You post a job for a junior accountant. A few days later you have hundreds of CVs. Reading them all is slow, and tired readers skim. They start to favour famous colleges or familiar names. The AI can help with the first read. It only helps if you give it a fair scorecard first, and you check what it gives back.
The mistake most people make
Asking the AI "Which candidate is best?" with no scorecard. Or letting it send rejections that no person has read.
“A founder pasted 10 CVs into a chat and asked: "Pick the top 2 for our marketing lead role."”
“The AI picked the two candidates from well-known engineering colleges. It passed over a candidate who had run paid ads for a clothing brand in Surat for four years.”
How it works: the marking scheme
An examiner does not mark an answer sheet on a feeling. They mark it against a written scheme: this point earns two marks, that step earns one. Two examiners using the same scheme give similar marks. Your scorecard is the marking scheme. Write it before the AI reads a single CV.
Four steps to a fairer first read
1. Write the scorecard
List three or four skills the job truly needs. For example: Tally, GST filing, well-written emails in English. Leave out anything the job doesn't need.
2. Ask for evidence
For every score, the AI must quote the line in the CV that supports it. No quote means no score.
3. Ask what's missing
For each shortlisted candidate, ask the AI which skills on your scorecard the CV doesn't show. Those become your interview questions.
4. A person decides
The AI summarises and quotes. A person chooses who to interview, who to hire and who to turn down.
Rules for fair hiring with AI
I am hiring for this role: [Job title] Here is the scorecard I wrote. It lists only the skills this job needs: """ 1. [Skill the job needs, e.g. Tally and GST filing] 2. [Skill the job needs, e.g. handling supplier payments] 3. [Skill the job needs, e.g. clear written English] 4. [Skill the job needs, or delete this line] """ Here is the candidate's CV, with name, photo, age, gender and address removed: """ [Paste CV text here] """ Your task: 1. Score the candidate from 1 to 5 on each scorecard line. 2. For each score, quote the exact line from the CV that supports it. If there is no line, write "No evidence" and give no score. 3. List up to 2 skills from the scorecard that the CV doesn't show. Phrase each as an interview question. 4. Put the candidate in one group: Strong match, Borderline or Weak match. Ignore career gaps, age, gender, name, photo, college name and location, unless a scorecard line needs them. Do not recommend hiring or rejecting anyone. A person decides.
The AI can only score what it can quote. It ignores the things that often lead to unfair choices. You wrote the scorecard, and you make the decision.
Working File Templates
Copy or save these plain text files to use directly in your workspace.
A standardised 4-tier rubric for consistent evaluation across all applicants.
# Candidate Scoring Rubric: Commercial Operations Executive Department: Operations & Supply Chain · Level: Mid-Senior ## Scoring Scales (1 to 5) - 1: No evidence found in CV. - 2: Mentions the skill in passing, but quotes no metrics, outcomes, or scale. - 3: Demonstrates standard execution with clear tools and timelines. - 4: Demonstrates independent problem solving and measurable business impact. - 5: Built or restructured systems, led cross-department initiatives, or trained others. ## Core Competencies & Evidence Requirements ### 1. ERP & Inventory Reconciliation - Minimum standard (Score 3): 2+ years daily reconciliation of multi-warehouse physical stock against software ledgers. - Evidence indicator: Names specific ERP (e.g. SAP, Tally, Zoho) and batch cycle audit frequency. ### 2. Vendor Dispute Resolution - Minimum standard (Score 3): Managed commercial communication for supplier shortages, transit damages, or delayed deliveries. - Evidence indicator: Mentions credit notes, debit notes, or transporter dispute settlements. ### 3. Clear Commercial Communication - Minimum standard (Score 3): Drafts unambiguous email follow-ups and WhatsApp status updates without corporate filler. - Evidence indicator: Concise project descriptions and well-structured responsibility bullets.
Additional Prompts for This Lesson
Tailored prompts for specific workflows and departments.
Evaluates an anonymised CV strictly against competency requirements without making automated hiring decisions.
You are an executive hiring assistant helping evaluate candidate evidence. Here is the role scorecard: """ [Paste the 3-4 essential competencies with required evidence standards] """ Here is the candidate work history (anonymised): """ [Paste the work experience text here] """ Your task: 1. Create a 3-column table: | Competency | Quoted Evidence from CV | Preliminary Score (1-5) | 2. If no verifiable metric or direct responsibility is mentioned, score 1 and write "No verifiable evidence found". 3. Write three specific probing questions for the hiring manager to ask during the live interview to test the candidate's weakest area. 4. Conclude with a neutral 2-sentence summary of the candidate's demonstrated track record. Do not state whether to hire or reject.
Forces the evaluation to rest entirely on direct quoted facts rather than stylistic polish or credentials.
Why should the AI quote a line from the CV for every score?
Show the answer
So you can see why it gave the score. If there is no quote, the AI may be guessing, or favouring something like a famous college. A quote lets you check it in seconds.