Gen AI Transforms Performance Reviews, Risks Persist
Harvard Business Review reports that enterprises are deploying generative AI to draft and polish performance reviews, with some systems reportedly cutting review-writing time by 40%. The HBR article warns that polishing narratives can make evaluations feel more consistent than they are and may mask longstanding inconsistencies and blind spots. HBR outlines an alternative use case: using generative AI to surface direct evidence of work-what people actually did, decided, and influenced-rather than producing better-written narratives. The article describes persistent deficiencies in traditional reviews, including selective memory and inconsistent manager descriptions, and argues that current AI usage often amplifies existing limitations instead of addressing them. HBR recommends reorienting implementations toward evidence gathering and structured signals, rather than only automating prose.
What happened
According to an article in Harvard Business Review published May 15, 2026, enterprises are increasingly deploying generative AI to streamline performance reviews. The HBR piece reports that some internal systems and vendor tools draft evaluations, support year-end review workflows, and reportedly reduce managers' review-writing time by 40%. The article states that, in practice, many organizations use AI primarily to produce polished narrative reviews more quickly rather than to change what gets evaluated.
Technical details
The HBR article documents two technical and process gaps in current review practice: inconsistent manager narratives shaped by selective memory and storytelling, and a reliance on easily measured outputs that miss higher-order contributions such as mentorship, strategic insight, and conflict resolution. HBR argues that generative models can smooth and standardize prose, which increases perceived consistency but may obscure underlying inconsistency in evidence.
Editorial analysis
Companies applying generative AI as a writing assistant commonly see short-term productivity gains while increasing opacity around provenance and evidence. For comparable document-automation use cases, observers note trade-offs between polish and auditability, and a corresponding need for provenance, metadata, and traceable source signals.
Context and significance
For practitioners, the HBR article reframes value from rhetorical improvement to evidence capture. Shifting performance systems toward structured event logs, decision traces, and outcome-linked artifacts would align tooling with what the article identifies as hard-to-measure but high-value contributions. This shift raises engineering and privacy questions around data sources, annotation, and consent.
What to watch
Indicators that organizations are moving beyond polishing include: integration of systems that surface decision timelines and communications, adoption of provenance metadata for review inputs, and tooling that links reviews to measurable team outcomes. Absent those signals, HBR warns that generative AI deployments may entrench the same evaluation blind spots while making them look cleaner.
Source attribution
All reported facts and figures above are drawn from the Harvard Business Review article "Gen AI Could Fix Performance Reviews-or Make Them Even Worse," May 15, 2026.
Scoring Rationale
The story matters to ML practitioners and HR technologists because it highlights common trade-offs when applying generative models to people processes: productivity gains versus auditability and evidence quality. The coverage is notable but not frontier-level model news.
Practice interview problems based on real data
1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
