Step01
Set the Agent Trigger
Create a Zapier Agent triggered by a new note in the authorized Granola interview folder. Limit the trigger to the intended role or hiring loop and pass the stable candidate and interview identifiers.
IntermediateOperations
How to Build an AI Agent to Evaluate Job Candidates with Zapier
Build a hiring thought partner that evaluates interview evidence against a job rubric, removes unnecessary personal data, and gives the interviewer a traceable recommendation and coaching without making the decision.
Source episode · 14:00
How I AI: Zapier CEO Wade Foster's Playbook for AI-Powered Recruiting and Culture Buildingwith Claire Vo
Wade triggers a Zapier Agent from a new Granola interview note, compares the transcript with a job description and values rubric, emails a yes, no, or maybe recommendation, removes PII, and adds interviewer coaching.
Before you start
What you need
- Authorized Granola interview notes
- A role-specific job description
- A reviewed, behavior-based values rubric
- Zapier Agents and an approved email recipient
- A hiring owner responsible for the final decision
What you’ll make
A private evaluation email with an evidence-linked recommendation, missing evidence, confidence, and interviewer coaching, delivered as decision support rather than an automated hiring action.
Tools used
- GitHub CopilotVisitGitHub Copilot
AI pair programmer
- ZapierVisitZapier
Workflow automation platform
- Granola
AI-powered meeting note-taker and summarizer
Step by step
The workflow
Follow the sequence once, then adapt the prompts, checks, and handoffs to your own setup.
5 steps
Step02
Provide Agent Instructions
Instruct the agent to compare only job-related interview evidence with the job description and behavior-based rubric. Require citations, confidence, missing evidence, and a recommendation that cannot change candidate status.
Example prompt
Act as a hiring evaluation assistant, not the decision-maker. Review this interview against the attached role rubric. For each criterion, cite transcript evidence or mark not assessed. Exclude protected traits and unnecessary personal details. Return YES, NO, MAYBE, or INSUFFICIENT EVIDENCE with confidence, 3 to 5 sentences of reasoning, open questions, and no automatic hiring action.
Step03
Upload Knowledge Sources
Attach the current job description and reviewed values rubric as versioned knowledge sources. Define each criterion, rating anchor, and prohibited signal so evaluations use the same standard.
Step04
Define the Final Action
Send the report only to the accountable interviewer or hiring owner. Put the recommendation and confidence in the subject, and include criterion evidence, missing evidence, and the source interview link in the body.
Step05
Refine with Copilot (Optional)
Use Copilot to add an interviewer-coaching section and strengthen privacy rules. Test the agent on representative and deliberately incomplete interviews before relying on it.
Example prompt
Update the instructions to remove candidate names and unnecessary identifying details from the evaluation. Add interviewer coaching that names rubric areas not assessed and suggests job-related follow-up questions. Never infer an answer from silence, communication style, accent, age, location, family status, disability, or other protected information.
Putting the YES/NO/MAYBE decision directly in the email subject line allows for much faster triaging of interview feedback in a busy inbox.
What good looks like
- Every assessment cites job-related evidence from the transcript.
- Protected traits and unnecessary personal information are excluded.
- Missing or weak interview evidence lowers confidence instead of becoming a negative inference.
- A human hiring owner makes and records the final decision.
Build your next product with ChatPRD
Turn an idea into a PRD, user stories, and a plan.
After the steps
Runbook notes
How to recover when the loop fails and where human judgment helps.
Recover
If it goes sideways
- Candidate identity or sensitive personal data appears in prompts, logs, or examples
- Minimize fields before model access, restrict retention and recipients, and redact output that is not needed for the decision.
- The agent infers values, ability, or fit from style or protected characteristics
- Use anchored job-related criteria, prohibit protected-trait inference, and audit recommendations across comparable candidates.
- The recommendation sounds certain when the interview did not cover a criterion
- Require citations, missing-evidence fields, and confidence; return maybe or insufficient evidence instead of guessing.
- An AI recommendation advances or rejects a candidate automatically
- Remove downstream status changes and route the report to the accountable hiring owner for a recorded human decision.
Keep building
Related workflows
advancedEngineeringClaude
How to Implement AI Governance with Context-Aware Project Controls
Create context-aware guardrails for your enterprise AI agent using a 'Project' governance layer. As a project owner, define rules, scope tools, and require human-in-the-loop approval for sensitive actions, ensuring safe and effective AI use.
Open
intermediateOperationsKai (Stripe)
How to Turn a One-Off AI Chat Session into a Reusable Workflow for Your Team
Transform a successful AI chat session into a reusable skill that anyone in your company can use. Package prompts, tools, and code into a standardized workflow, making it easy to share and scale automation across your organization.
Open
beginnerDataKai (Stripe)
How to Create a Custom Data Dashboard On-the-Fly with a Natural Language AI Agent
Use a natural language AI agent to instantly generate and iterate on custom data dashboards. This workflow queries your company's data securely, allowing non-technical users to perform their own 'last mile' data analysis without filing tickets.
Open
Start shipping
better products.
Join 100,000+ product managers who use ChatPRD to write better docs, align teams faster, and build products users love.
Free to start
No credit card
SOC 2 certified
Enterprise ready