Agents Lookup

Use cases / Decision models (Jev)

Triage a shared inbox by what each email needs

Every email to a shared inbox gets one intent label and an urgency flag from a decision model, so confident ones are labelled and routed in Gmail while unclear ones wait for a person to sort.

DifficultyBeginner
Setup time1–3 hours (estimate)
CategoryDecision models (Jev)

Author's figure: Author's figure: 500 emails classified for 3.5 cents (Riley Brown). Reported by the builder, not measured by us.

The use case

A shared inbox for a small business collects orders, questions, complaints, invoices, sales pitches, and spam, and someone spends the first hour of each day sorting it. Jev reads each new email and answers which intent it has, from a fixed list, and whether it is time-sensitive. Emails labelled with a confidence of 0.85 or higher are tagged and moved to the right Gmail label. Lower-confidence emails stay in the main inbox with a needs sorting tag. Nothing is replied to or deleted automatically. The team still reads and answers every email, decides what counts as urgent for a customer, and can retag anything Jev got wrong. Start with the threshold high, review a week of labels against what your team would have chosen, and lower it only when the labels hold up.

Example workflow

  1. A Zapier trigger fires on each new email in the shared Gmail inbox.
  2. The subject, sender, and first part of the body are sent to Jev as the state.
  3. Jev answers a Choice question, what does this email need, with the allowed answers new order, order question, complaint, invoice or billing, sales pitch, spam, and other, and a Noul question, the sender needs a reply today.
  4. Zapier applies the matching Gmail label when the confidence clears the threshold, and adds needs sorting when it does not.
  5. Emails marked time-sensitive with high probability are posted to a Slack channel so someone picks them up early.
  6. Each week a person reviews a sample of labels and corrects any that were wrong, adjusting the threshold if needed.

Tools

Risks and guardrails

A similar pattern in a public build: https://madewithjev.com/builds/500-emails-3-cents. See also Decision models.

Related recipe

Support Ticket and Review Triage with Policy-Cited Draft Replies: An agent tags every helpdesk ticket and public review by intent, urgency, and sentiment, routes it, and drafts a reply that cites your help center. People approve refunds and all public replies.

Want it set up for you? Request a quote.

Sources

Written by Agents Lookup. Credits and official pages for the tools this use case names: