← Selected work

    Gradezy / EdTech

    Remove the work around the work.

    A targeted automation tool for assessment review, designed around the workflow people already use.

    01 / The tension

    Assessment review crossed spreadsheets, learning platforms, email, and repeated file handling. People spent time processing records that did not all require the same level of attention.

    I mapped the workflow to separate repetitive data handling from decisions requiring human judgment. The opportunity was to reduce the path to a useful review, without asking the institution to replace its entire assessment process.

    02 / The decisions

    01

    Automate the repetition. Preserve the judgment.

    I focused on structured processing and identifying records that needed attention. Human review remained part of the workflow where exceptions or interpretation mattered.

    The trade-off

    End-to-end automation promised a larger change, but would also absorb decisions that could not responsibly be treated as routine processing.

    02

    Fit the institution’s workflow.

    I built targeted automation around the existing process and kept its underlying systems as the source of truth where possible.

    The trade-off

    Replacing the wider assessment platform would offer more control, at the cost of a much larger integration and adoption burden.

    03 / What I built

    I owned workflow mapping, product definition, the data flow, application development, and iteration. Delivery included GitHub Actions and deployment automation, with improvements focused on removing unnecessary data and file handoffs.

    04 / What changed

    Working

    automation workflow

    A functioning assessment-review tool built around a concrete operational process.

    Fewer

    manual handling steps

    The project reduced unnecessary handling. A quantified time-saving or error-rate improvement has not yet been recorded.

    The lesson I’m taking forward.

    The important product choice was the boundary of the automation. The tool should bring the right records to a person’s attention, rather than make every decision for them. I would establish a before-and-after measure of review time to make the operational benefit more precise.

    A good place to begin

    Something worth
    thinking about?

    I’d love to hear about it. Open to thoughtful conversations about product, systems, and what comes next.