Intake-to-Procure: What Breaks in the Real World

• • The “front door” asks for too much too early, so people avoid it
• • Routing and approvals aren’t owned (no decision rights, no service levels)
• • Workarounds become faster than the designed path, so behavior follows
• • Preferred suppliers and buying channels aren’t embedded, so compliance stays optional
• • KPIs exist, but there’s no weekly ownership or action loop

• • Start with the 5–7 most common request types and make those paths effortless
• • Define decision rights (RACI) before you configure anything
• • Add a routing owner: someone accountable for “where does this go next?”
• • Embed preferred suppliers and policy guidance into the request path (not a PDF link)
• • Treat exceptions as data: track reasons, fix the root cause, then update the flow

• • Time-to-first-action (from request submitted → someone starts work)
• • Approval cycle time (by request type and value/risk tier)
• • Exception rate + top 5 exception reasons (then kill them one by one)
• • Spend routed to preferred channels (directionally; definitions matter more than perfection)

Use LLMs to summarize requests, classify categories, and draft missing fields. Then a human validates before it changes routing, suppliers, or approvals. Done well, this speeds up triage without eroding trust.