While artificial intelligence (AI) investments offer significant growth and profitability potential for insurance companies, they also bring new risks that, if not managed properly, could cause significant financial and have a negative impact on credit ratings, according to Morningstar DBRS analysts.
Insurers have long used machine learning, natural language processing and predictive analytics for underwriting models. Now AI-powered technologies have become more widespread and essential to maintaining competitiveness.
As a result, according to a survey by Wipro Limited (Wipro), North American insurers have increased the proportion of investment technology (IT) budgets allocated to AI technology to more than 20% in about three to five years, from 8% in 2024.
In a recent commentary, Morningstar DBRS provides an overview of key areas of the insurance value chain that stand to benefit from AI adoption, as well as the risks and challenges that this brings – which also present an important part of its credit risk analysis.
Analysts highlight that establishing robust governance risk frameworks is crucial for harnessing AI’s advantages while preserving the stability of credit ratings.
Across the value chain, AI provides potential benefits to insurers by enhancing operational efficiency, by taking on simple but high-volume repetitive tasks such as generating policy templates, summarizing customer interactions, and extracting key information from large datasets.
AI could also benefit the customer experience, by identifying customer behaviour, patterns, and preferences, helping to streamline sales and marketing efforts thereby reducing the overall customer acquisition cost; as well as providing support in core functions like underwriting and claims management.
Some insurers have deployed AI- powered chatbots and virtual assistants, which can benefit both employees and customers. These tools simplify the often-complex process of buying insurance by recommending policies based on individual preferences or similar customer profiles.
For life insurance and annuity providers, virtual assistants can even suggest options tailored to reported risk tolerance. However, challenges can arise when chatbots are used for processes that need to be highly customised, such as specific claims settlement interactions.
Despite these limitations, AI tools, when used appropriately, can significantly help insurers manage larger claim volumes by flagging complex cases for human intervention, ultimately improving overall efficiency, Morningstar DBRS analysts state.
Within the property and casualty insurance industry, AI can be used to assess damages to the vehicle/property using digital pictures and provide repair-cost estimates to the customer before the vehicle/property is even inspected.
AI can also be used for loss assessment during natural catastrophes. It can assess a company’s exposure to a specific catastrophic event in a relatively shorter timeframe.
Fraud is a major pain point for the insurance industry and is an area where AI can also provide value, given that the industry loses billions of dollars to fraud every year, analysts note.
AI models, trained on historical company data or third-party provider data, can identify suspicious claims. Flagging these claims for further review allows insurers to expedite payments for legitimate claims.
“However,” Morningstar DBRS warns, “we would also note that companies using AI assessments to reject claims could be exposed to legal and reputational risk if those AI models turn out to be unreliable.”
Adding: “By synthesizing granular data and improving existing risk models, AI can complement human expertise in underwriting and provide valuable insights for risk selection and pricing. This may ultimately help insurers write more policies with consistent pricing for similar risk profiles.
“On the other hand, since underwriting decisions have a direct impact on profitability, AI models need to be carefully selected, trained and tested as otherwise mispriced policies could result in very serious reputational and financial implications.”
AI adoption by insurers has many applications and can contribute to a company’s growth and profitability, but it also introduces new risks that could cause significant damage.
Some activities can be less risky compared to others. “For example, determining marketing leads based on AI recommendations is generally a low-risk proposition. At other times, benefits and risks are closely related such as in terms of how AI impacts customer experience,” analysts explain.
Noting: “But, in our view one of the most serious challenges arises when AI is used extensively in underwriting and pricing of policies as those decisions are directly related to profitability.
“In those situations, the insurer could be subjected to various costly errors and biases (i.e., quoting unreasonably high/low premiums for characteristics that are not well represented in the data used in training AI models). Additionally, there could also be regulatory fines amid the evolving regulatory landscape. Equally concerning could be certain decisions related to claims processing.”
Smaller companies often face challenges in managing risks due to less developed frameworks, limited resources and restrictive data, leading to potential decision-making errors.
Wipro’s research highlights a concerning trend: even with extensive AI adoption, many companies lack clear data usage policies, a problem particularly prevalent among smaller businesses.
As an industry that leverages large datasets of information, it is not surprising that insurers are increasing investments in AI, Morningstar DBRS said, which is also needed to stay competitive.
“At the same time, they must not lose sight of the importance of having commensurate risk management frameworks. From a credit rating perspective, AI can both enhance and damage franchise strength by affecting customer experience. Moreover, while it may improve profitability through efficiency gains, it generally also contributes to higher operational risks, including legal and compliance risk,” Morningstar DBRS concluded.
