Devin Huber
Zapier · 2 agents

Guest Application Flywheel

Two agents that screen incoming podcast guest applications, draft personal declines for weak fits, and send leadership a monthly view of the pipeline.

For a businessCuts hours of manual screening so strong podcast guests aren't lost in the pile, while a person keeps the final say on every decision.

Built for
Leadership-development institute's podcast (current work)
Record
Google Sheets
Output
Recommendations, decline drafts, monthly digest
Trigger
New application · monthly
15point scoring rubric
5scored criteria
2agents with separate tool sets

01 · ContextThe problem

Guest applications arrive through a form and were screened by hand. Strong fits got buried among weak ones, and every decline deserved a personal reply that took time to write.

02 · PipelineHow it works

  1. New application

    The screening agent starts the moment a form is submitted.

  2. Research the applicant

    It looks the applicant up online for context the form doesn't give.

  3. Score against the rubric

    Five criteria, scored out of 15, plus the agent's own judgment.

  4. Write the record

    The result is written to a Google Sheet that both agents share.

  5. Recommend

    Leadership gets a pass, consider or fail recommendation within moments of submission, and can overrule any score.

  6. Monthly digest

    A second agent reads the sheet each month and summarizes the pipeline for leadership.

03 · ChoicesDesign decisions

Safety by limiting tools

Each agent holds only the actions its job needs, and nothing more.

Applications are data

Application text is treated as untrusted input, with defenses against instructions hidden inside it.

No half-written rows

Writes happen in a fixed order so the sheet never shows a partly processed application.

People keep the final say

Any score can be overruled, and decline emails are drafted, not sent automatically.

04 · EvidenceHow I tested it

Shipped with a written build sheet and test plan, and a step budget per run.

Want to see this one run?

I'll screenshare the workflow, the test set and the results, including what didn't work.

Get in touch