Intake-to-Owner Auto-Router

The problem this kills

A request lands - a new project, a service ticket, a change form, an access ask - and then it floats. Which team owns this? Whose queue does it go in? Is it big enough to need the senior person, or can it go to whoever's free? Did anyone actually pick it up, or is it sitting in a shared inbox that everyone assumes someone else is watching?

In most intake teams the "routing process" is a coordinator squinting at each request, half-remembering the rules ("anything from Legal over $50k goes to Priya, I think?"), and forwarding it on with a hopeful note. Things sit unassigned for days. Easy ones land on the wrong queue. The same request gets routed twice because two people grabbed it. And when someone asks "why did this go to that team?", the answer is a shrug.

This tool replaces the guesswork with a clear, repeatable routing engine - one that reads each request, proposes the right owner from your rules, explains why, and won't commit anything until a coordinator confirms it.

What you'll build

A small internal web app, just for your team, that:

• Imports a batch of incoming requests from a CSV or sheet (request ID, type, department, description, value, requester).
• Reads your routing-rules table - your own rules for "this type, from this department, with these keywords, in this value band, goes to this team / queue / owner" - and proposes the right routing for each request automatically.
• Shows a plain-English reason for every proposal ("matched rule: type = Access Request + department = Finance -> IAM queue, owner Dana").
• Puts each proposal in front of a coordinator who approves it as-is or reassigns to a different team, queue, or owner.
• Commits the routing only after the coordinator confirms - then emails the assigned owner via Resend.
• Lets you edit the rules table in the app, no code required, so routing keeps up with reorgs and new request types.
• Dedupes on request ID so the same request can't be routed twice.
• Exports a clean routed-assignments CSV in the exact columns your system of record expects.

What's inside the Implementation Plan

The plan is a single markdown file you paste into Claude Code (a free AI coding agent). It walks the agent through building the whole tool, step by step, each step ending with a ready-to-paste prompt.

The most important part: the plan opens by interviewing you about your business. Before it writes a single line, the agent asks how requests reach you today, what your routing rules actually look like, the real field names and codes in your data, your teams and queues and who owns what, your typical and peak request volumes, and your messiest edge cases (a request that matches two rules, an owner who left, a type nobody has a rule for yet). It reads a short tailored spec back to you, you confirm it, and only then does it build - so you get a router shaped to your intake process, not a generic template you have to bend to fit.

Inside you'll find:

• The discovery interview and how the agent turns your answers into the data model and the rules engine.
• The full build: database, login, CSV import with duplicate guards, the rules engine with explainable reasons, the coordinator review-and-reassign screen, the commit-and-notify flow, and the editable rules table.
• The hard human approval gate - the AI proposes, a coordinator confirms, and only then is anything committed.
• Verification steps so you can prove it works, and the CSV-export fallback so it's fully usable even before you connect it to your system of record.

The governance it includes (this is the point)

This isn't a toy. The plan builds in the controls an intake team actually needs:

• Login so only your team can see or touch anything.
• Row-level security so people only see the requests and assignments that belong to your organization.
• A complete audit trail - every import, proposal, approval, reassignment, commit, and notification is logged with who and when.
• A hard human-in-the-loop gate - the AI proposes the routing, but a coordinator must review and confirm; nothing is committed to the system of record until a person says yes.
• Duplicate guards so the same request ID can't be routed or committed twice.

Who it's for

Intake coordinators, PMO and shared-services dispatchers, service-desk leads, and ops/BPM folks who are the human router for a stream of incoming requests - and who want a real, auditable routing tool without hiring a developer or buying a heavyweight workflow platform. You don't need to write code. You need your requests CSV, your routing rules, your teams directory, and an afternoon.

You've got this - paste the first prompt and let the agent interview you.