LunaPromptsThe agent says it handled scheduling. The recruiter assumes the calendar is confirmed. The candidate thinks they only shared their availability. Everyone acted reasonably, but the workflow never defined what “done” meant.
Useful recruiting automation starts with a smaller, clearer promise. Give each agent a defined task, an observable completion state, and somewhere to send an exception.
LunaPrompts AI copilots support candidate evaluation and routine recruiting work. AI Judge helps reviewers interpret assessment evidence, while calling agents can collect information, coordinate interview slots, conduct defined phone screens, and answer routine process questions.
Coordinate an interview without inventing a booking
Recruiting coordinatorIllustrative scenario01 / 03Confirm the candidate’s availability
The candidate can meet Thursday afternoon in their local time zone.
Start with a scheduling workflow because its outcome can be checked. Define the inputs: candidate identity, interview type, expected duration, relevant calendars, and time zone. Define the output: a confirmed event or a clearly described exception.
The calling agent gathers availability and communicates the next step. If the candidate says “Thursday afternoon,” the workflow needs a time zone and a specific proposed time. If the calendar cannot confirm the slot, the message should say it is proposed or awaiting confirmation. It should not imply that the interview is booked.
Give the recruiter enough context to resolve an exception without repeating the entire conversation. A useful handoff includes what the candidate requested, what the agent attempted, and the remaining decision. Avoid collecting unrelated personal detail just because the conversation makes it possible.
Apply the same discipline to AI Judge. Define the criteria it assists with, the evidence a reviewer can inspect, and how disagreement is handled. The useful output is an evidence-linked review aid. The recruiter and hiring manager still own the hiring decision and the relationship with the candidate.
AI Judge: evaluate against the role's expectations
AI Judge assists with reviewing a candidate's performance for the enterprise and role being assessed. Its evaluation should connect to the work, the scoring criteria, and the standards your team has defined.
AI Judge is designed to adapt to reviewer feedback so its assistance can reflect your team's documented evaluation standards. Use feedback to correct an interpretation, clarify a criterion, or identify missing evidence. Review the criteria themselves when a pattern of disagreement suggests that the team is measuring different things.
What a reviewer should get from AI Judge
A helpful evaluation makes the candidate's demonstrated strengths, gaps, and unanswered questions easier to inspect. The recruiter can use it to prepare a handoff. The hiring manager can check the supporting work and decide what to explore next.
When the AI and a reviewer disagree, the discrepancy should prompt examination of the evidence or criteria. AI Judge adds an assessment perspective; the human team owns the decision.
Calling agents: four recruiting workflows
1. Gather candidate information
A calling agent can collect the basic details needed for a defined hiring step. For example, the conversation may confirm a candidate's relevant experience, availability, or understanding of the role's working arrangement.
Prepare the information requirements in advance. Collect details that serve the hiring process, and route unclear or sensitive questions to a person rather than improvising an answer.
2. Coordinate interview slots
The agent can discuss availability and help coordinate an interview time. A useful workflow establishes the relevant time zone, candidate preference, and whether the proposed slot has actually been confirmed.
Coordination should reflect the calendar and scheduling capabilities configured for the organization. A proposed slot and a booked interview are different states, and the candidate needs to know which one applies.
3. Conduct a phone screening round
A defined L0 phone screen can explore basic role fit and initial understanding before a longer assessment or interview. The conversation should use role-related questions and produce evidence the recruiter can review.
For example, an agent might ask a backend candidate to describe a service they worked on and a problem they helped resolve. A human reviewer can then decide whether the answer supports progressing to practical evaluation.
4. Help with routine information
Candidates may need to understand the next step, an assessment instruction, or the interview process. A calling agent can assist with the approved information available to it and hand off questions that require a recruiter.
Avoid letting a process assistant invent hiring outcomes, compensation commitments, or company policies. A useful agent knows when the available information is insufficient.
Make the handoff clear
Candidates should know when they are interacting with AI and how to reach the recruiting team. Reviewers should understand what the agent completed, what it could not resolve, and what needs human follow-up.
Before deployment, agree on the agent's purpose, approved information, escalation path, and the evidence the team wants to retain. This makes the assistance useful in the actual workflow instead of creating another stream of disconnected activity.
A review framework your team can use
The following is an editorial checklist for this workflow. Adapt it to the role, the candidate journey, and your configured environment.
| Review area | What to establish |
|---|---|
| Task | The bounded job the agent should perform |
| Completion | What the system can actually confirm |
| Exception | What the agent must hand off |
| Review | What a person can inspect afterward |
Automation can create extra work when it hides incomplete tasks behind a success message. Make “needs attention” a normal, useful outcome.
Put the guide to work
Choose one repetitive recruiting task. Write its inputs, completion state, and three exceptions that should reach a person.
Download the editable worksheet to record the decisions and open questions with your team. The examples in this guide are illustrative, not customer results or validated selection rules. Product availability and setup depend on the workflow agreed for your account.
A focused checklist, a worked scenario, and questions to resolve together.
Frequently asked questions
Does AI Judge make the hiring decision?
No. It supports evaluation. Recruiters and hiring managers review the evidence and decide who progresses or receives an offer.
Can calling agents replace every recruiter conversation?
No. They are suited to defined tasks. Complex questions, exceptions, relationship building, and consequential discussions may need a recruiter.
How does AI Judge adapt to our evaluation style?
The intended workflow uses enterprise criteria and reviewer feedback for calibration. Confirm the currently available feedback controls, review process, and data boundaries during configuration.
How is a calling agent different from a full AI interview?
A calling agent can perform an operational task or a short phone screen. A full AI interview is a broader evaluation stage designed around the role's interview framework.
Keep building your hiring workflow
- Candidate analytics: The scores match. The follow-ups should not.
- AI interviews: More first conversations. A clearer handoff to people.
- Interview copilot: They already took the test. Make the interview add something.
Explore recruiting copilots with LunaPrompts or browse all 19 hiring field guides.
