Overview
At a senior-healthcare company, I handled proposals for government demonstration projects and support programs. Once it became clear that writing quality was the bottleneck, I turned proposal writing itself into a repeatable system.
Problem
I was writing proposals alone, and each one required interpreting the call for proposals, market research, technology trends, budgeting, and responding to reviewers. Evaluation feedback on earlier proposals consistently pointed to writing quality as the bottleneck.
Approach
I broke proposal writing down into an AI agent team with 7 roles (call analysis, market research, technology strategy, business strategy, budget, editing, and evaluation), and added an output numbering rule (00 call analysis → 06 final draft) and a loop of 5 internal mock reviews, creating a system that raises the score before submission. Along the way, I discovered two things. First, the choice of industrial technology classification code determines whether personnel costs can be budgeted in cash (I settled on “SW solution” to secure the basis for cash-budgeted personnel costs). Second, switching the demonstration-site track from public to private changes the scoring sheet itself, from 40 points for public value to 40 points for business viability. I rebuilt the entire narrative around private-sector metrics and applied a non-medical-device positioning principle throughout (no wording about diagnosis, healing, or treatment).
Results
Submitted the demonstration-project proposal before the deadline (late June to early July). The internal mock review score went from 82 to 86, and the business viability item from 22 to 31 out of 40. The market research assets (TAM KRW 7.7 trillion, SAM KRW 85.3 billion, SOM KRW 2.56 billion; 11 competitors; 32 sources) are shared across IR and business planning, and the 7-agent workflow remains a writing system that can be rerun as-is for the next project. In the same period, the company was selected for 4 overseas expansion and acceleration programs (Singapore PoC, age-tech, North America expansion, and an India trade mission); among them, I identified the Singapore PoC program and wrote the submission for North America expansion.
Lessons Learned
The evaluation feedback that earlier proposals received became the reviewer agent for the next one. Knowing that each track has a different scoring sheet moves the score more than writing good sentences. The workflow can be rerun as-is on the next project.