How to revolutionize the insurance value chain with generative AI

Insurers face significant barriers in their pursuit of strong returns on their investments in GenAI, as EY survey made clear.

1. Regulatory ambiguity

Two-thirds of respondents cite regulatory ambiguity as the top barrier to establishing a dedicated team for GenAI, though there is large variance by line of business. Among life and annuity insurers, 89% of our respondents cite regulatory ambiguity as the greatest barrier, compared to only 39% of P&C carriers. The difference may be attributed to the EU AI Act’s framework, which defines life and health insurance providers that use AI technology as high-risk, although questions remain about how the AI Act will be enforced.

Despite the lack of clear or comprehensive rules, insurers, like their peers in other sectors, are making major investments in AI. However, lingering regulatory uncertainty may stifle investments in some specific capabilities, particularly areas with a direct impact on customer experience and pricing decisions.

The broad-based guidelines issued to date, for example President Biden’s October 2023 executive order and the EU AI Act, acknowledge the challenges associated with the evolving GenAI regulatory environment. Global insurers are also tracking the conduct-related regulations established by several authorities across the Asia-Pacific region, including in Australia, Hong Kong, Japan and Singapore. In a heavily regulated sector like insurance, it’s understandable that executives are cautious about future regulation, even as they respond to the imperative to invest in AI.

2. Uncertain ROI models

While there are many options for deploying GenAI across the business, developing clear models for return on investment (ROI) can be challenging, according to the survey respondents. In fact, more than half indicate that uncertainty about expected ROI is a top barrier to establishing a dedicated GenAI team. The pressure to develop clear ROI models is likely linked to the urgency many executives feel to adopt GenAI quickly and at scale. Further, it’s difficult to estimate GenAI’s potential impact on loss ratio, for example through more sophisticated underwriting and more accurate pricing, and other key metrics.

However, it’s worth noting that 53% of insurers that have implemented GenAI did so because of expected productivity gains, without necessarily factoring in tangible cost savings. Our results suggest testing and learning with initial deployments may hold the key to setting definitive ROI targets.

3. Data privacy, quality and security

GenAI will only add to the pervasive concerns about data privacy, quality and security. In fact, more than half of respondents identify such concerns as the top barrier to exploring GenAI initiatives. The most common worry is that GenAI applications will expose sensitive customer or corporate data. But many business leaders recognize that GenAI apps built on low-quality, out of date or incomplete data will lead to “garbage in, garbage out” scenarios, greatly limiting the value of GenAI. These are legitimate concerns given the longstanding data management integration challenges insurers face.
In light of these issues, some carriers, as well as companies in other sectors, have prohibited the use of publicly available GenAI tools. These bans are understandable, considering that insurers, and life and health insurers in particular, possess huge volumes of sensitive customer data and employees may be tempted to feed that data into external tools such as during GenAI experimentation. Over time, however, at least some insurers will lift the prohibition and sanction some usage within defined guardrails.

EY results show optimism about addressing these concerns over the long term; for instance, more than 50% of respondents indicate that their firm should establish regular monitoring, designated teams and governing bodies when implementing GenAI. Additionally, 68% of insurers expect to use closed-source large language models (LLMs) while 24% prefer open-source LLMs. Establishing trust and security in all types and sizes of LLMs – from the largest, scaled platforms to smaller, proprietary models – is a priority.

4. Integration with legacy IT systems

Integrating with existing technology and data sources is another common hurdle to GenAI adoption, particularly at companies with fragmented data infrastructures and limited analytical capabilities. Siloed systems can increase implementation costs, too. More than half of our respondents, 54%, cite cost of implementation as a challenge to GenAI adoption, and 46% indicate high energy consumption as another key barrier.

5. Risk of deploying in client-facing applications

The vast majority of insurers – 95% in the survey – are prioritizing administrative and back-office use cases. A significant number, 43%, are awaiting further development and testing before deploying GenAI in client-facing, front-office applications. Not surprisingly, InsurTechs will likely be the first movers for front-line deployments; 56% of respondents from InsurTechs say they are very excited about client-facing use cases for GenAI.

Insurers have plenty of experience with disruptive technologies and can apply lessons from the past to chart the right course forward. The immediate priorities are to systematically identify and pilot use cases, while simultaneously establishing the necessary governance, infrastructure and operating models to scale adoption successfully. As both these tracks are critical, leaders must strike a balance in managing them. 

Findings from EY research confirm our experience working with insurers around the world – that it’s critical to both take quick action to start on the GenAI journey and methodically build a stable foundation for future success. Insurers that move too slowly risk being leapfrogged by the competition. And those that fail to build core infrastructure, governance and capabilities will struggle to consistently develop and scale GenAI applications and to maximize returns on their investments. Firms that embrace a dual-track approach will be well positioned to unlock immediate value even as they set the stage for sustainable success at scale.

1.  Encourage experimentation by equipping your people with GenAI tools

Accelerate grassroots innovation by deploying GenAI tools, including both internally and externally developed copilots and sandbox environments, and empowering users to explore and develop potential use cases. Appropriate guardrails can promote a safe and secure environment, such as protections on using internal data with public tools. Start with talent upskilling and foundational training, then prioritize quick-win use cases that can be tested and launched efficiently and with low risk, such as data summarization and report generation across near-term time horizons, for example eight to 18 months.

• Take immediate action but prepare to fail fast: Leaders of strategy and innovation teams can champion a grassroots culture and spearhead initial pilots as they understand where targeted GenAI applications can be applied to address specific business problems. Encourage all staff levels and functions to test use cases autonomously with minimal oversight in sandbox environments. EY teams are working with a speciality insurer that has formed cross-functional teams to identify unmet business needs and assess how GenAI solutions can help, for example automating the policy renewal process.
  
  
• Prioritize and pilot quickly: “Lite” frameworks can help determine the viability of GenAI solutions for starting-point use cases. To identify quick wins, leaders can weigh business value against the ability to execute. EY teams recently supported a life insurer in evaluating and prioritizing a lengthy backlog of use cases; identifying “clusters of value,” for example enhanced underwriting, ease of doing business, was critical to short-listing use cases aligned to enterprise goals.

Rapid piloting throughout the organization can quickly validate assumptions and hypotheses. For other insurers, EY teams are conducting six- to eight-week sprints to streamline the development and launch of high-priority use cases.

• Set up guardrails: Determine the organization’s risk appetite and develop clear guidelines on what is and is not allowed for employee use of GenAI. Deploy off-the-shelf models in a sandbox environment to accelerate innovation through “test-and-learn” iterations. For a leading global insurance carrier, EY teams helped in creating  a dedicated, fit-for-purpose development environment to promote experimentation and testing with innovative ideas.

Compliance with the EU AI Act will be a baseline expectation. Other EY research shows senior leaders are focused on establishing the right guardrails and policies. In a recent survey of insurance chief risk officers (CROs), 81% said their firms had already enhanced or were in the process of enhancing policies around development, validation and use of GenAI. A similar number, 82%, said they had established a governance framework, policies and procedures for LLM and GenAI adoption or were working to establish them. You can see the full results here.  

2. Develop the enterprise strategic and technical infrastructure for GenAI

Even as they pursue quick wins and seek to build confidence with AI, insurers should craft the overall vision and determine how GenAI can transform the business over the long term. Key elements of the GenAI strategy include the delivery model, data governance, computing infrastructure (i.e., on-premise vs. cloud) and a prioritization framework to support desired business outcomes.

Ultimately, success with GenAI will be determined by innovative, externally oriented use cases that provide tangible long-term benefit such as, revenue growth and more accurate risk pricing. For many insurers, conducting a thorough assessment of organizational readiness and the maturity of current capabilities will be a sensible first step.

• Mobilize cross-functional teams and centralize management: Multidisciplinary teams with different competencies, such as business, IT, security, compliance, should be charged with designing and delivering end-to-end solutions. We have seen first-hand how centralized management, coupled with enhanced capabilities and select competencies, provides a strong foundation for successful GenAI programs.

As GenAI adoption expands across the value chain, cross-functional teams should have the support of senior-level sponsors. Increasingly, we expect more heads of dedicated GenAI teams will report directly to the C-suite, including the CEO.

Most firms will need to fill key competency gaps with external talent. Engaging with third parties and tapping into an ecosystem of GenAI players can further accelerate adoption; large insurers may establish a framework to inform build-buy-partner decision making.

• Establish data infrastructure and governance practices: As insurers mature their GenAI use, they should enhance governance and policies in line with their own needs and shifting regulations. Continuous monitoring of the regulatory landscape is essential.

Strong data governance starts with understanding the source of data, where it resides and who has access to it. Identify critical data that is used to generate judgements and establish clear controls for decision-making processes. Conduct regular reviews of AI-generated content. By maintaining data transparency and accountability, insurers can avoid “garbage-in, garbage out” scenarios.

Insurers should establish a strong data environment, such as closed- or open-source LLMs, to create, train and deploy GenAI models and solutions, and plan to enhance it on an ongoing basis. In parallel, they should design a capability for processing and managing unstructured and semi-structured data that supports reusability in line with strategic priorities.