The pain

Inbound is a good problem until it's a tax.

You spend an hour each morning reading messages that all look hopeful — and all turn out to be tire-kickers, students, peers, or people pitching you. Two of them are real. You'll find those on Thursday, you tell yourself.

By Thursday, the real prospects have moved on. Inbound feels generous and behaves like rent.

Who has it

Anyone with any inbound surface — a contact form, a DM funnel, a "book a chat" link, an email address listed publicly.

Volume isn't the threshold. Ten messages a week is enough to lose the signal. Past that, signal drowns first; ego ("look at all this interest") drowns second.

The fix

One Claude prompt that scores every inbound against a fixed rubric — and routes them. Call now. Schedule. FAQ. Archive.

The rubric is yours. The scoring is the agent's. The decision stays cheap.

Setup (45 min, once)

  • Define your rubric. Be specific. The whole leverage is in this list being yours, not a generic template:
    • - Budget signal — they name a budget, a timeline, or a specific project (yes/no) - Fit signal — they describe a problem you actually solve (yes/no) - Effort signal — the message shows effort, not copy-paste (yes/no) - 3/3 → call now. 2/3 → schedule. 1/3 → FAQ. 0/3 → archive.
  • Build the prompt:
    • > Read this inbound message. Score it against the rubric below. Output: score, one-sentence reasoning, suggested next action, and (if calling) a 3-line draft reply.
  • Wire it up. Easiest path: Zapier → Claude → back into Notion + a Slack ping for 3/3 hits.

    • Running it (10 min, daily)

    Open the Notion view filtered to "score 3/3" first. Reply to those. Then 2/3. Skip the rest. Trust the rubric — the rubric does not get tired.

    What it looks like running

    Friday's inbound dump, sorted before you've finished coffee:

    3/3 — Call now "Mai (Saigon, runs 12-person Pilates studio): looking for someone to set up an automated booking + WhatsApp reminder system. Budget mentioned: $1,500. Timeline: this month."
    >
    2/3 — Schedule "Tuấn (HCMC freelance designer): wants AI workflows for client revisions. No budget yet, but specific problem."
    >
    1/3 — Send FAQ Generic "loved your post, want to chat" — no specifics.

    Twenty messages collapsed into three replies. The 3/3 reply takes three minutes. The two real prospects, the ones you would have lost on Thursday, are the ones you talk to first.

    Why this works

    This is a Fixer-tier workflow — AI removes a measurable bottleneck (response time on real leads) with a measurable result (faster replies, higher close rate on hot inbound).

    The trust ask is small and specific: "Can AI read a message and tell me how serious it is?" Yes — better than you can on a tired Thursday afternoon. The rubric doesn't get tired. Your gut does.

    Don't skip the rubric. The whole leverage is in it being yours, specific, and stable. If you can't articulate the rubric, you can't outsource it. And maybe — uncomfortable thought — that's why your inbound feels overwhelming in the first place. Not too many messages. Not enough rubric.