Sales research that stops at a budget
Prospect proposal engine
One prospect domain in; site audit, five-vendor AI study, SEO audit and three design comps out, then a proposal.
It cuts a week of agency research to one run with a hard spend ceiling. Per-run ceilings and per-vendor ledgers mean the cost is known before and after every run.
- When
- Jul to Nov 2026
- Status
- Live
- proposals produced
- 13
- lines in 73 Python files
- 26,700
- AI vendors studied per prospect
- 5
- highest spend ceiling on a run
- $50
What this does not claim
One signed contract is the only recorded win. Close rate across the thirteen is not measured.
01
The problem
A good proposal needs real research on the prospect: how their site performs, what five AI engines say about them, and what the design could be. Done by hand it takes days, and API costs can run away.
02
What I built
A pipeline with human gates for scope and ceiling, spend, report acceptance, publish and pricing. It audits the site, runs the five-vendor AI study, scores the results, builds three design comps, and publishes a private page. A spend ledger tracks every vendor call against the ceiling.
03
What it produced
Thirteen proposal packages. One run spent $27.87 of a $50 ceiling. One of the proposals led to a signed university contract.
Measured from the repository
73 Python files, about 26,700 lines, used for 13 proposals; run ceilings of $25 to $50.
The stack
- Python
- Five AI vendor APIs
- DataForSEO
- Headless Chrome
- Next.js delivery host