How generative AI helped Nippon Life elevate stewardship

For more than a decade, Nippon Life, one of Japan’s largest life insurers, has been a leading steward in the country’s corporate governance reform, engaging roughly 700 investee companies annually in the effort. As the scale of stewardship activity grew, the company needed a clearer way to understand whether its engagement efforts were effective and where progress was being made.

The challenge — Measuring stewardship effectively at scale

To better understand the effectiveness of its stewardship activity, Nippon Life needed a way to evaluate more than a decade of investee company dialogue in a consistent, objective manner. The volume of records, the qualitative nature of the conversations, and the wide range of themes made manual assessment difficult. The company needed an approach that could analyze dialogue quality at scale and surface patterns that were not visible through anecdotal review alone.

Nippon Life embedded stewardship as a core responsibility following the introduction of Japan’s Stewardship Code for Institutional Investors (2014) and the Corporate Governance Code (2015), which accelerated governance reform. Over time, discussions with investee companies expanded from governance structures and financial performance to environmental and social themes, reflecting broader shifts in expectations around sustainable value creation.

This expansion increased the diversity and complexity of the conversations. Each stewardship officer was engaging with close to 100 companies each year, often across multiple themes and time horizons. Nippon Life needed a way to understand how these dialogues were progressing and where further focus was required — based on consistent evidence rather than anecdotal impression.

The solution — Applying generative AI to assess dialogue quality

Addressing this challenge, we partnered with the insurer to develop an AI-enabled framework capable of evaluating dialogue quality across a decade of engagement records and linking those interactions to observable corporate outcomes.

The collaboration focused on turning a large body of unstructured information into evidence that could guide future stewardship priorities. We began by training AI to understand Nippon Life’s investment philosophy, and then defined what constitutes a high-quality dialogue — something constructive, specific, and capable of influencing behavior. A pilot covering 300 records was used to refine prompts so the model could distinguish between different levels of dialogue quality without over-interpreting the content. The final prompt framework captured the detailed criteria needed for consistent assessment.

Once calibrated, the AI analyzed a total of 8,830 dialogue minutes. Human-led validation and governance checks were incorporated to ensure credibility. This combined approach enabled consistent measurement of dialogue quality over time, themes, and companies, revealing patterns that had previously been hidden in unstructured records.

Key metrics 

Measuring dialogue quality was only half the challenge. The next step was to determine whether Nippon Life’s governance requests were followed by any observable change within their investee companies. To do this, we classified more than 10,000 requests as quantitative (financial metrics) or qualitative (disclosures, governance changes) and pre-loaded relevant financial data and public reports into the system environment. The AI cross-referenced these requests with subsequent changes in disclosures and financials to assess directional alignment.

Nippon Life approached this step with healthy skepticism, recognizing the complexity of linking dialogue to corporate outcomes. But through a combination of system design and human-led validation, the assessment produced results that were consistent, repeatable, and credible.

The impact — Evidence-based stewardship outcomes and measurable improvements

The analysis produced clear, actionable insights. Year over year, the proportion of dialogue rated High Quality increased, confirming progress in the effectiveness of Nippon Life’s engagement activity. A pattern also emerged: financial and governance topics more frequently achieved High, Good, or Effective ratings, while environmental and social dialogues showed more mixed results. Crucially, when dialogues were effective and companies responded positively, the likelihood of observable behavioral change was higher.

The combined approach enabled Nippon Life to understand where progress was being made, where further focus was needed, and how dialogue quality related to subsequent corporate outcomes. For a long-term institutional investor, stewardship is not about quick wins — improvements often take three to five years or more to materialize. With clearer evidence on what works and why, Nippon Life is now better equipped to support investee companies and drive sustainable corporate value over the long term.