ProcureFlow AI

rocureFlow AI solves a common enterprise workflow problem: purchase approvals are slow, fragmented, and hard to audit. A single request often moves across finance, vendor risk, policy review, and management sign-off with no clear shared state. ProcureFlow AI turns that process into a coordinated multi-agent workflow on Band. IntakeAgent parses and normalizes the purchase request. RiskAgent evaluates vendor and spend exposure using API-compatible reasoning. PolicyAgent checks procurement controls and can flag or block the request. FeatherlessReviewAgent runs an independent open-source model review through Featherless AI before final approval. ApprovalAgent gathers all context into a human-readable decision memo and waits for explicit human sign-off. Each agent posts its handoff/result into the same Band chat room using External Agent API keys, so judges can inspect the coordination live. After the human decision, the system writes an audit packet containing the original request, AI/ML risk report, policy verdict, Featherless review, approval summary, signer, timestamp, and SHA-256 hash. The frontend provides a complete enterprise workflow: Submit, Dashboard, Approvals, Agent Timeline, and Audit Packet. The result is a practical internal enterprise workflow where Band is the visible coordination layer, humans remain in control, and every decision is traceable.