✅ Built a 30+ agent orchestration system turning raw client artifacts into a complete BRD, user stories, effort estimates, and proposal deck in one command
✅ Scrum Poker skill generates Traditional vs. Gen-e2 Fibonacci estimates using 6 developer-persona agents
✅ Transformed the pre-commitment process — from idea to multiple 90%-complete POCs, reducing RFP friction
What is this project about?
The AI-Powered Presales Pipeline is an internal PALO IT tool that transforms raw client artifacts — PDFs, Word documents, PowerPoint decks, Excel sheets, images, and meeting notes — into a complete Business Requirements Document, user stories, effort estimates, and a 12-slide proposal deck in a single command.
The pipeline runs a 30+ agent orchestration system: an orchestrator coordinates group leads (Discovery, Requirements, Delivery), each spawning parallel sub-agent teams. A Scrum Poker skill pits 6 developer-persona agents against each other in Fibonacci estimation rounds — comparing Traditional vs. Gen-e2 delivery estimates. A Brainstorm skill deploys client, sales lead, and tech lead personas to resolve contested scope or strategy decisions.
This fundamentally changed how PALO IT approaches the pre-commitment phase. From a project idea alone, we can now generate multiple 90%-complete POCs to show clients exactly what we can deliver — a genuine competitive edge at the RFP stage.
Takeaway #1: The hardest part of building an agentic system isn't the agents themselves
The hardest part of building an agentic system isn't the agents themselves — it's designing the context handoff between stages so each agent has exactly what it needs and nothing it doesn't.
Takeaway #2: AI-powered presales only works if the output is client-ready without human polish
AI-powered presales only works if the output is client-ready without human polish — that raised my bar for what 'done' means in an agent pipeline significantly.
🛠️Tech Stack
Backend
Project Management
