Two colleagues discussing handwritten notes beside a laptop.
Illustrative image · AI-generated

A draft kit, not a scoring engine

The distinction worth holding onto: drafting an interview kit is writing the questions and the rating guidance in advance. Automated candidate scoring is a machine reading applicant material and producing a number or a rank. This workflow does the first and never the second.

Concretely, no resume, application, transcript or note about a named person is pasted into the prompt. The AI never sees a candidate. It sees what the job involves, and it proposes wording. Interviewers rate what they heard, and a person weighs those ratings.

  • Input is job information only, never candidate material.
  • Output is criteria, questions and evidence anchors, not scores or shortlists.
  • Every line is a proposal to be edited, including the ones that read well.
Reference

Start from a source pack with IDs

A source pack is the small set of documents you are willing to stand behind: the duty list, the team's own notes on what the role does day to day, anything already written down and agreed. Each one gets an ID so the draft can point at where a criterion came from.

Our worked example uses a fictional pack, R1 v1, describing a support coordinator at an invented company called Northstar. It lists three duties: clarify customer requests, document a handoff with an owner and a next step, and prioritize according to an approved policy. Instruct the AI to cite those IDs, then verify that it has not added a degree requirement, a number of years of experience, a personality profile or a culture fit criterion. Move unsupported requirements into a To confirm list. A prompt does not guarantee that the output follows these rules.

  • Give every document an ID and a version, so the draft can cite it.
  • Anything the AI cannot trace to an ID belongs in the gap column, not in the criteria.
  • No invented qualifications, experience thresholds, personality traits or culture fit language.
Reference

Same questions for everyone, one shared scale

The structure this workflow borrows is the plain one used in structured interviewing: ask every candidate for a role the same predetermined questions, and rate the answers against a common scale agreed in advance.

That is easier to promise than to keep, which is why the kit exists as an artifact. When the questions and the scale are written down before the first interview, the follow up conversation is about what candidates said rather than about which interviewer asked the harder question.

  • Questions are fixed before the first interview, not improvised per candidate.
  • One rating scale, defined in words, used by every interviewer on the panel.
  • Agree any permitted follow-up probes and interviewer instructions before use.
Reference

Evidence anchors, and why Not observed is its own option

For each criterion the draft proposes an evidence anchor: a short description of what an interviewer would actually hear in a useful answer. These anchors are editorial proposals that need calibration against how your team really works. They are not findings, and they are not lifted from any source.

Record Not observed when there is insufficient evidence to assess a criterion, including when it was not covered. Keep that separate from evidence that meets a defined low anchor. Do not convert missing evidence into zero. This template has no composite score or automatic decision threshold.

  • Not observed is recorded as its own value, never converted to zero.
  • No weighting, no total, no automatic pass or fail line.
  • Anchors describe observable answer content, not impressions of the person.
Reference

Manager review, then hand the kit to your ATS

Approval is one named person reading the whole draft: the criteria, the question wording, the rating scale, the interviewer instructions and the gap column. Items marked To confirm stay unusable until the missing policy or definition is supplied. That review is the step that makes the kit yours.

After approval, the kit belongs in the system your interviewers already open. In Workable, interview kits attach the questions and the scorecard to a stage so every interviewer sees the same thing and leaves feedback in the same place, and kits are available on all plans. In Breezy, custom scorecards sit on Growth, Business and Pro, and changing a scorecard's settings removes feedback already left in custom sections, so build a new draft outside any active kit, preview it there, and confirm what is preserved before editing something a live role depends on.

  • One named approver, and a date, before the kit is used.
  • Anything marked To confirm is not rated until it is resolved.
  • In Breezy, preview a new draft outside the active kit and check preservation before editing.
  • The hiring decision stays with the manager, informed by the notes, not produced by the form.
Reference

A practical starting point

From role brief to a reviewed interview kit.

Fictional source R1 v1: Northstar’s support coordinator clarifies customer requests, documents handoffs with an owner and next step, and prioritizes according to an approved policy. That policy has not been supplied.

The questions and evidence below are editorial proposals for calibration. They are not a validated assessment or a record of an AI product test.

Clarifies a customer request before acting on it

Source: R1 v1 (fictional Northstar support coordinator duties)

Draft question: Tell me about a request that arrived without enough information to act on. What did you do before you started working on it?

Evidence to discuss: Names the specific detail that was missing, describes how they went back for it, and says what changed once they had the answer.

Before use: The pack does not say which channels requests arrive on. Confirm the wording matches how inbound requests actually reach your team.

Documents a handoff with a named owner and a next step

Source: R1 v1 (fictional Northstar support coordinator duties)

Draft question: Walk me through the last time you handed something to a colleague. What did you write down, and where did you put it?

Evidence to discuss: Identifies the receiving owner and next step, and explains where the handoff was recorded. Confirm any additional expectations before adding them to a rubric.

Before use: The pack does not identify a system of record for handoff notes. Confirm the tool before interviewers judge what good documentation looks like here.

Prioritizes according to the approved policy

Source: R1 v1 (fictional Northstar support coordinator duties)

Draft question: Describe a time two requests needed your attention at once. How did you decide which one went first?

Evidence to discuss: To confirm after the approved priority policy is supplied. The question is a draft only; no scoring anchors are set for this criterion.

Before use: Rubric: To confirm. R1 v1 states that prioritization follows an approved policy but does not include it. Do not rate this criterion until the policy defines the levels, and do not let the draft invent urgency definitions.

Download the prompt & fictional source pack Download the blank scorecard template

Voice of the customer

Bring the hiring conversation together.

Explore individual customer perspectives on shared candidate records and structured interview feedback, then turn each experience into a question for your own team.

Your recruiting questions, answered.

How to create an interview scorecard?

Start from the duties of the job as written down, turn three to five of them into criteria, write one question per criterion, describe what a useful answer contains, agree a rating scale in words, and add a Not observed option. Then have the hiring manager review and approve the whole thing before the first interview. AI can draft each of those pieces, but the approval step is the one that makes it usable.

Reference
Does the AI score or rank candidates?

No. It drafts the questions and the rating guidance before anyone is interviewed. Interviewers do the rating, a manager reads the results, and no number produced here decides anything.

Reference
Can I paste a resume or interview notes into the prompt?

Not in this workflow. The input is job information only. Once candidate material enters the prompt you are doing something different, with different stakes, and this is not the process for it.

Reference
What exactly is a source pack?

The set of documents you are willing to stand behind about the role, each given an ID and a version. Our example pack, R1 v1, is fictional and describes a support coordinator at an invented company.

Reference
What happens when a duty has no rubric, like prioritization?

It gets marked To confirm and stays unrated. In the example, the source pack says prioritization follows an approved policy but does not include the policy, so no urgency definitions are written in. Someone supplies the real one first.

Reference
Why is Not observed separate from a low rating?

Not observed records insufficient evidence to assess a criterion. A low rating requires evidence matching a defined low anchor. Keep those separate and record the reason for any missing evidence.

Reference
Do I add the ratings up?

This workflow does not produce a composite score, a weighting or a cut off. The ratings are structured notes that make a conversation easier to have, not a formula that returns an answer.

Reference
Who approves the kit before it is used?

The manager who owns the role. They confirm the criteria, the question wording, the rating scale, the interviewer instructions and everything sitting in the gap column. Without that pass, the draft stays a draft.

Reference
Should every candidate really get the same questions?

Use the same predetermined core questions and rating guidance for the role. Agree interviewer instructions and any permitted probes before use.

Reference
Where does the approved kit live afterwards?

In the ATS your interviewers already use. Workable's interview kits keep the questions and the scorecard attached to a stage so feedback lands in one place, and kits are available on all plans.

Reference
Anything to watch when editing a Breezy scorecard?

Yes. Custom scorecards are on Growth, Business and Pro, and changing a scorecard's settings removes feedback already left in custom sections. Build the new version as a separate draft outside any active kit, preview it, and check what is preserved before touching a live one.

Reference
How should we maintain the scorecard after a role changes?

Check the source version, review affected criteria and questions, and approve a revised draft before use. Confirm how your ATS preserves existing feedback before editing an active kit.

Reference

Keep the process connected.

Compare Breezy HR and Workable, choose your ATS starting point, then plan the onboarding handoff.

Breezy HR reviewWorkable review

Sources & method

Official references checked September 20, 2026. Product boundaries come from vendor documentation; suitability and the proposed workflow are editorial judgments. Confirm current scope with the vendor.

One FAQ topic was selected from Google US People Also Ask results collected through DataForSEO on September 20, 2026; the remaining eleven are editorial. Candidate coaching and unrelated questions were excluded.

How we compare tools