Reimagining Insurance with Generative AI:

The Ultimate Guide for Business Leaders

In the span of just a few years, generative artificial intelligence (GenAI) has transitioned from just another concept in AI research to a major driver of business value. Unlike traditional AI, GenAI can create unique outputs — from text, images, audio and video, to code and data — through simple human commands executed within seconds. This revolutionary shift has spurred organizations in industries as diverse as banking, healthcare, and manufacturing to invest in GenAI for improved customer interactions, enhanced operational efficiency, and unprecedented product innovation.

Generative AI is quickly becoming the new normal in the insurance industry. Nationwide is exploring the use of GenAI as an assistant for customer service representatives. The GenAI platform analyzes customer calls in real time, providing prompts and data to help representatives resolve issues swiftly and improve the customer experience. HDI Global is utilizing its proprietary GenAI system, HDI-GPT, to extract real-time insights from unstructured data, be it text or images. GenAI assists the staff in interpreting legal contract terms and conditions through content summarization and real-time assessment of insurance claims.

The potential of generative AI in insurance is immense. According to estimates from the McKinsey Global Institute, insurance organizations are projected to be one of major beneficiaries of GenAI, with an annual potential of $50B to $70B. This represents about 3% of the total industry revenue, with increases driven by better operational efficiency and the automation of micro-tasks such as claims processing and underwriting. GenAI use cases in customer-facing operations, marketing and sales, software engineering, and risk and compliance are also cited as major value drivers.

For the insurance industry, adopting generative AI means more than keeping up with technology trends. It is about making a strategic choice to innovate beyond traditional AI to remain competitive. Tried-and-true AI use cases such as predictive analytics, anomaly detection, and automated document processing can now be complemented by GenAI solutions for hyperpersonalized customer service and insurance advice; intelligent document search, summarization, and generation; and custom knowledge agents for business teams. With generative AI, insurance leaders have a once-in-a-lifetime opportunity to redefine their business.

This guide offers C-level executives in insurance a clear, actionable roadmap for adopting generative AI. From understanding basic principles and potential applications to tackling the complexities of building their first solution, this Guide will help executives to navigate the GenAI landscape, to drive growth, efficiency, and innovation.

General Considerations for Generative AI in Insurance

Generative AI's ability to mimic human creativity and intelligence has prompted a significant — and often unwarranted — shift away from traditional AI. Both AI and GenAI are at a critical juncture, and leaders need to look beyond the hype to gain a nuanced understanding of how these technologies can complement each other for the benefit of their organizations.

The following considerations should be on the radar of insurance leaders as they evaluate the adoption of generative AI:

Data Availability

Insurance organizations are ideally positioned to leverage both traditional “predictive” AI and generative AI. They possess vast amounts of highly diverse customer- and business-centric data of high quality that can be used to develop custom GenAI solutions to build significant competitive advantages in decision-making, optimization of systems and processes, and enhancement of productivity and creativity for both technical and non-technical users.

Balance of Use Cases

Generative AI is not a replacement for AI. Traditional AI use cases in insurance such as risk assessment and pricing, fraud detection and prevention, and claims processing (among many others) should be leveraged for their strengths in prediction, classification, anomaly detection, and personalization tasks. Generative AI should serve to complement existing AI capabilities, helping to build a more customer-centric organization and modernize operations.

Innovation S-curve

The adoption of generative AI in insurance will likely follow an S-curve pattern: from its use in text creation and research by individual users to augmenting specific operations (e.g. data extraction from claims, report generation), to transforming core processes and functions (e.g. GenAI-powered contact center, GenAI assistant for insurance underwriting). GenAI can augment well-established AI solutions to support both tactical and strategic decision-making.

Adoption Costs

When demand for GenAI compute outstripped supply by 10x, access to affordable resources was a critical factor in GenAI adoption. As more organizations began using GenAI services and models available via the cloud and FM providers, it became apparent that, while development costs remain a consideration, operational costs do not surpass those of traditional AI systems. In fact, GenAI use cases, such as report summarization, are more cost-efficient than their AI counterparts.

Compliance and Security

When adopting any form of AI, insurance organizations should prioritize addressing threats to data security and privacy, system availability, operational continuity, interoperability, auditability, and compliance with legal requirements. The availability of GenAI's foundation models (FMs) in the cloud ensures the technology's safety. By default, cloud providers enforce strict security measures while adhering to regulatory standards.

Ethics and Bias

Addressing ethics and bias requires a multifaceted approach, including diverse and inclusive training data, ethical guidelines for AI development, transparency about how models work, continuous monitoring for biased outcomes, and mechanisms for accountability. The adoption of Reinforcement Learning with Human Feedback (RLHF), Explainable AI (XAI), and Human-in-the-Loop (HITL) are recommended to ensure fairness and transparency.

Overview of Generative AI Use Cases in Insurance

Like organizations in many other industries, insurance companies are discovering opportunities to leverage generative AI in use cases that are organized horizontally by “broad” business function. These “horizontal” use cases encompass such business functions as:

Marketing and sales
Customer operations
Risk and compliance
Software engineering

Short-term Impact

Strategy and finance
Supply chain and operations
Corporate IT and product R&D
Talent organization

Long-term Impact

In practice, “horizontal” GenAI use cases empower a variety of applications, including dialogue generation for virtual assistants, automated code generation, personalized messaging for marketing and sales, and much more.

However, the insurance industry has specific requirements. “Vertical” GenAI use cases — the ones that are organized by unique and more focused, sector-specific tasks — present more opportunities to create sustainable and measurable value. But these opportunities require domain knowledge, contextual understanding, and expertise, often necessitating the fine-tuning of existing foundation models. Some “vertical” GenAI use cases include:

Claims processing

Q&A agents for SMEs and BUs

Insurance underwriting

Group plan customization

Loss control and prevention

Quote and policy generation

Product personalization

Intelligent document agent

Insurance pricing

Insurance advice agent

Both horizontal and vertical use cases are more likely to realize at-scale business value for an organization if they are used holistically; that is, if they are seamlessly integrated with other systems, and into processes and teams.

Considering that the insurance industry holds vast amounts of sensitive customer data, it is crucial to strike the right balance between harnessing value from generative AI and managing the associated risks, like hallucinations, biases, phishing, and prompt injections. Consider also the potential for long-term vs. short-term profitability and growth, cost savings, the balance between impact and investment, and the time to achieve measurable impact.

Choosing the Right Foundation Model for Insurance

Choosing the right foundation model involves striking a balance between scale and specificity. This choice can be complex because:

There are numerous models available from cloud providers like Amazon Web Services (AWS), or from FM providers such as Anthropic, Cohere, Meta, Stability AI, AI21 Labs, and many others.

The effectiveness of FMs, especially Large Language Models (LLMs), hinges more on how they are fine-tuned to specific tasks, rather than the models being inherently focused on insurance.

The required infrastructure capabilities, along with considerations of cost, security, and performance, are critical factors to weigh. Navigating these choices is complex, requiring careful evaluation of trade-offs.

When choosing a foundation model, insurance organizations should begin GenAI adoption with a specific use case (their intended business goal). First, data and infrastructure is assessed and evaluated. Second, the discovery phase begins, to create a list of use cases with the highest ROI. Third, once the most impactful use case is selected, a Prototype is built to determine if the GenAI solution can drive business value in production.

There are additional factors to consider when selecting the appropriate foundation model for a generative AI use case, and vice versa:

The use case discovery phase begins by identifying the problem and a specific action that the organization wants to achieve. Different use cases require various types of foundation models, and it is essential to establish clear goals and metrics from the start. These will direct the selection of the most appropriate model type, as well as the model’s performance, knowledge and understanding levels, latency, and required security measures.

GenAI action to complete:

Text generation
Text summarization
Text translation
Text classification
Information retrieval
Question answering
Sentiment analysis
Vector search

Foundation models differ not only in type but also in size, which is primarily determined by the number of parameters they contain. In general, large models provide robust performance and offer superior reasoning abilities at increased cost. Smaller, more specialized models are more affordable and easy to deploy, yet they can be challenging to train, maintain, and scale across numerous GenAI use cases.

Foundation model size factors to consider:

Supporting architecture
Training data (volume, variety, quality)
Optimization technique (e.g. quantization)
Transformer efficiencies
Choice of learning frameworks
Model compression techniques
Cost of inference (run-time)
Environmental impact

Selecting a foundation model for a GenAI use case should be a forward-thinking process. Organizations should consider the immediate task and its potential scalability organization-wide — if such scalability is feasible. Given the specifics of “vertical” GenAI use cases, insurance organizations should understand the implications of scaling on costs, performance and ROI, while also balancing it with GenAI experimentation to find a better FM.

Scaling components to pay attention to:

Data residency requirements
Estimates of the traffic volume  that will be served
Talent: upskilling, recruitment, or both
Costs: minimal up-front investment vs. significant spend down the road
Operations: in-house vs. managed AI service

Choosing the right foundation model is a nuanced process that may require collaboration with a third-party AI service provider for guidance. The selection journey should be an integral part of GenAI exploration, allowing insurance leaders to delve into the specifics of cost, performance, and risks associated with the technology. By doing so, insurers will be able to confidently harness measurable value from the GenAI applications they intend to build.

Customer Success:
 Streamlining Insurance Underwriting with Generative AI

Our client is an international specialty insurance company that combines its experience and reputation with a legacy-free balance sheet, building a range of insurance and reinsurance products, and digital risk management solutions.

Challenge

The client wanted to reinvent its underwriting operations by enabling its underwriters to assess risks and make faster and more accurate decisions about extending or denying policies to potential clients with the help of Generative AI. Previously, its subject matter experts (SMEs) would spend days reviewing 100-page technical survey reports of the objects to be insured before making an underwriting decision. With generative AI, the survey reports can be summarized in minutes, dramatically accelerating policy issuance while balancing risks.

The undifferentiated heavy lifting associated with manual insurance underwriting

Subject matter experts spend precious time on report processing, not on decision-making and policy issuance

The number of reports that can be processed per week is limited by SMEs on staff

Solution

Provectus developed a GenAI-powered solution to summarize text and extract information from technical survey reports, including such highlights as object information, positive and negative points of interest, and object KPIs. At the core of the solution is Cohere’s Command Large Language Model (LLM) hosted on Amazon Bedrock. Provectus adjusted the in-context learning for the model to perform summarization and extraction tasks, specifically tailored to the engineering reports in PDF format.

Architecture diagram of the GenAI solution

Outcome

The client received an advanced GenAI solution that can accurately process technical survey reports and generate rich summaries, shortening the time needed to make policy decisions from days to minutes. The summarization process takes about 10 minutes per report, helping the client to scale decision-making with less human intervention. Generative AI enables the client to insure more objects, reduce the cost of products sold, and increase annual recurring revenue (ARR).

Time to insight reduced from about 10 days down to less than 10 minutes

GenAI’s ability to highlight key aspects of technical survey reports is as good as that of subject matter experts

GenAI’s summarization quality is similar to or on par with that of subject matter experts

How to Get Started:
 Generative AI Adoption Journey with Provectus and Cohere

At Provectus, we believe that a generative AI use case should serve as the essential first step for any company looking to adopt and scale generative AI, while realizing business outcomes. Provectus’ subject matter experts and technical team, in close collaboration with Cohere, have developed an approach to help enterprises quickly adopt the most impactful GenAI use cases. It encompasses two phases:

Phase I

Data readiness and use case prioritization workshop that unites IT and business teams to help identify the best place to start for the highest return.

Phase II

Building the Prototype in your organization’s AWS account to test the use case value and determine how impactful generative AI can be.

Our approach is designed to encourage cross-functional collaboration between C-suite Executives, Business Unit Leaders, Data and AI/ML Leaders, and IT professionals. The goal is to streamline the adoption process and effectively measure success criteria.

Learn more

Deliverables:

Data and infrastructure discovery
Prototype of the selected use case deployed on AWS
GenAI use case prioritization and adoption roadmap
Solution architecture documentation
Expected business value and success criteria research
Detailed plan to deploy to production
Estimation of use case complexity and risks
TCO estimates for AWS and FM services

Conclusion

Generative AI is transforming industries rapidly, and leaders in insurance need to closely monitor this technology, poised to add value of between $50B and $70B annually from its adoption. Generative AI presents a prime opportunity to compete in a highly dynamic market where InsurTech startups are challenging the dominance of established enterprises.

However, generative AI adoption is not always straightforward. It necessitates comprehensive organizational changes business- and technology-wise, starting from the adoption of the first GenAI use case, to scaling the technology across various processes and functions. Provectus is ready to support your insurance organization as a valuable partner in your GenAI adoption journey.

Leveraging its deep expertise in the AI domain, Provectus excels in end-to-end AI transformations for enterprises, and in implementing AI solutions on a use-case basis. With pre-built GenAI solutions, Provectus offers its customers risk-free Prototypes, ready to go live within weeks so you can quickly evaluate the impact of generative AI on your operations.

Provectus pioneers the adoption of generative AI technologies with AWS. As an AWS Premier Tier Services Partner with the AWS Machine Learning Competency, AWS Data & Analytics Competency, AWS DevOps Competency, and AWS Migration Competency designations, Provectus is one of the few partners to add AWS Generative AI Competency to its list of specialized credentials.

Our strategic alignment with AWS extends beyond technological advancements. To drive the adoption of generative AI, AWS works closely with Provectus and other partners to fund their customers’ projects. Given AWS’s strong focus on the banking, finance, and insurance industries, insurance organizations are well-placed to receive funding, to help mitigate the costs associated with developing, adopting, and maintaining GenAI solutions.

To address the existing and potential risks of GenAI adoption for its customers, Provectus has launched The AI Landing Zone accelerator, designed to help enterprises take their first steps toward the application of generative AI for real-world use cases while keeping potential risks in check.

From day one, Provectus has stayed true to its mission to help businesses reimagine the way they operate, compete, and deliver customer value with AI. Generative AI marks an exciting new chapter for Provectus, creating unprecedented opportunities for our customers. We encourage all forward-thinking executives to explore, adopt, and leverage generative AI to solidify the future success of their businesses.

Begin your organization’s generative AI journey by reaching out to us today!

Download The Full Guide

See the Provectus privacy policy for details on how we collect, use, and share information about you.