← Selected WorkBook a Call →
Professional Services · Case Study
Proposal Generation Pipeline
RAG over 12 years of engagement history — partners now draft scoped proposals in 20 minutes instead of two days, with citations to every prior win.
Duration
6 weeks
Stack
ClaudePerplexitySupabaseHermesGoogle Cloud
Outcome Metrics
6×
Faster drafting
3×
Pipeline throughput
12 yrs
Indexed history
100%
Citation coverage
01 · The Problem
A consulting firm was losing pipeline velocity because senior partners were the bottleneck on every proposal. Each new RFP required hours of digging through SharePoint for relevant case studies, scoping precedents, and rate cards. Junior staff couldn't draft autonomously because the institutional memory lived in partners' heads. Win rate was strong, but throughput was capped.
02 · The Approach
- 01Indexed 12 years of proposals, statements of work, and post-engagement reviews into a structured RAG corpus with provenance metadata on every chunk.
- 02Built a guided drafting agent that interviews the partner about the prospect, retrieves the 5 most analogous prior engagements, and drafts a scoped proposal with inline citations.
- 03Added a scope-validation step that flags pricing or timeline assumptions inconsistent with historical precedent before the partner reviews.
- 04Integrated with the firm's CRM so the agent pre-populates client context and writes the final proposal back as a versioned document.
03 · The Outcomes
- Proposal drafting time fell from ~14 hours to under 2 hours per RFP.
- Junior consultants now produce first drafts independently — partners only review and refine.
- Pipeline throughput tripled in the first quarter post-launch.
- Win rate held steady, indicating quality was preserved while speed multiplied.
"It doesn't write the proposal. It writes the proposal we would have written, if we'd had three uninterrupted days. Then we make it better."
— Managing Partner, Consulting Client
More Case Studies
Your Turn
Have a workflow that looks like this?
Book a free 30-minute consultation. We'll map whether AI is the right answer — and where to start.