Scaling GenAI: Considerations for Insurance Leaders
Moving Ahead with GenAI
It’s hard to believe how quickly Generative AI (GenAI) technology has gained traction across industries worldwide. Since OpenAI introduced the ChatGPT API in March 2023, the technology has seen widespread adoption, with an estimated 2 million developers building GenAI applications by November 2023.
In the insurance industry, carriers have been actively exploring and gradually embracing GenAI technology. Many started their GenAI journey with the goal of providing their employees with secure access to public large language models (LLMs) like ChatGPT. From there, they proceeded to develop custom applications primarily aimed at enhancing internal productivity. While most carriers are still in Proof of Concept (POC) mode, a growing number have begun releasing solutions into production, with plans to scale GenAI usage across the insurance value chain.
As insurance leaders look to expand and scale their use of GenAI they should consider several important factors:
1.Build vs. Buy: Evaluate whether to develop your own LLM or use existing models, and whether to buy external vendor solutions to complement your own capabilities.
Carriers can choose to build their own custom LLM models or leverage existing open source or commercially available models like ChatGPT. Building custom LLMs offers maximum flexibility and control, reducing risks such as hallucinations. However, the cost and complexity of constructing a Large Language Model (LLM) can be challenging for all but the largest companies. As a result, most carriers currently prefer to develop their own solutions on top of third-party LLMs.
When using existing LLMs, insurers are strongly advised to use techniques like Retrieval Augmented Generation (RAG) to fine tune their accuracy, relevancy and explainability. Implementing RAG will require data engineering skills and the appropriate infrastructure.
Insurers should also consider GenAI offerings from Insurtech and software vendors for specialized use-cases such as customer service chatbots, and code generation tools such as Microsoft GitHub Copilot.
2.Consider the Future of GenAI: Define a GenAI strategy that positions you for the future
On its current trajectory, GenAI will soon become pervasive throughout the insurance life cycle. However, many of the GenAI capabilities that are being widely implemented today are rapidly becoming table stakes. To get greater utility from this technology it will need to eventually work seamlessly with other AI technologies and become an integral part of enterprise workflows and business processes. To achieve this, enterprises will need to adopt a multi-agentic approach, utilizing multiple specialized LLM models, and traditional AI/ML models working together as a team to solve complex problems and activating tasks across multiple API-enabled applications.
All of this will require modern data infrastructure and in-house AI and data engineering capabilities.
3.Establish a Modern Data Foundation and AI Capabilities: Establish a robust data infrastructure and AI capabilities to support effective scaling of GenAI throughout the enterprise.
To fully capitalize on the transformative potential of GenAI, insurers will need to establish a modern cloud data foundation and develop the necessary capabilities to scale and integrate GenAI technology throughout the enterprise. This includes having a strong data infrastructure, in-house data engineering and AI talent, MLOPS and AI modeling tools, RAG frameworks, API integration frameworks, and data and AI governance practices.
By having these foundations in place, insurers can industrialize their GenAI development with end-to-end management of data flows and models, embedded controls and governance, and increased automation. In addition, it positions them for future use-cases including Multi-Agent systems.
4.Planning for AI Regulation and Responsible AI: Establish AI and Risk Governance frameworks.
Insurers need to stay abreast of the rapidly evolving AI regulatory landscape and the importance of responsible AI practices to help mitigate regulatory, legal, and reputational risks.
As insurers begin to scale their GenAI efforts they should ensure that they have appropriate AI governance and risk management frameworks in place to ensure responsible and ethical use of AI technologies throughout the end-to-end development lifecycle. The NAIC Model AI Bulletin and the NIST AI Risk Framework provide guidelines for responsible AI use in insurance. The NIST framework has recently been extended to include the unique risk considerations for GenAI.
For more information, refer to the Celent report: Data Governance For Insurers In the Age of AI
GenAI technology offers immense potential for the insurance industry, but scaling and realizing its benefits will require careful planning and investment. Insurance leaders will need to establish responsible AI practices, ensure they have modern data foundations and capabilities to support existing and future use-cases, and carefully consider their build vs buy options.
For more leading industry insights in the rapidy evolving field of Gen AI, please join us at Celent’s Generative AI Symposium event on Sept 18, 2024, in New York.