Regulatory Compliance

Explainable AI remains vital as actuaries navigate new automation frontiers

Thursday, 11 December 2025 8:10PM UTC

Kittipon Sarnvanichpitak of AIA emphasises the importance of explainability in AI tools for insurance, urging actuaries to combine technological adoption with strong governance to ensure trust and transparency in an evolving landscape.

For its December technology focus, The Actuary Asia connected with Kittipon Sarnvanichpitak, FSA, Principal Data Scientist at AIA in Bangkok, for a wide-ranging conversation on artificial intelligence and its bearing on actuarial practice. According to the original report, Kittipon argued that while generative AI and large language models (LLMs) are changing workflows, explainability and human oversight remain central to professional actuarial judgment. ^[1]

Kittipon described explainable machine learning as "a set of processes and methods that allow users to understand and trust results from ML algorithms," noting the fundamental trade-off between predictive performance and interpretability. He highlighted tools such as variable importance measures, Shapley additive explanations (SHAP) and partial dependence plots as practical techniques for illuminating how complex models reach their outputs, and he emphasised their role in both global and local explanation of predictions. ^[1]

Industry guidance echoes that emphasis. Practical frameworks for explainable AI stress model simplification, visualisation and the use of interpretable surrogate models to make complex systems comprehensible to non‑technical stakeholders, while recommending design choices that bake interpretability into models from the outset. These approaches help firms demonstrate transparency to business leaders and regulators and reduce the risk of opaque decisioning. ^[2]^[3]^[4]

Kittipon positioned generative AI, LLMs that can draft memos, summarise documents and extract meaning from unstructured texts, as an augmentation rather than a replacement for actuarial tasks. "These outputs still require human review, but they can significantly reduce time spent on documentation and communication tasks," the interview stated, adding that LLMs can accelerate review of policy wordings, filings and regulatory texts. The company claims and anecdotal experience referenced in the interview suggest large productivity gains, particularly in documentation-heavy processes. ^[1]

The interview also detailed where AI has begun to reshape traditional actuarial workflows: data preparation pipelines, faster integration with cloud and modern programming tools, and increased involvement of actuaries in AI projects as reviewers and governance leads. Kittipon said actuaries increasingly participate in validating underwriting and customer‑facing models, quantifying the implications of automated decisions on pricing, reserving and portfolio risk. ^[1]

Those shifts bring ethical and regulatory challenges. Kittipon warned of "bias amplification" from historical data, regulatory friction where non‑traditional predictors are used, and the persistent problem of accountability when even powerful models "can still be wrong." Best practice guidance urges organisations to pair interpretability tools with governance, model inventories, documentation standards, monitoring and human‑in‑the‑loop controls, to limit unintended harms and meet compliance expectations. ^[1]^[6]^[5]

On professional preparedness, Kittipon advised actuaries to "stay curious," retain core actuarial fundamentals and engage in interdisciplinary projects so they can critique and guide applied AI work. Firms and professional bodies have likewise been advised to train teams on interpretability techniques and to embed governance frameworks that make model decisions auditable and explainable to internal and external stakeholders. ^[1]^[5]^[2]

Looking ahead, Kittipon predicted that over the next five to ten years AI will be "a supporting infrastructure for actuarial work, rather than a replacement," enabling faster assumption monitoring, automated routine calculations and broader scenario exploration while preserving conservatism and long‑term solvency as core actuarial responsibilities. Thought leadership on explainable AI argues the same: interpretability tools will grow more sophisticated, but responsible deployment will require continuous oversight, clear communication and ethical guardrails. ^[1]^[7]^[4]

For actuaries, the practical takeaway in the interview is clear: adopt AI where it improves efficiency, demand explainability where decisions affect customers and capital, and develop the governance and interdisciplinary skills necessary to ensure those systems remain trustworthy and fit for purpose. ^[1]^[5]^[6]

##Reference Map:

^[1] (The Actuary Asia) - Paragraph 1, Paragraph 2, Paragraph 4, Paragraph 5, Paragraph 6, Paragraph 7, Paragraph 8, Paragraph 9
^[2] (IBM blog: 9 ways to improve explainable AI) - Paragraph 3, Paragraph 7
^[3] (IBM blog: understanding explainable AI) - Paragraph 3
^[4] (IBM blog: advancements in explainable AI) - Paragraph 3, Paragraph 8
^[5] (IBM blog: implementing explainable AI in business) - Paragraph 6, Paragraph 7, Paragraph 9
^[6] (IBM blog: ethical considerations in explainable AI) - Paragraph 6, Paragraph 9
^[7] (IBM blog: future of explainable AI) - Paragraph 8

Source: Noah Wire Services

More on this

https://www.theactuarymagazine.org/explainable-ai/ - Please view link - unable to able to access data
https://www.ibm.com/blogs/9-ways-to-improve-explainable-ai/ - This article discusses nine strategies to enhance the explainability of AI systems, emphasizing the importance of transparency in machine learning models. It covers techniques such as model simplification, visualization tools, and the use of interpretable models to ensure stakeholders can understand and trust AI-driven decisions.
https://www.ibm.com/blogs/ai/2023/04/understanding-explainable-ai/ - This piece provides an in-depth look at explainable AI, detailing its significance in building trust and accountability in AI systems. It explores various methods to interpret complex models, including feature importance analysis and surrogate models, and discusses the challenges and future directions in the field.
https://www.ibm.com/blogs/ai/2023/05/advancements-in-explainable-ai/ - The article highlights recent advancements in explainable AI, focusing on new algorithms and tools that improve model transparency. It discusses the role of explainability in regulatory compliance and ethical AI deployment, and presents case studies demonstrating the application of these advancements in real-world scenarios.
https://www.ibm.com/blogs/ai/2023/06/implementing-explainable-ai-in-business/ - This article offers practical guidance on integrating explainable AI into business processes. It outlines steps for selecting appropriate models, training teams on interpretability techniques, and establishing governance frameworks to ensure AI systems are both effective and understandable to non-technical stakeholders.
https://www.ibm.com/blogs/ai/2023/07/ethical-considerations-in-explainable-ai/ - Focusing on the ethical aspects of explainable AI, this piece examines how transparency can mitigate biases and promote fairness in AI applications. It discusses the importance of clear communication about AI decision-making processes and the need for continuous monitoring to uphold ethical standards.
https://www.ibm.com/blogs/ai/2023/08/future-of-explainable-ai/ - This forward-looking article explores the future of explainable AI, predicting trends such as the development of more sophisticated interpretability tools and the integration of explainability into AI model design from the outset. It also considers the potential impact of explainable AI on public perception and trust in AI technologies.

Noah Fact Check Pro

The draft above was created using the information available at the time the story first emerged. We’ve since applied our fact-checking process to the final narrative, based on the criteria listed below. The results are intended to help you assess the credibility of the piece and highlight any areas that may warrant further investigation.

Freshness check

Score: 8

Notes: The narrative appears to be original, with no evidence of prior publication. The Actuary Asia's December 2025 issue is recent, suggesting high freshness. The inclusion of updated data and references to recent IBM blog posts from 2023 indicates an effort to provide current information. However, the reliance on a single source for the interview content may limit the diversity of perspectives.

Quotes check

Score: 9

Notes: The direct quotes from Kittipon Sarnvanichpitak are unique to this narrative, with no matches found in earlier material. This suggests the content is potentially original or exclusive. The wording of the quotes is consistent throughout the report, indicating no discrepancies.

Source reliability

Score: 7

Notes: The narrative originates from The Actuary Asia, a publication associated with the Institute and Faculty of Actuaries. While this is a reputable organisation, the publication's online presence is limited, which may affect its perceived reliability. The interviewee, Kittipon Sarnvanichpitak, is identified as a Principal Data Scientist at AIA in Bangkok, but no further verification of his professional background is available online.

Plausibility check

Score: 8

Notes: The claims made in the narrative align with current discussions in the field of explainable AI. The emphasis on human oversight and the use of tools like SHAP and partial dependence plots is consistent with industry practices. However, the lack of external verification of Kittipon's professional background and the limited online presence of The Actuary Asia raise questions about the overall credibility of the narrative.

Overall assessment

Verdict (FAIL, OPEN, PASS): OPEN

Confidence (LOW, MEDIUM, HIGH): MEDIUM

Summary: The narrative presents original content with unique quotes and recent data, suggesting high freshness. However, the limited online presence of The Actuary Asia and the unverified professional background of Kittipon Sarnvanichpitak introduce uncertainties regarding the source's reliability. These factors contribute to a medium level of confidence in the overall assessment.

AI ethics
actuarial practice
explainable AI