Original text here from Patrice Bernard (LinkedIn)
McKinsey consultants occasionally surprise us. While everyone is buzzing about artificial intelligence—generative AI being the latest trend—they offer a more measured view of its potential in the insurance industry. It may be less flashy, but it’s probably more realistic than most other forecasts.
To kick things off, they pose the classic question: which use cases are most promising for the industry and, more importantly, most likely to create real value? Four possibilities emerge, but only two are truly specific to insurance, after excluding those focused on assisting with software programming and optimizing customer service—issues Gartner also highlighted, as I mentioned yesterday.
Subscribe to our newsletter:
Focusing more on the core business, one opportunity lies in the search and extraction of valuable information from unstructured data sources (documents, forms, and other exchanges). This might involve analyzing the details of a claim or summarizing a company’s policy. The other opportunity involves generating communication materials, such as updating a policyholder on the status of their claim or negotiating terms with an agent.
The scope may seem limited, but it’s pragmatic. Once we move beyond the fantasy of a magic solution—often oversold by many—we need to come back to reality and assess the technology’s actual capabilities. McKinsey wisely reminds us that generative AI alone won’t deliver the expected benefits; it needs to be combined with other approaches—like data science and robotic process automation (RPA)—to fully realize its potential.
Despite this clearly defined scope, most companies remain stuck in the "pilot purgatory," unable to move projects beyond the experimental phase into full-scale production. The primary reason is familiar: traditional organizations lack the expertise in testing protocols, particularly in objectively determining success criteria and shutting down projects that don’t meet them.
It’s true that the discipline required is even harder to master when dealing with artificial intelligence, given its unique risks and uncertainties—such as sensitive data protection, inherent imperfections (like hallucinations), potential biases, and the associated ethical and regulatory challenges—which are especially daunting for financial institutions. These challenges demand the creation of appropriate management plans, which are currently lacking.
In summary, those who will successfully harness the potential of generative AI will do so by integrating three key elements: a clear understanding of the functions that the technology can and cannot fulfill (and leveraging other options for the latter), establishing an innovation governance framework (applicable to other emerging concepts as well), and finally, creating a formal, adaptable risk management structure.
