SparkBox

The suggestion box in a large organisation collects raw ideas that managers cannot process. Ideas arrive in uneven prose, with no cost, benefit or owner attached, and most are never read closely.

The specification is the only artefact. An employee submits an idea in plain language. A hosted language model turns it into a structured proposal with a problem, a solution, a benefit, an effort and a category. Managers review it through a status workflow (Submitted, In Review, Solution Given or Rejected), and an Innovation Head dashboard tracks it to implementation. Role-based access and an admin “view as” feature are part of the design.

The documents describe a Base44 version used in teaching, and a planned self-hosted rebuild in React, Node, PostgreSQL and Prisma. No model is named.

It has not been measured. I found no application code on either of my drives, no users and no results. It is a specification and a teaching reference build, not a deployed product.

A specification is worth writing even when nothing gets built, because it forces the decisions: who submits, who reviews, what a structured idea must contain. In teaching, its value is as a brief that participants can build against. Its honest limit is that nobody has shown it improves the quality of ideas an organisation acts on.