Actuaries may have to defend AI-generated decisions

Actuaries may soon be responsible for AI-generated decisions, even when the models and data come from third-party providers.

That’s one of the key implications of a Canadian Institute of Actuaries (CIA) review of its Standards of Practice, as the profession grapples with the growing use of artificial intelligence in actuarial modelling.

The proposed changes would place greater emphasis on transparency, explainability and accountability, while requiring actuaries to pay closer attention to the quality of the data feeding AI systems.

The review intends to align with the Office of the Superintendent of Financial Institutions (OSFI)’s growing focus on governance and oversight of higher-risk AI models, signalling a broader shift in how the insurance industry approaches AI-driven decision-making.

The CIA’s actuarial standards board is considering changes that would address data quality, model validation, governance, documentation, and professional accountability. It is also reviewing whether current standards adequately address an actuary’s responsibilities when relying on AI tools and outputs developed by third parties.

Also in the news: Why your commercial clients should err on the side of caution

That focus reflects growing concern throughout the financial services sector about how AI-driven decisions are made, whether those decisions can be explained, and who is accountable when things go wrong.

In its notice of intent, the CIA said it will consider alignment with several domestic and international frameworks, including OSFI’s Guideline E-23 on model risk management. The guideline takes a risk-based approach to model oversight, requiring stronger governance, monitoring and controls for higher-risk models. The CIA also plans to examine issues such as model transparency, explainability and interpretability as part of its review.

Among other things, the CIA said it’s considering data quality, representativeness, and potential bias in the information used to train and operate AI models. The review will also examine whether existing guidance adequately addresses reliance on AI tools and outputs developed by others, including third-party vendors.

In practical terms, actuaries may be expected to stand behind the output of an AI model even when the underlying technology, methodology, or data comes from an external provider.

That could place greater responsibility on actuaries to understand not only how a model works, but also the quality, reliability and potential limitations of the data feeding it.

The review also focuses on model validation, testing and governance. The CIA is assessing whether existing quality assurance requirements remain sufficient in an environment in which AI models can evolve over time, rely on massive datasets, and produce outcomes that may be more difficult to interpret than traditional actuarial models.

The emphasis on explainability mirrors concerns increasingly being raised throughout the insurance industry. Brokers, insurers and regulators have questioned how AI-driven decisions are reached and whether those decisions can be adequately explained to customers, regulators, and other stakeholders.

Brokers need to specialize. But where should they focus? Image

Insights Paid Content

Brokers need to specialize. But where should they focus?

Specialization is becoming table stakes, but there’s not a one-size-fits-all solution, advises specialized commercial insurer Sovereign Insurance.

By 

Sponsor Image

For insurers using AI in pricing, risk selection, reserving and predictive modelling, those questions are becoming harder to ignore.

The conversation is no longer centred on whether AI will be used in actuarial work. Instead, the focus is shifting toward transparency, governance and accountability — ensuring professionals can explain, validate and stand behind the outputs generated by increasingly complex AI systems.

For actuaries, that may mean greater responsibility not only for the models they build themselves, but also for the third-party tools and data on which they choose to rely.

Sonia Sache

Sonia is an award-winning multimedia journalist, producer TV personality and content strategist with more than a decade of experience across television, radio, digital media, and communications.