The Essence of “Organizational Redesign” Revealed by the AI-ification of Performance Reviews

The Dawn of an Era Where AI Manages Performance Review “Operations”

Performance reviews are considered “critical periodic tasks” in many companies, yet the process is prone to becoming a hollow formality and overly dependent on individuals. Issues like inconsistent evaluation criteria, reviewer burden, and uneven feedback quality have long remained unaddressed.

A new service has been announced that challenges this status quo: “AI人事評価ジョブオペ(R),” which utilizes generative AI to make performance reviews “operational.” The core of this news is not that “AI will ‘conduct’ the evaluations.” It lies in the AI’s ability to “structure the evaluation process into an operational state and sustain it.”

This transcends the category of a mere efficiency tool. It can be seen as a symbolic case indicating that AI is beginning to intervene in “organizational design” itself, taking on a role that ensures its reproducibility and sustainability.

AI’s Inherent Strength: “Standardizing Evaluation Criteria”

What is the biggest hurdle in implementing traditional performance review systems? It is the difficulty of customizing a uniform evaluation flow to fit “your company’s context” and maintaining its operation. Many SaaS-based performance review tools provide excellent “frameworks,” but the content—defining specific criteria, writing evaluation comments, ensuring feedback quality—was ultimately left to humans.

The groundbreaking aspect of this AI performance review service is the deep involvement of AI in generating and supporting the operation of this “content.” Specifically, it is expected to support processes such as the following.

Specifying Evaluation Items and Generating Text

The task of translating abstract evaluation items like “leadership” or “communication skills” into specific descriptions tied to actual work behaviors (e.g., “Identified the risk of project delay early and conducted interviews with relevant parties”) places a significant burden on reviewers. Generative AI can link the competency requirements defined by the company with an employee’s actual performance data (e.g., from activity reports or project management tools) and propose suitable specific examples or text for the evaluation.

Bias Mitigation and Consistency Assurance

Human evaluations are inevitably subject to biases like the recency effect (stronger impressions from recent events) or similarity bias (rating subordinates similar to oneself higher). By generating evaluation text based on data from the entire review period and applying consistent criteria, AI is expected to help mitigate these unconscious biases. It can function as a de facto “standard gauge” for maintaining a consistent “measuring stick” for evaluations across the company.

The Management Challenge Shifts from “Efficiency” to “Redesign”

Let’s reframe this trend within the context of what we advocate: “Moving Beyond SaaS Dependence.” Traditional performance review SaaS operated on a business model of renting out the “framework”—the “box.” Companies paid high license fees while still bearing the crucial operational burden and personalization risk internally.

The AI-based approach changes this structure. It shows a path to embedding the company’s core “evaluation philosophy” or “desired employee profile” into an AI engine, semi-automating the operation itself. This is closer to the idea of internalizing your company’s core elements as “AI-driven mechanisms” rather than borrowing a SaaS “finished product.”

In fact, even at the company of this media’s operator (the author), we have begun experimenting with delegating some aspects of evaluation and goal management (OKR) progress checks to an AI agent integrated with Slack. The AI collects and summarizes weekly progress reports, analyzes deviations from goals, and alerts managers as needed. Such “automation of operations” cannot be achieved by merely buying an evaluation “framework.”

The First Step to Practice: Starting “AI-ification of Reviews” In-House

Before suddenly introducing a dedicated service, how far can you go with your own resources? Here are concrete, realistic first steps that executives and back-office managers should consider.

Step 1: “Verbalizing” and “Structuring” Evaluation Criteria

First, organize your current evaluation system into a form AI can understand. Instead of inputting the evaluation sheet directly into AI, compile “evaluation items,” “definitions,” “specific expected behaviors (examples),” and “rating scales (definitions for 1-5)” into a table format. This exercise itself helps uncover ambiguities in your criteria. The cost is only the man-hours required for this organization work.

Step 2: Using Generative AI for “Draft Generation” of Evaluation Text

Utilize general-purpose generative AI like ChatGPT or Claude. The reviewer (manager) inputs the subordinate’s specific achievements or episodes (e.g., “Handled a sudden client specification change in Project A and met the deadline”) as simple bullet points. Provide the AI with the pre-organized evaluation criteria definitions as a prompt, and have it generate a draft like, “This episode corresponds to Level 4 in ‘Problem-Solving Ability.’ The following evaluation text is suggested.” The reviewer can then add more personalized feedback based on this. Monthly costs can start from around a few thousand yen per user, even using high-performance models like GPT-4.

Step 3: AI Checks to Ensure Feedback “Quality”

As a final step, you can also use AI to analyze the feedback text written by the reviewer. Have the AI check against pre-set guidelines: “Is this feedback based on specific actions?” “Does it include suggestions for improvement?” “Is there an excessive use of negative expressions?” This helps raise and level the overall quality of feedback.

Balancing Subsidy Utilization and In-House Development

As related news, it’s also noteworthy that an AI platform has been certified as an eligible tool for the “Digitalization and AI Introduction Subsidy.” This signifies the government’s support for introducing the AI “foundation” itself.

A crucial decision for executives is choosing between “introducing an external service using the subsidy” or “strengthening the company’s own AI foundation (e.g., data integration platform) with the subsidy and aiming to internalize the evaluation system on top of it.” The latter path may seem high-barrier at first glance, but with the advancement of code-generating AI (Claude Code, GitHub Copilot, etc.), it has become a much more realistic option than before.

Not limited to performance reviews, all business processes—management, sales support, customer service—are loops of “design → operation → improvement.” The true value of AI lies in reducing the burden of the “operation” part of this loop, creating an environment where humans can focus on more creative and strategic tasks: “design” and “improvement.”

Conclusion: AI Elevates HR from “Management” to “Design”

The emergence of AI performance review services vividly demonstrates that AI adoption is shifting phases from mere “task automation” to “redesigning the business process itself.”

There may be cautious views about AI entering the domain of performance reviews, which is fundamental to corporate culture. However, what AI takes on is precisely the part humans are not good at and find burdensome: “fair and continuous operation based on defined criteria.” The “philosophy” of evaluation and the final “judgment” remain entrusted to human hands.

What executives and CTOs should consider now is not “which AI evaluation tool to choose,” but a more fundamental question: “How can we leverage AI’s power to redesign our company’s HR system and organizational management, making it more robust and sustainable?” The first step begins with small-scale practice: verbalizing your company’s evaluation criteria and experimenting with existing generative AI.