Inspiration
Procurement involves too many manual steps: predicting demand, checking contracts, comparing suppliers, getting approvals, and placing orders. We wanted to build an AI system that does the work, not just talks about it.
What it does
Our platform uses five AI agents to autonomously manage procurement:Forecast -Contract RAG -Negotiate - Approve -Execute It predicts stockouts, retrieves supplier contract terms, evaluates suppliers, handles approval decisions, and executes procurement actions. High-value purchases are routed to humans for approval.
How we built it
We built the system using Gemini, Google ADK, Vertex AI RAG, Cloud Run, Firestore, Pub/Sub, FastAPI, and Expo/React Native. Google ADK orchestrates the specialized agents, while Vertex AI RAG grounds decisions in real supplier contracts.
Challenges we ran into
The biggest challenge was coordinating multiple agents into one reliable workflow while maintaining state between steps. We also had to balance autonomy with safety, ensuring that high-impact purchasing decisions can still require human approval
Accomplishments that we're proud of
We built a working Taskmaster-style autonomous workflow rather than a chatbot. The system can take a procurement request through multiple AI-driven stages and produce an actionable procurement outcome.
What we learned
We learned that successful agentic systems depend as much on workflow and state design as they do on the underlying AI model. Grounding decisions with real business data and keeping humans in control of critical actions were also essential.
What's next for Procurement Platform
We plan to connect the platform to real ERP systems and supplier APIs, enabling live inventory monitoring, supplier interactions, automated purchase orders, and longer-running autonomous procurement workflows.
Log in or sign up for Devpost to join the conversation.