Recruiter Interview Scorecard Spreadsheets: Use case and AI Confidentiality

After a round of interviews the consultant gathers the panel’s scores into one interview scorecard spreadsheet: criteria and weights down the side, candidates across the top, comments beside the scores and a weighted total. Consultants upload the sheet to ChatGPT or Claude to rank the field, write the debrief for the client or draft feedback for an unsuccessful candidate. The sheet holds three people’s scores, a client manager’s candid comments, a remark about seniority in a young team and a row of agency notes that should never have been in the file. This page marks each item that the Privacy Act and discrimination law touch, and shows how the same sheet is worked on Ermos inside the agency.

The interview scorecard spreadsheet

Built by the consultant the day after first-round interviews from the panel’s score sheets, sent to the client in a cleaned version and kept for the debrief and the feedback calls. The working copy with the agency notes is the one that tends to get uploaded.

fx
ABCDEF
1CriterionWeightPanel comment
2All three strong.
3
4
5
6GL: Callum knows our ERP.
7
8
9Weighted score100%Out of 5. Panel: ME and GL, Tue 27 Oct.
10Panel decision
11TH notes (internal)
Round 1 scores

An invented spreadsheet. The candidates, the panel and the client are fictional. ME is Marcus Ellery and GL is Gareth Lowe of the client; TH is the consultant. Hover or tap any highlighted item for the analysis; the full list is below.

Interview scorecard builder

An Agent+ that builds the interview scorecard spreadsheet from the panel's score sheets, using your agency's layout and criteria. It keeps agency notes on a separate working sheet and flags comments, such as remarks about age or personal circumstances, for the consultant to review before the client copy is prepared.

What it does

  1. Reads the panel's score sheets and the criteria
  2. Builds the scorecard in your agency layout
  3. Splits the client copy from the working notes

What stays with your people. The consultant checks the scores and comments and sends the client copy. The agent builds the sheet; it does not rate candidates.

An Agent+ is built once and anyone in the practice can run it. It works from your own documents and runs on Ermos Dominion, the AI unit in your office, or on Ermos Edge, your own instance in an Australian data centre.

What leaves the agency

  1. 1
    Three named candidates

    Three people who interviewed in confidence, each still employed elsewhere. Personal information held by the agency, and the key that ties every score and note in the sheet to a real person. Privacy Act guide.

  2. 2
    Scores, weighted totals and decisions

    Seven scores per person, a weighted total and the panel’s decision. The Privacy Act defines personal information as information or an opinion about an identified individual, whether true or not, so scores are personal information. APP 12 requires access on request unless an exception applies. The totals check: 5 × 20% + 5 × 15% + 4 × 15% + 4 × 20% + 4 × 10% + 5 × 10% + 4 × 10% is 4.45 for the first column, and the others come to 4.00 and 3.50. Privacy Act guide.

  3. 3
    The criteria and weights

    Seven headings and their weights, which add to 100 per cent. A scoring template with no person in it, and the part of the sheet that can be shared or improved anywhere.

  4. 4
    A manager’s comment on an answer

    The client’s Head of Finance Operations on one candidate’s interview. An opinion about her, given to the agency in confidence, and the source of the feedback the agency later gives her. Privacy Act guide.

  5. 5
    A question about a candidate’s home life

    A panel member wondering what lies behind a candidate’s questions about travel, and asking the consultant to find out. Family or carer’s responsibilities and pregnancy are attributes in s 351(1) of the Fair Work Act, which applies to a prospective employer that refuses to employ or offers worse terms (s 342(1), item 2). Under s 361, if a complainant alleges action was taken for a prohibited reason, that reason is presumed unless the person who took the action proves otherwise. Privacy Act guide.

  6. 6
    Remarks about age and seniority

    A comment that a candidate is very senior for a young team, beside a note of her age. Section 49ZYB of the Anti-Discrimination Act 1977 (NSW) covers the arrangements an employer makes for deciding who is offered employment, and s 49ZYH covers employment agencies. Section 550 of the Fair Work Act treats a person knowingly concerned in a contravention of a civil remedy provision, such as s 351, as having contravened it, and s 56 of the Age Discrimination Act 2004 treats a person who causes, instructs, induces, aids or permits an unlawful act as having done it. Privacy Act guide.

  7. 7
    The weighted total formula

    A SUMPRODUCT of the weights and one candidate’s scores. A formula with no personal data in it. A tool can check it safely on its own; the sheet it sits in is another matter.

  8. 8
    A pregnancy, told in confidence

    A due date and the consultant’s promise to keep it from the client. Information about a pregnancy is likely to be health information, which is sensitive information under s 6 of the Privacy Act, and pregnancy is an attribute in s 351(1) of the Fair Work Act. The candidate told one person; an upload tells an AI company and anyone the sheet is later shared with. Patient data guide.

  9. 9
    A conviction and a police check

    A conviction the candidate says is spent, and the client’s condition. Criminal record is sensitive information under s 6 of the Privacy Act, and APP 3.3 governs collecting it. What, if anything, the client should hear about it is a decision for the agency and the candidate. Privacy Act guide.

  10. 10
    Other tabs

    The panel’s own score sheets and the round 2 plan. A spreadsheet file generally carries every tab, hidden columns, comments and properties such as the author, so an upload sends what is off screen as well.

What happens in Frontier Models (ChatGPT, Claude)

  • The whole document is transmitted to the provider’s servers. On personal plans the user cannot choose where it is stored or processed, and for the major providers that has generally meant outside Australia, under that country’s legal process. Some business plans offer storage in Australia, but the provider still receives the content.
  • On consumer tiers, prompts can be retained and used to train future models.
  • The provider’s staff and systems can reach the content under the provider’s own policies.
  • Once sent, the agency no longer controls it. Retention and deletion happen on the provider’s terms.

The same document on Ermos

Ermos Dominion

The document is drafted on a unit in your office, from your own templates and the candidate file. Every marked item stays on the unit, which keeps working with the internet switched off. Nothing in it is transmitted to an AI provider, so there is no disclosure to analyse: the confidentiality work is the same as for your file server.

Ermos Edge

Your documents, the index and your access boundaries stay on your own instance in an Australian data centre. When a staff member asks for this draft, only their question and the excerpts they are permitted to see go to a named Australian provider, who is contractually barred from keeping the content or training any model on it. Nothing is processed outside Australia.

Frontier Models (ChatGPT, Claude)Ermos EdgeErmos Dominion
Where the document is processedProvider servers, generally offshore on personal plansYour instance in an Australian data centre; answers generated in AustraliaThe unit in your office
Who can read itProvider staff and systems, under policyYour staff, within the boundaries you set. The named Australian provider processes only permitted excerpts, under contract, and keeps nothingYour staff only
Is it keptOn the provider’s termsKept on your instance; the provider is barred from retaining itStays on the unit
Can it train a modelOn consumer tiers, yesNo, by contractOnly your own unit learns from it, on your hardware

Part of Use cases in recruitment agencies.

Common questions

Do ChatGPT and Claude use the information in an interview scorecard spreadsheet?
Yes, everything pasted or uploaded is sent to the AI company's servers and processed there to produce a reply. As at October 2026, whether a conversation on a personal plan (ChatGPT Free, Plus or Pro; Claude Free, Pro or Max) can be used to train the models depends on a setting each user controls. Business plans (ChatGPT Business and Enterprise, Claude Team and Enterprise) and the developer APIs sit under commercial terms, and OpenAI and Anthropic each say they do not train on that content by default (OpenAI, Anthropic). On either product, the thumbs up and thumbs down buttons can send a whole conversation back for training. An uploaded scorecard carries three candidates' scores, their weighted totals and the panel's decision to decline Jessica Whitlam. It also holds the consultant's internal notes row, with Nina Castellano's pregnancy told in confidence and a conviction Callum Frasier says is spent.
Can we use ChatGPT or Claude for an interview scorecard spreadsheet if we remove the candidates' names?
No, because deleting the header row still leaves Nina, Callum, Jessica and Marcus named in the comment cells. Removing names rarely makes a document anonymous. Information stays personal information while the person is reasonably identifiable, and the OAIC warns that information can be at risk of re-identification even when it has been de-identified or anonymised. The notes row adds a candidate aged 52 who would step down in title, a baby due in June and a 2012 conviction. Each is enough to pick out one of three people interviewed for one role. Information about a pregnancy is likely to be health information, and health information and criminal record are both sensitive information under section 6 of the Privacy Act. The Office of the Australian Information Commissioner (OAIC) recommends in its guidance on commercially available AI products that organisations do not enter personal information, and particularly sensitive information, into publicly available generative AI tools, because of the significant and complex privacy risks.
Can ChatGPT or Claude be trusted to rank the candidates in an interview scorecard spreadsheet?
Not without the consultant rechecking the sums. Language models can produce fluent text with details not in the source, wrong figures and references that look real but are not. Here a mis-copied score or misread weight would do most harm, because the totals of 4.45, 4.00 and 3.50 set the order. An upload also hands the model the pregnancy and age in the notes row. From 10 December 2026, APP 1.7 to 1.9 require a privacy policy to describe the kinds of personal information and decisions involved where a computer program uses personal information about a person to make, or substantially and directly assist, a decision that could significantly affect them. A ranking that decides who reaches a client may be such a step.
How can a recruitment agency use AI for an interview scorecard spreadsheet without sending it to ChatGPT or Claude?
Ermos is built so the document stays under your control. On the Dominion AI unit it is drafted inside your office and none of it is sent to an AI provider. On Ermos Edge your documents and index stay on your own instance in an Australian data centre, and only the question and the excerpts the user is permitted to see go to a named Australian provider that is contractually barred from keeping them or training any model on them. The ranking, the client debrief and the feedback for each candidate are drafted from the panel's score sheets and the agency's own scorecard template. The consultant rechecks every weighted total and signs off the cleaned version before it reaches Northgate.
Can a recruitment agency be liable for a client's discriminatory hiring decision?
It can be. Section 550 of the Fair Work Act treats a person knowingly concerned in a contravention of a civil remedy provision, such as section 351, as having contravened it. Section 56 of the Age Discrimination Act 2004 and section 122 of the Disability Discrimination Act 1992 treat a person who causes, instructs, induces, aids or permits an unlawful act as having done it. Both federal Acts also bind employment agencies directly.
Can a candidate see her interview scores?
APP 12 requires an APP entity to give an individual access to the personal information it holds about them on request, unless an exception applies. The Privacy Act defines personal information to include an opinion about an identified individual, whether true or not, so scores and panel comments are covered. Write and keep them on the basis that the candidate may ask to read them.
Do we have to disclose that software helps score candidates?
From 10 December 2026, APP 1.7 requires an APP entity's privacy policy to describe the kinds of personal information used, and the kinds of decisions involved, where it has arranged for a computer program to make a decision, or do a thing substantially and directly related to making one, that could reasonably be expected to significantly affect an individual's rights or interests, using personal information about that individual. A ranking that decides who reaches a client may be such a step.

See Dominion answering from your own documents.

A demonstration takes 30 minutes, uses no client data, and comes with no obligation.

Book a demo