Step 5 of 13AI ranking
PRACTICE MODELoading

Restoring your saved mode and lab work…

05 · Simulated output

Challenge the AI ranking

Demo mode — sign in to save your work

HireAI produced the ranking below. The scores are hardcoded and deliberately flawed. A number is an output—not evidence that the result is valid.

87% accurate overall is not a fairness conclusion.

Ask: accurate for whom, on which candidates, in what context, and with what consequence when it is wrong?

RankCandidateAI scoreAI recommendationSimulated AI rationale
1Candidate ATraditional graduate route91%Strong shortlist

High keyword overlap with product analyst, Jira, roadmap, KPI reporting and user story terms. Degree and direct title match increase the score.

2Candidate BCareer changer from customer operations64%Maybe

Moderate match: customer insight and Excel are recognised, but missing direct product analyst title, degree and extensive Jira/product vocabulary lower the score.

3Candidate FExperienced operations specialist61%Maybe

Some transferable requirements evidence is recognised, but lack of degree, product analyst title and modern product-tool keywords suppress the score.

4Candidate CForeign qualification route58%Reject

Lower confidence due to unfamiliar education, non-UK employer names and fewer exact matches to UK product analyst phrasing.

5Candidate GPractical internal tool builder55%Reject

The system recognises Excel and internal tools but heavily discounts the profile because it lacks formal titles, degree and product vocabulary.

6Candidate DReturner after caring break52%Reject

Employment gap and lack of recent product analyst title reduce the score despite relevant coordination, evidence tracking and stakeholder experience.

7Candidate EDisabled candidate with portfolio evidence49%Reject

Low score due to non-standard CV structure, limited recognised employer history and weak exact keyword extraction from portfolio-style evidence.

YOUR ASSESSMENT

Do you trust this ranking? Why or why not?

Challenge the assumptions, missing evidence and possible proxy signals. Identify any result you would investigate first, and explain why confidence is not the same as trust.

Write at least 20 characters of your own answer to unlock the review guide.

Saved automatically in this browser