JPMorgan Leads Corporate Adoption of AI-Powered Performance Reviews

JPMorgan Pioneers AI-Assisted Performance Reviews

JPMorgan Chase & Co. has become one of the first major financial institutions to officially permit managers to use artificial intelligence for writing employee performance reviews, marking a significant shift in corporate human resources practices 1

1

. The bank's internal guidelines allow supervisors to utilize an internal chatbot to help compose their evaluations, though the technology is explicitly described as "not a substitute for human judgment" 2

2

.

The policy comes with clear boundaries: managers are prohibited from using AI tools to assign performance scores or make critical decisions regarding pay or promotions 2

2

. This approach reflects the company's attempt to harness AI's efficiency while maintaining human oversight over consequential employment decisions.

The Promise of More Objective Evaluations

Proponents argue that AI-assisted reviews could address longstanding issues with traditional performance evaluations. Peter Cappelli, a management professor at the University of Pennsylvania's Wharton School, notes that human managers are "prone to bias and faulty memory" and often overweight recent events when reviewing longer periods 2

2

. AI systems can potentially provide more objective assessments by analyzing comprehensive data throughout the review period.

Benjamin Levick, who leads AI and operations at corporate card company Ramp, emphasizes the practical benefits for time-constrained managers. He suggests that failing to use available AI tools might actually disservice employees if it means managers can only review a small fraction of their work due to time limitations 2

2

.

Risks of Dehumanization and Generic Feedback

Despite potential benefits, experts warn of significant risks associated with AI-powered performance reviews. The primary concern centers on the potential for "AI workslop" – generic, impersonal feedback that fails to provide meaningful guidance to employees 1

1

. Employees may perceive AI-assisted reviews as less credible, potentially thinking, "It said I did a good job. Does the boss really think that?" according to Cappelli 2

2

.

Research indicates that when people have access to AI tools, they often "offload analysis to the technology and rely less on their own judgment" 2

2

. This dependency could result in employees feeling "processed rather than personally evaluated," potentially deepening existing cynicism toward performance review processes.

Emerging AI-First Evaluation Systems

Beyond assisting human managers, some companies are developing AI systems designed to take primary roles in employee evaluation. Payroll platform Rippling has introduced "Talent Signal," a system that automatically assesses new employees' first 90 days, categorizing workers as "high potential," "typical," or "pay attention" based on their task performance and work complexity 2

2

. Managers can approve or reject these AI-generated assessments with simple thumbs up or down responses.

Meanwhile, established companies like consulting firm KPMG and e-commerce platform Shopify have begun incorporating AI usage itself as evaluation criteria, assessing how effectively employees utilize artificial intelligence tools in their work 2

2

.

Industry Caution and Best Practices

HR experts emphasize the importance of maintaining human involvement in the review process. Levick describes his approach as using AI to "summarize lots of data, such as peer feedback and work employees submitted throughout the year" while ensuring that "the review is fundamentally from me and includes my thoughts as the primary driver" 2

2

.

The technology's tendency toward sycophantic responses presents another challenge, as AI-assisted reviews may be "overly positive and fail to adequately address performance issues" 2

2

. This characteristic underscores the need for human oversight to ensure reviews maintain appropriate balance and address areas for improvement.🟡 communicated by: 🟡Okay, I've reviewed the summary and the image placement rules. Since no specific images were provided, I've used placeholder image IDs for what would be appropriate visuals based on the content.

Here's the summary with the selected images placed: