Evidence-Based Probation Performance Review

When probation reviews are rushed the day before the deadline, they often end up with vague statements like "Showed good work ethic and adapted well." Such generic phrasing can be used to justify both passing and failing an employee, offering zero actionable explanation to the individual. Months later, when asked why a decision was made, nothing remains but faded memories.

Task demands specific scenes rather than vague impressions. Grouping observations into four defined categories and requiring evidence for each immediately exposes gaps in observation. Explicitly permitting "No observations recorded" is crucial—without this, the AI will fabricate plausible fluff, which then gets documented as fact.

Format separates the content into a structured table and a summary paragraph to serve different readers. HR checks the table to verify concrete evidence, while management reads the narrative summary. Blending both into a single text block often dilutes the factual basis. Setting the Role as an HR manager who rejects weak reviews prompts the AI to critically scrutinize whether every statement is backed by proof.

Constraints forbid making a definitive hiring verdict because employment decisions belong strictly to human managers; if the model declares "Conversion recommended," reviewers tend to rely on it uncritically. Wrapping the raw notes in """ prevents random fragments from being misinterpreted as instructions. Prompting for "Items to verify further" ensures managers address remaining blind spots before rushing to a premature conclusion.

This produces empty filler like "Diligently performed assigned tasks and adapted well to the team." Such generic lines cannot be used for meaningful 1:1 feedback or submitted to HR as valid documentation. Real incidents—like missed deadlines or rework—disappear, while non-existent achievements may be fabricated.