A Finished Screening Call Is Not a Draft to Rerun

A second AI interview verdict is what a recruiter asked me for in July 2026: wipe a No Go and mail the student a fresh link that evening. The conversation had finished. The notes were specific. The hiring manager’s objection was that the student had sounded nervous in the opening minutes. Another outcome would have made the evening feel kinder. It would also have made the file dishonest.

This is the rule I give teams who hit that moment. Some second chances are real. A do-over because the answer was inconvenient is not one of them.

What a second AI interview verdict creates in HireQwik

A second AI interview verdict is a new outcome on a new conversation for a person who already has an outcome on that job. HireQwik did not build a “try again until it looks better” button. Teams invent one anyway, usually with a reset, sometimes with a second job description that is really the same role.

The damage is not the extra 10 to 20 minutes. The damage is that the panel now has two stories and no rule for which story wins. If the second call is Strong Go and the first was No Go, someone will forward the Strong Go and forget the No Go. If you delete the first, you have edited history. If you keep both and do not say which one binds, you have a fight in the debrief.

People reach for a redo because a single conversation feels final, and final feels harsh at campus volume. Harsh is a reason to read the evidence. Rolling the dice again does not fix harshness. Workable’s guidance on writing interview feedback down is aimed at human interviewers, and the lesson still fits: a vague “let them try again” note is how alignment between recruiter and hiring manager dies. Write what you heard. Then decide.

The design this sits next to is the rule that each opening keeps a single description, a single conversation and a single outcome. I wrote that as its own page, three locks on a screening call. This page is only the failure mode: what to do when someone asks to mint another outcome for the same person on the same role.

When HireQwik’s booking reset is the fair second chance

There is a real retry. I do not want this post read as “never touch the row again.”

Reset the booking when the first meeting is not a sample you would defend. The usual cases:

• The call dropped and the status is a broken or screened session that did not become a real conversation.
• The verdict is Incomplete because too little was assessed, and you have listened long enough to believe the cause was technical rather than a walk-away.
• The candidate could not join from the booked window for a reason you accept, and they never reached a scored call at all.

The reset works from a scheduled or screened state. It refuses with an error if the person is inside a live interview, so a click cannot end the conversation. It keeps the old meeting, marks that meeting superseded, reopens booking, and puts the reschedule count back to zero. The cap itself is 3 attempts inside 72 hours of the first booking, which is explained in the reschedule limit. A genuine connection failure should not burn that cap. A fishing expedition should not get a fresh cap either.

What people do instead, and what fails, is resetting every On Hold “so they get a calmer morning slot.” On Hold is already the pile where a person should listen. A new slot does not tell you whether the first answers were thin. It tells you that your team is uncomfortable closing. Listen to the recording. If you would move them, record On Hold or Go as the human decision and write the sentence. Leave the AI outcome where it is.

A conversation that stops halfway is easy to misread as cruelty. Sometimes the product refuses to tier it. Sometimes it tiers what was actually said and the missing part simply adds nothing. I will not restate those rules here. They are in why a partial call is not a fail. Read that page before you reset, or you will retry people who already have a legitimate tier and leave a true walk-away sitting in the wrong queue.

Why a redo after a finished HireQwik call hides the first result

Assume the call completed. The student answered the questions. The tier is No Go at 4.2, under the 4.5 line on the weighted path, and the note says the examples were generic. The hiring manager likes the college and asks for another attempt tomorrow.

If you reset, three things happen that your debrief will not survive.

First, the new conversation is scored on its own. Nerves may be lower. The student may also have looked up the questions, if a friend from the morning batch described them. You will not know which effect you measured. Campus batches talk. A same-day redo is not a clean retest.

Second, the superseded row is still there, but busy reviewers skip superseded rows. The Strong Go from the evening becomes the only pill anyone mentions in Slack. You have not deleted the morning. You have buried it, which is worse, because a buried row looks like you are hiding it when someone finally opens it.

Third, you teach the team that tiers are negotiable by volume of complaints. The next request arrives faster. I have watched a “one exception” become the unofficial process by Thursday of a drive week.

A University of Washington study found that people followed biased AI hiring suggestions up to about 90% of the time (the UW News write-up). A redo-until-it-matches is the same pull in the other direction: you keep sampling until the pill matches the person you already wanted.

Close the finished call. Open the recording. Read the reason on a No Go, using the same habit as reading a rejection reason. If you disagree, record your own decision and say why. If the question was unfair, fix the screener for the next invite batch. Do not launder a rubric bug through a student’s second afternoon.

How a superseded HireQwik meeting should be talked about

Superseded is a status with a job. It means “this attempt is not the one we are scoring now, and we are not pretending it never existed.”

Use it in three sentences, and no more, when a placement officer asks what happened.

1. The first session did not complete in a way we can stand behind, and here is the concrete reason (drop at two minutes, no candidate audio, Incomplete after a confirmed outage).
2. We kept that session on the record and marked it superseded.
3. The verdict we are using is the later completed conversation, and here is the tier.

If you cannot say sentence 1 without squirming, that row is not a supersede. It is a redo. Stop and put the human decision on the original row.

There is a cousin of this mess that is not a reset at all. Someone creates a second job description with the same title, invites the same student, and calls it a different campaign. The inbox then shows two verdicts for what the business thinks is one role. Labels and archive-instead-of-delete exist so you can tell jobs apart and so old interviews are not orphaned. They are not a back door for a second bite. If two labels are really one requisition, you have broken the job side of the rule, and the verdicts will not agree because they may not even share bands.

Regenerate Evaluation is another cousin. It re-scores a finished call from the transcript, keeps the earlier marks on file, and will not run mid-conversation or on a blank transcript. I use it when the screener itself was wrong and we need a before-and-after trail. I do not use it as a mood button. If you regenerate until the tier moves, you are doing the redo without the honesty of a second conversation.

The request that should never open a new HireQwik link

I keep a short list of sentences that sound reasonable and should still get a no.

• “They were nervous. Let them sit again tonight.” Nervousness can be real. It is a cue to listen, and sometimes a cue to move On Hold to a human Go. Discarding a completed scorecard is a different act.
• “The hiring manager did not watch the recording, but they know this student.” Then the manager owes the file a decision, not the student a second exam.
• “We need a Strong Go for the director review.” Manufacture that and the director review is theatre. Change the band on the job if the band is wrong. Do not shop candidates into it.
• “Their friend scored higher and the questions were the same.” Same questions are the point of a structured screen. Different answers are allowed to land in different tiers.
• “Just one more, this candidate only.” One-only policies do not stay one-only. Write the exception down or do not make it.

A HireQwik inbox tree before anyone clicks reset

Use this before anyone clicks reset. It takes less time than the argument.

On Hold rows that nobody opens are not a cue to re-invite. They are a cue to set a deadline, which is the point of giving every On Hold a decision date. A deadline creates a decision. A new link creates a second verdict.

Apply the tree on the next uncomfortable row, not on a policy slide. Open the interviews inbox, find one completed No Go that someone wants retried, and write the human note before you discuss a new link. If the note is “sounds nervous, please redo,” you are not done listening.

What HireQwik reviewers should tell the panel when two records exist

Sometimes you inherit the mess. A coordinator reset three students yesterday. You cannot unsend the evening.

Do this in order.

Name the binding record out loud. “We are using the completed call from Tuesday. The Monday session is superseded because the join failed at the start.” If you cannot give the because, say the harder sentence: “Monday was a full call. Tuesday should not have been booked. We will use Monday, and we will not book a third.”

Stop further resets on that job until the standup agrees the criterion. A criterion I will accept is narrow: Incomplete or a broken join, confirmed by the recording or the status, one retry. A criterion I will not accept is “manager request” or “tier below Go.”

Tell the student something true if they already sat twice. Skip the speech. You need to avoid implying that the company runs surprise retests until the score improves. Surprise retests are how a process starts to feel arbitrary, which is the opposite of why we structured the call.

Share the binding scorecard, not both, unless someone asks for the history. A shared scorecard should be the record you mean. Sending two links and saying “see what you think” exports the confusion to a person with even less context.

Then fix the screener if the redo revealed a bad question. The rewrite applies forward. It does not entitle the morning batch to a new exam.

When a placement officer asks for a reattempt, answer with the roll number and the concrete reason, then hold the line. A dropped join can earn a reset. A finished call that sounded nervous earns a human note, not a fresh exam.

The HireQwik line I want on the drive doc

Write this above the invite list:

One completed screening call per student per job. A reset is only when that session failed to finish. Disagreement with a finished verdict is a human decision on that verdict, not a new link.

If a stakeholder crosses it out, ask them which decision they want deleted from the file, and who will explain that deletion in August when someone asks. Most people put the line back.

If you want to pressure-test the line against a role you are about to open, send the edge case through the demo form: a drop, a knockout, or a manager who wants a redo. The useful outcome is a written rule for that role, not another sample conversation.