The commercial property and casualty (P&C) insurance sector has been slow to adopt artificial intelligence (AI), but new trends in data augmentation could help pick up the pace, according to experts on a recent Insurtech Insights panel.
Ilya Kolmogorov, group chief pricing actuary at Zurich Insurance Group, noted that while machine learning models are entering the commercial insurance (CI) space, the progress is happening at a slower rate compared to other industries. However, he highlighted a significant advancement in data augmentation, stating, “We are no longer relying just on the available historical data. Instead, we are trying to augment the data as much as possible.”
Balancing AI Implementation with Risk Considerations
The key for American P&C insurers lies in balancing AI implementation with risk considerations while keeping a close eye on compliance and regulation, panelists emphasized. Christy Kaufman, VP of P&C risk and chief compliance officer at USAA, pointed out that compliance professionals often focus on potential downsides, but she believes the biggest risk is not taking enough risk and failing to move quickly enough to stay competitive.
“We have to go forth and go forward, but we also have to know where the pitfalls are and how we’re going to guard against them,” Kaufman said. She recommended that insurers invest in training, tailor their approach to AI implementation, and involve independent teams to ensure models account for crucial factors like fairness, data quality, and transparency.
Unique AI Considerations in Commercial Insurance
Kolmogorov acknowledged that commercial insurance has unique considerations that make AI adoption more complex. “We have a much more heterogeneous exposure type on the CI side than on the personal line side, and that makes modeling much more complex,” he explained. The large amounts of data available can sometimes lead to “overfeeding” an AI model, making it less effective.
However, Kolmogorov suggested that data augmentation could be a solution. This involves not just adding more data, but rather better structuring existing data and adding variation to help models learn. “We’re not only looking at historical records in terms of policy exposure and claims records, but we’re really pulling all available information that’s out there,” he said.
Regulatory Challenges
A major consideration for P&C insurers is the changing regulatory environment in the United States. Kaufman advised insurers to develop a “programmatic, systematic way to ingest new regulation as it comes along.” She acknowledged that this could be challenging due to the rapid pace of AI advancements and regulatory changes.
Kaufman suggested a “train-the-trainer” approach to working with model developers, ensuring they understand compliance requirements from the outset. This approach is necessary because it’s not realistic to be involved in every model development process.
The Insurtech Insights panel highlighted the need for commercial P&C insurers to carefully consider their approach to AI adoption, balancing the potential benefits with the risks and regulatory challenges involved.
