What You Will Learn

• How LLMs are used as employee co-pilots in insurance
• What "computable contracts" are and their benefits
• AI-driven claims automation with computer vision and telematics
• Managing AI risks, bias, and ethical deployment

Study Guide

Introduction

Welcome to the comprehensive guide on "How The Insurance Industry Is Using AI To Optimize Business," inspired by insights from a Forbes video course. This course offers a deep dive into how artificial intelligence (AI) is revolutionizing the insurance sector. With leaders from AXA, Allstate, Hosta AI, and Fai AI sharing their expertise, you'll learn about current AI applications, future transformations, risk management, and innovation strategies. Understanding these concepts is invaluable for anyone in the insurance industry or those interested in AI's transformative potential.

Current AI Applications and Use Cases

Large Language Models (LLMs) for Internal Productivity
One of the most significant advancements in AI is the deployment of Large Language Models (LLMs) to enhance productivity within insurance companies. AXA, for example, has integrated secure access to OpenAI's LLM backend for its 150,000 employees. These models act as "co-pilots," assisting with small tasks that collectively improve efficiency. This approach not only uses the intelligence of LLMs but also leverages collective user intelligence through feedback mechanisms.
Example: Employees might use LLMs to draft initial policy documents or respond to customer inquiries, significantly reducing time spent on these tasks.

"Computable Contracts"
AXA is also pioneering the use of LLMs to translate insurance contracts into system specifications, a method known as "computable contracts." This approach allows for querying contracts once per contract rather than per claim, offering significant cost advantages.
Example: By automating the interpretation of contract terms, call centers can provide faster and more accurate responses to customer queries.

AI in Claims Processing
Allstate is focusing on AI to enhance data collection at the claim point, using telematics and computer vision to assess accident severity and streamline claims processing.
Example: AI can automatically assess damage from accident photos, reducing the need for manual inspection and speeding up claim settlements.

Computer Vision for Claims Automation
Hosta AI utilizes foundation vision models to automate claims by creating a spatial understanding of 2D images.
Example: This technology can automatically detect and document damage in property claims, reducing the time adjusters spend on-site.

AI for Customer Acquisition
Fai AI enhances customer acquisition for property and casualty insurers by understanding risk profiles and using generative AI for personalized content delivery.
Example: Insurers can tailor marketing messages to individual risk profiles, improving conversion rates and customer satisfaction.

Future Transformation of Insurance Functions

Pricing
In the short term, AI is expected to refine existing pricing models through better data utilization. In the next five years, pricing may become more dynamic, leveraging real-time data such as telematics to adjust policies based on behavior.
Example: A driver with safe driving habits could receive lower premiums based on real-time telematics data.

Underwriting & Risk Assessment
AI will enable more granular risk assessment using data from various sources, including satellite and appliances. This could redefine insurability, especially in high-risk areas affected by climate change.
Example: Satellite imagery could help assess flood risks more accurately, influencing underwriting decisions.

Claims Processing
AI models are maturing to improve data collection, fraud detection, and claim processing. In the long term, fully automated claims processes are anticipated.
Example: AI could automatically flag fraudulent claims by analyzing patterns across multiple data sources.

Customer Acquisition
AI will drive more personalized strategies, integrating risk assessment and pricing from the start. Transparency in rating factors will be crucial.
Example: Potential customers could receive personalized quotes based on a comprehensive risk assessment conducted at the initial contact.

Shifting Balance of Risk

New Forms of Risk
AI introduces risks such as model biases, which can lead to discriminatory outcomes. For example, a lawsuit against State Farm highlighted biases in fraud models.
Example: AI models trained on biased data could unfairly penalize certain demographics, necessitating careful oversight.

Changing Nature of Traditional Risks
The rise of autonomous vehicles shifts risk from individual drivers to potential systemic risks.
Example: A software flaw in autonomous vehicles could lead to widespread incidents, requiring new risk management approaches.

Managing Model Risk
With the proliferation of AI models, managing "model risk" becomes a complex challenge.
Example: Insurers must ensure that interconnected AI systems function correctly and do not amplify errors.

The Role of Human Capital

Talent Dynamics
As AI adoption grows, the insurance industry must attract talent capable of strategic thinking and innovation. Roles may evolve, with a fusion of actuarial and data science skills.
Example: Actuaries may need to develop skills in data analytics to complement traditional risk assessment methods.

Ethical AI Deployment
Human expertise is crucial for managing complex risks and ensuring ethical AI use.
Example: Teams dedicated to AI ethics can help prevent biases and ensure fair outcomes in AI-driven processes.

Innovation Strategies and Startup Engagement

Building to Spinoff
AXA employs a "build to spinoff" strategy, creating internal innovations that can scale with external capital.
Example: A successful AI tool developed in-house could become an independent entity to attract further investment.

Partnering with Startups
Allstate emphasizes the importance of partnering with startups that offer clear value and have executive sponsorship.
Example: A startup providing innovative fraud detection solutions could partner with Allstate to pilot and refine its technology.

Data Access Challenges
Startups often struggle to access the vast datasets held by insurers. Successful collaborations focus on specific, high-value problems.
Example: A startup might develop a niche solution for property damage assessment and seek partnerships to access necessary data.

Startup Opportunities

Insuring AI
Opportunities exist in developing insurance products that address AI-related risks, such as transparency and fairness.
Example: A startup could offer policies covering potential damages from AI model failures.

Proactive Risk Management
Startups can focus on predicting and preventing risks, shifting from reactive to proactive insurance models.
Example: AI-driven insights could help policyholders mitigate risks before they result in claims.

Conclusion

By completing this course, you now have a comprehensive understanding of how AI is transforming the insurance industry. From enhancing productivity with LLMs to redefining risk management and customer acquisition, AI offers numerous opportunities for innovation and efficiency. However, thoughtful application and management of AI technologies are crucial to address new risks and ethical considerations. As the landscape evolves, staying informed and adaptable will be key to leveraging AI's full potential in the insurance sector.

Podcast

There'll soon be a podcast available for this course.

Frequently Asked Questions

Introduction

Welcome to the FAQ section for the 'Video Course: How The Insurance Industry Is Using AI To Optimize Business by Forbes.' This resource is designed to provide comprehensive answers to common questions about the integration and impact of AI within the insurance industry. Whether you're new to AI or a seasoned professional, this FAQ aims to offer insights and practical advice on leveraging AI for business optimization in insurance.

How is AI currently being used within the insurance industry, and can you provide specific examples?

AI is being implemented in various ways. Large Language Models (LLMs) are being deployed as secure co-pilots for employees across vast organisations like AXA, assisting with smaller tasks and leveraging collective intelligence through user-generated and rated prompts for areas like distribution and underwriting. LLMs are also being used to translate complex insurance contracts into computable specifications, enabling more efficient processing in areas like call centres and offering the potential for contract analysis and simplification. Additionally, AI, including computer vision and multimodal models, is being applied to automate claims processing (as seen with Hosta AI's spatial understanding of images), improve fraud detection, and optimise customer acquisition by intelligently understanding risk profiles and tailoring marketing efforts (as with Fai AI).

Looking ahead one year and then five years, how is AI expected to transform key insurance processes such as pricing, underwriting, and claims?

In the next year, AI will likely lead to more mature models for claims processing, focusing on faster and more effective data collection (potentially using telematics and computer vision), improved identification of severe claims (like those involving bodily injury), and strategies to reduce attorney representation. Over the next five years, pricing could become more dynamic, potentially using real-time data like telematics to price policies based on behaviour. AI's ability to better understand risk through image analysis (e.g., satellite data) could lead to more accurate and scalable risk assessment in underwriting. In claims, AI could facilitate faster settlements, particularly in catastrophe situations, by enabling quicker damage assessment through image analysis and data triggers.

How might AI impact the insurability of certain risks and geographies in the future, particularly considering factors like climate change?

AI's enhanced risk assessment capabilities may lead to the identification of areas or risks that become increasingly difficult or even impossible to insure due to escalating risks, such as flood damage in coastal regions. The ability to model and predict the impact of factors like climate change more accurately will force insurers to make difficult decisions about coverage in high-risk zones. This also raises questions about the role of government intervention to ensure continued insurability in affected areas.

With the increasing use of AI, how do you foresee the balance of risk shifting within the insurance industry?

AI introduces new types of risks. For example, with the advent of autonomous vehicles, the risk shifts from individual driver error to potential systemic risks arising from software bugs or widespread technological failures. Furthermore, the use of AI models in decision-making processes, such as fraud detection, carries the risk of bias against certain populations if the models are trained on biased historical data, leading to systemic risks that were not as apparent with individual human decision-making. The management of model risk itself, with the proliferation and integration of numerous AI models within an organisation, becomes a significant challenge.

What role will human capital play in the insurance industry as AI adoption grows? Will the skills and roles of actuaries, underwriters, and claims adjusters evolve?

While AI will automate many routine tasks, human expertise will remain crucial. There's a possibility of roles fusing, such as actuaries and data scientists becoming more integrated. Human talent will be needed for tasks that require empathy, complex decision-making beyond current AI capabilities, and overseeing and validating AI-driven processes. Moreover, dedicated internal teams focused on the success of AI initiatives and partnerships with startups are essential. The democratisation of AI tools may shift the focus towards product management and actuarial skills in defining and implementing AI solutions.

For established insurance companies looking to innovate with AI, what are the key considerations when deciding whether to build solutions in-house, partner with startups, or invest in them?

Key considerations include the speed of development (is it faster externally?), whether the solution is a commodity or offers potential for proprietary value, the cost of internal development versus external partnership, and the level of executive support for the initiative. A "build to spin-off" strategy is also emerging, where internal incubation leads to the creation of independent entities that can attract external capital. When partnering with startups, having a clear value-based use case and an executive sponsor are critical for successful engagement. Internal dedicated teams focused on the success of startup collaborations are also vital to ensure prioritisation and effective execution.

Startups often face challenges in accessing the vast amounts of data held by incumbent insurance carriers. What strategies can startups employ to effectively collaborate and demonstrate value to these larger organisations despite this data gap?

Startups can focus on developing very clear, value-driven point solutions in specific areas where they excel and secure executive sponsorship within the insurance carrier. Leveraging industry-wide data providers can help initially, but ultimately, demonstrating a compelling use case can facilitate access to carrier-specific data for pilots and further development. Startups also need to consider integration with existing insurance systems and workflows to make adoption easier for users. Forming strategic partnerships with tier-one or tier-two carriers can help drive necessary integrations and validate their solutions.

If you were to start a company focused on AI within the insurance industry today, what specific area or problem would you be most excited to tackle?

Potential startup areas include "insuring AI" itself, focusing on understanding and mitigating the risks associated with AI technologies and developing insurance products tailored to these risks. Another exciting area is shifting the core insurance model from simply covering losses to actively predicting and preventing risks by leveraging AI to analyse behaviour, identify potential hazards, and influence positive changes through personalised insights and interventions. Addressing the challenges of model risk management within insurance organisations, with the increasing complexity of interconnected AI models, also presents a significant opportunity.

What are the two main ways AXA is currently thinking about and deploying Large Language Models (LLMs)?

AXA is deploying LLMs in two primary ways: firstly, by providing secure access to an OpenAI backend for all employees as AI co-pilots to assist with smaller tasks, leveraging collective user intelligence through prompt sharing and feedback. Secondly, they are using LLMs to translate contract wording into specifications for their systems, creating a "computable contract" approach for more efficient processing, particularly for call centres.

According to Christopher Pette, what are some key areas in claims processing where AI can create value in the next year?

In the short term for claims, AI can create value by improving the collection of necessary information upfront and by building models that can more effectively identify claims with bodily injury and those likely to involve attorney representation. This allows for quicker and more appropriate handling of claims.

Henrietta Flashman suggests that in five years, what might happen to areas that are increasingly susceptible to risks like flooding? Why?

Henrietta Flashman suggests that in five years, some areas might become uninsurable altogether due to the increasing risks associated with climate change and events like flooding, as evidenced by insurers opting out of high-risk areas. This is because the frequency and severity of these events make it too costly to provide coverage.

How does Hank Frecon believe Generative AI can significantly change customer acquisition for insurance companies in the short term?

Hank Frecon believes that the introduction of autonomous agents for simple tasks can lead to massive efficiency gains in the short term. Additionally, Generative AI allows for the production of personalised and individualised content, making targeted outreach to specific audiences for acquisition much easier and faster.

Christopher Pette describes Allstate's multi-channel approach. How does AI help them determine where a potential customer belongs within these channels?

AI helps Allstate intelligently route potential customers across their direct online/mobile, captive call centre, exclusive agent, and independent agent channels by analysing various factors to understand the risk profile and the most suitable channel and products for that individual. This ensures a more tailored and efficient customer experience.

Marcin Piwowarski explains AXA's new approach to internal innovation. What is unique about their current thinking regarding the incubation and future of internally developed AI solutions?

AXA's new approach involves incubating use cases by matching technology to business needs and then spinning them off relatively early to be funded by external capital, allowing for faster scaling. This is a shift from solely building internally, recognising the competitive capital landscape.

According to Christopher Pette, what are the two critical factors for a startup to work successfully with a large insurer like Allstate?

The two critical factors for a startup to work successfully with Allstate are having a very clear value-based use case that they execute exceptionally well in a specific area, and having an executive within Allstate who is willing to sponsor their efforts.

Henrietta Flashman discusses the challenges startups face when trying to integrate their solutions within the insurance industry. What is one of the biggest hurdles she identifies?

One of the biggest hurdles Henrietta Flashman identifies for startups is the difficulty in achieving data integration with established insurance carriers and other key industry players like pricing information providers. This lack of seamless integration can hinder the value and efficiency of the startup's solution.

Marcin Piwowarski describes AXA's platform approach to integrating startups. How does this strategy aim to address the challenge of startups solving only small parts of a larger problem?

AXA's platform approach involves building integrated platforms for specific lines of business (like health and commercial PNC) that connect data and coordinate various actors, including startups. This aims to address the issue of point solutions by creating an ecosystem where different specialised AI applications can work together to solve broader challenges.

If Marcin Piwowarski were to start a company today, what area related to AI would he focus on, and why does he find it exciting?

If Marcin Piwowarski were to start a company, he would focus on insuring AI itself. He finds this exciting because it connects to his research interests in AI risk, transparency, and fairness, and explores how to provide coverage for the unique risks associated with artificial intelligence.

Discuss the transformative potential of Large Language Models (LLMs) within the insurance industry, considering both the immediate applications and the more significant, long-term changes they might bring to areas like contract management.

LLMs have the potential to revolutionise contract management by translating complex legal jargon into computable formats, allowing for automated processing and analysis. In the long term, this could lead to more standardised and transparent contracts, reducing disputes and enhancing customer trust. Furthermore, LLMs can serve as AI co-pilots, assisting employees in decision-making and improving operational efficiency across various functions.

Analyse the impact of Artificial Intelligence on the core functions of insurance: pricing, underwriting, and claims processing. Consider the perspectives of both established insurers and emerging startups in your discussion, and explore the potential shifts in risk assessment.

AI is reshaping the core functions of insurance by enabling dynamic pricing, where premiums are adjusted in real-time based on individual risk profiles. In underwriting, AI enhances risk assessment through data-driven insights, improving accuracy and scalability. Claims processing benefits from automation and faster settlements, particularly in high-volume scenarios. Startups bring agility and innovation, often focusing on niche solutions, while established insurers leverage AI for broad-scale efficiency improvements.

Evaluate the challenges and opportunities for collaboration between established insurance enterprises and AI-focused startups. What are the key factors that contribute to successful partnerships, and how can both types of organisations benefit from these collaborations?

Successful collaboration requires clear communication, alignment of goals, and mutual trust. Startups offer innovative solutions and agility, while established insurers provide market access and resources. Key factors include having a well-defined use case, executive sponsorship, and seamless integration with existing systems. Both parties benefit from shared knowledge, accelerated innovation, and improved market competitiveness.

Explore the evolving landscape of risk in the insurance industry as a result of the increasing adoption of AI. Consider both traditional areas of risk and the emergence of new, AI-specific risks, such as model bias and systemic failures in autonomous systems.

AI adoption introduces new risks, such as model bias and systemic vulnerabilities, which can lead to unfair outcomes and widespread disruptions. Traditional risks, like fraud and underwriting errors, are being mitigated through AI-driven solutions. Insurers must balance leveraging AI's capabilities with managing these new risks by implementing robust validation processes and ensuring transparency and fairness in AI models.

Discuss the future of talent within the insurance industry in the age of AI. How might the roles of actuaries, data scientists, underwriters, and other professionals evolve, and what skills will be most critical for success in this changing environment?

The roles of insurance professionals are evolving towards more strategic and analytical functions. Actuaries and underwriters will need to develop data science and AI proficiency to leverage advanced analytics tools. Skills in machine learning, data interpretation, and AI ethics will become increasingly important. Human expertise in empathy, complex decision-making, and oversight of AI systems will remain critical to complement automated processes.

What are some common challenges faced by insurance companies when integrating AI into their operations, and how can they overcome them?

Common challenges include data privacy concerns, integration with legacy systems, and cultural resistance. To overcome these, insurers should prioritise data governance, invest in modernising IT infrastructure, and foster a culture of innovation through training and change management initiatives. Engaging stakeholders early and demonstrating AI's value through pilot projects can also facilitate smoother integration.

What are some practical applications of AI in the insurance industry that are currently being explored or implemented?

Practical applications of AI in insurance include automated claims processing, where AI analyses images and data to expedite settlements. Fraud detection is enhanced through pattern recognition and anomaly detection. Customer service is improved with AI chatbots providing 24/7 support. Predictive analytics in underwriting and pricing allows for more accurate risk assessments and tailored policy offerings.

How is AI driving personalization in the insurance industry, and what benefits does this bring to both insurers and customers?

AI enables insurers to offer personalised products and services by analysing customer data, such as behaviour and preferences. This allows for tailored policy recommendations and pricing, enhancing customer satisfaction and loyalty. For insurers, personalisation leads to improved customer retention, better risk management, and increased profitability through targeted marketing efforts.

How are regulatory compliance and ethical considerations being addressed in the context of AI adoption in insurance?

Regulatory compliance and ethical considerations are critical in AI adoption. Insurers must ensure transparency, fairness, and accountability in AI systems. This involves adhering to data protection regulations, such as GDPR, and implementing bias detection and mitigation strategies. Collaborating with regulators and industry bodies to develop guidelines and standards is essential for maintaining trust and promoting responsible AI use.

In what ways is AI enhancing the customer experience in the insurance industry?

AI enhances the customer experience by providing faster, more efficient service. Chatbots and virtual assistants offer immediate support and guidance, while AI-driven insights enable personalised interactions and product offerings. Automated processes reduce wait times for claims and policy changes, improving overall customer satisfaction. AI also empowers customers with self-service options, allowing for greater control and convenience.

How is AI improving fraud detection and prevention in the insurance industry?

AI improves fraud detection by analysing large volumes of data to identify patterns and anomalies that may indicate fraudulent activity. Machine learning algorithms can learn from historical data to predict and flag potential fraud cases, enabling insurers to respond proactively. This reduces false claims, saves costs, and enhances the overall integrity of the insurance system.

What role does AI play in enhancing risk assessment and management for insurers?

AI enhances risk assessment by providing more accurate and comprehensive analyses of data, including real-time information from telematics and IoT devices. This allows insurers to better understand and predict risks, leading to more informed underwriting decisions and pricing strategies. AI also supports proactive risk management by identifying potential threats and suggesting preventive measures.

How is telematics being used in conjunction with AI to improve insurance offerings?

Telematics, combined with AI, allows insurers to collect and analyse data on driving behaviour, enabling usage-based insurance models. This provides more accurate risk assessments and personalised pricing based on individual driving habits. Insurers can offer incentives for safe driving, enhancing customer engagement and reducing accident-related claims.

How does AI contribute to improving efficiency in the underwriting process?

AI streamlines the underwriting process by automating data collection and analysis, reducing the time and effort required for risk assessment. Machine learning models can evaluate complex datasets to identify risk factors and recommend appropriate coverage options. This leads to faster decision-making, improved accuracy, and more competitive pricing for customers.

In what ways is AI being used to manage and respond to catastrophic events in the insurance industry?

AI aids in catastrophe management by providing rapid damage assessment through satellite imagery and drone footage, enabling quicker claims processing and resource allocation. Predictive models help insurers anticipate and prepare for extreme weather events, reducing potential losses. AI also supports communication and coordination efforts during disasters, ensuring timely assistance to affected policyholders.