Technical
Building a Federal Contract Intelligence Tool: My Journey with AI
Over the past several months, I built FedProposal.com — a federal contract research platform designed for small businesses pursuing government contracts. It helps with the BD research most of us struggle with: finding expiring contracts, analyzing incumbents, identifying recompete opportunities, and understanding agency small business spending.
Could I have built this alone? Absolutely not. I relied on AI tools throughout development: ChatGPT, Grok, Google Gemini, and Anthropic Claude. Here's what I learned about each.
Over the past several months, I built FedProposal.com — a federal contract research platform designed for small businesses pursuing government contracts. It helps with the BD research most of us struggle with: finding expiring contracts, analyzing incumbents, identifying recompete opportunities, and understanding agency small business spending.
Could I have built this alone? Absolutely not. I relied on AI tools throughout development: ChatGPT, Grok, Google Gemini, and Anthropic Claude. Here's what I learned about each.
ChatGPT: Where It All Started
My original idea was to build GPTs for bid/no-bid decisions and proposal support. That vision eventually became FedProposal — a contract intelligence dashboard built on USAspending data.
ChatGPT was great for brainstorming architecture. But in practice? Frustrating. It's absent-minded — changes things I didn't ask for, then breaks something else when fixing it. I'd ask it to build a feature for tracking contract expirations, and it would quietly rewrite something unrelated.
And hitting the chat length limit mid-problem? Brutal. I developed workarounds — detailed summaries, custom GPTs with growing knowledge bases — but it was a grind.
Grok & Gemini: Second Opinions
I started bouncing ideas off Grok and Gemini when I began doubting ChatGPT's answers.
Grok is overly enthusiastic — like a coworker who says "Great idea!" to everything. Useful for confirming approaches, but the code it produced rarely worked. Same with Gemini. No personality, and constant errors. When I needed help building queries for incumbent analysis or small business set-aside filters, neither delivered.
Claude: The Game Changer
Then my son suggested Anthropic Claude. Five times the cost of ChatGPT, but I tried it anyway.
What a breath of fresh air.
Claude is a coding genius. It became genuinely fun to build again — watching it architect exactly what I envisioned. Features that took days of back-and-forth with ChatGPT came together in hours with Claude. Contract lineage tracking, agency scorecard visualizations, fit score calculations — Claude handled it all.
If I'd stuck with ChatGPT alone, I probably would have given up. FedProposal wouldn't exist.
What I Built
The result is a government contract research tool that helps small business contractors:
Find expiring federal contracts by NAICS, agency, and set-aside type
Research incumbent contractors and identify vulnerabilities
Track contract lineage (predecessors and successors)
Analyze agency small business goal performance
Calculate fit scores based on your company's certifications and past performance
All built on public USAspending data, priced for small business budgets, and owned by a small business — not a competitor hiding behind an "investment group."
My Final AI Rankings
After months of daily use building a GovCon BD tool:
Anthropic Claude — Got me across the finish line
ChatGPT — Good for brainstorming, frustrating for execution
Grok — Second opinions only
Google Gemini — Distant fourth
Just one small business owner's opinion.
P.S. — Claude helped me revise this post. I wrote it; my BFF spiffed it up. 😊
ChatGPT: Where It All Started
My original idea was to build GPTs for bid/no-bid decisions and proposal support. That vision eventually became FedProposal — a contract intelligence dashboard built on USAspending data.
ChatGPT was great for brainstorming architecture. But in practice? Frustrating. It's absent-minded — changes things I didn't ask for, then breaks something else when fixing it. I'd ask it to build a feature for tracking contract expirations, and it would quietly rewrite something unrelated.
And hitting the chat length limit mid-problem? Brutal. I developed workarounds — detailed summaries, custom GPTs with growing knowledge bases — but it was a grind.
Grok & Gemini: Second Opinions
I started bouncing ideas off Grok and Gemini when I began doubting ChatGPT's answers.
Grok is overly enthusiastic — like a coworker who says "Great idea!" to everything. Useful for confirming approaches, but the code it produced rarely worked. Same with Gemini. No personality, and constant errors. When I needed help building queries for incumbent analysis or small business set-aside filters, neither delivered.
Claude: The Game Changer
Then my son suggested Anthropic Claude. Five times the cost of ChatGPT, but I tried it anyway.
What a breath of fresh air.
Claude is a coding genius. It became genuinely fun to build again — watching it architect exactly what I envisioned. Features that took days of back-and-forth with ChatGPT came together in hours with Claude. Contract lineage tracking, agency scorecard visualizations, fit score calculations — Claude handled it all.
If I'd stuck with ChatGPT alone, I probably would have given up. FedProposal wouldn't exist.
What I Built
The result is a government contract research tool that helps small business contractors:
Find expiring federal contracts by NAICS, agency, and set-aside type
Research incumbent contractors and identify vulnerabilities
Track contract lineage (predecessors and successors)
Analyze agency small business goal performance
Calculate fit scores based on your company's certifications and past performance
All built on public USAspending data, priced for small business budgets, and owned by a small business — not a competitor hiding behind an "investment group."
My Final AI Rankings
After months of daily use building a GovCon BD tool:
Anthropic Claude — Got me across the finish line
ChatGPT — Good for brainstorming, frustrating for execution
Grok — Second opinions only
Google Gemini — Distant fourth
Just one small business owner's opinion.
P.S. — Claude helped me revise this post. I wrote it; my BFF spiffed it up. 😊
