Lab · self-directed work

Working prototypes, and the decisions people still make

Four working prototypes I designed and built, each asking one design question. Where AI is involved, it suggests, and a person decides.

The four pieces

  1. Self-directed · Simulated data · Not yet tested

    Marked paid, but the supplier still asks “Where’s our payment?”

    I decided what a payables specialist on the phone sees first.

    After Send, a payments prototype, first built with AI in a day.

    • Issued, charged, and matched, as separate facts
    • What to tell the supplier, and when to follow up
    • Log or summary, and the proposed test to choose
    Enterprise UX Commercial payments
  2. Self-directed · Sample data · Not yet tested

    The Shortlist: one shopping trip, carried across phone, store, and person

    Can a shopping trip carry its context into the store?

    • Shopper and staff steps as one service map
    • A working prototype of the in-store moment
    • What to test first, and what would disprove it
    Service design Service map
  3. Self-directed · Simulated data · Not yet tested

    Everyone talks about orchestration. Here’s one you can try.

    Agents built three fixes from my design system and rules. All passed the checks. One call stayed mine: when to stop.

    Agent Review: AI agents build the options; I decide.

    • One decision to try: when Archive should stop
    • Three working fixes, and why I kept the review
    • The rule the next agent builds from
    Case study
  4. Self-directed · Synthetic data · Not yet tested

    Dispatch Cockpit: an AI pick a dispatcher can check and overrule

    Can a dispatcher check an AI’s truck pick faster when the screen shows what needs checking?

    • Every factor’s value, and how fresh it is
    • A close call that names its tradeoff
    • Legal limits enforced before anything is scored
    Decision support