Confidential · UX · AI Systems
Digital Infuzion writes proposals to win government contracts, which come with tight guidelines, strict evaluation criteria, and a lot of language recycled from previously submitted contracts. Over Summer 2025, I designed an internal AI Proposal Generator: a prompting interface meant to help their team search, adapt, and reuse past proposal content instead of rebuilding every document from scratch.
Presenting the project alongside the other interns at the end of the summer.
The harder part was keeping a person in the loop on tone, strategy, and what the client actually wanted, not just what checked a box. The first version of the interface prompted finished proposals, not winning ones. Technically complete, but flat; not leaving much room for the writer's own judgment.
I went back to the drawing board and rebuilt the flow around guided, Mad Libs style prompting, where the person writing the proposal stayed the strategist and the AI stayed the drafting assist. I also ran a short side study on where AI image generation could actually help with the graphic side of a proposal, and where I think it just can't.
This project is proprietary and confidential to Digital Infuzion, so I can't post the full case study publicly. The deck below illustrates the process, both interface iterations, and my specific findings.
