FlowPilot

FlowPilot is an AI-powered workflow approval platform designed to help organizations automate internal business processes through reusable workflow templates, structured approvals, and intelligent decision support. Instead of relying on fragmented email threads, spreadsheets, or chat messages, FlowPilot provides a centralized workflow engine that tracks every approval step, maintains complete audit trails, and ensures consistent execution across teams. Organizations can define custom workflow templates, create workflow instances, manage approval stages, and monitor the complete lifecycle of requests from creation to completion. Every workflow transition is recorded, providing transparency, accountability, and traceability for business-critical operations. To enhance decision-making, FlowPilot integrates an AI review engine powered by Qwen3-0.6B running on AMD Developer Cloud through a self-hosted vLLM OpenAI-compatible inference server. Before an approval decision is made, the AI analyzes the workflow context and generates structured insights across multiple dimensions, including Security, Compliance, Operational Risk, Cost Impact, and an Overall Recommendation. The AI acts as an intelligent assistant, while deterministic business rules ensure that core workflow logic always remains under application control. **Built as a full-stack monorepo using Express, TypeScript, Next.js, PostgreSQL, Prisma, and Tailwind CSS, FlowPilot demonstrates how modern AI can be integrated into enterprise workflow automation while maintaining modular architecture, auditability, and scalability. By leveraging AMD Developer Cloud for inference, the project showcases practical deployment of open-source language models using vLLM to deliver real-time AI-powered workflow analysis for business applications.