In short, complexity was front-loaded instead of progressively revealed: instructors faced high-stakes choices before the flow gave them the context to make them. The funnel data confirms this: the largest drop sits exactly at the moment users are asked to commit without preview.

AI is framed as a risky fork, not a helpful partner.

AI capabilities are positioned as branching paths: AI vs. non-AI. That framing amplifies fear of choosing wrong, makes AI feel irreversible rather than assistive, and turns grading into a high-stakes configuration problem.

Instructors don't want to choose an AI workflow. They want to grade, with AI available where it's helpful.

Small inefficiencies undermine the value proposition.

Editing layouts for questions and rubrics are constrained, in-context guidance is sparse, error messages explain what went wrong without showing how to recover, and small visual inconsistencies erode perceived quality. Each issue is minor on its own, but together they contradict the product's core promise of saving time.