Generative AI is here. Are you ready?

A guide to operationalizing Generative AI in insurance

WHITEPAPER GenAI is here are you ready? A guide to operationalizing GenAI in insurance

Executive summary GenAI is the world’s newest shiny toy. And if the hype is to be believed (which it is), it’s set to create opportunities for innovation, improve operational efficiency, and enhance customer experiences. The challenge right now, however, is that many professionals have only scratched the surface of what GenAI can do and are still unsure of how to leverage it across their organizations. This is especially true in the insurance industry. Insurers rely heavily on document-intensive processes, and GenAI presents an unprecedented opportunity to drive enhanced results. By embracing these tools strategically, the insurance industry can unlock new levels of efficiency, elevate customer experiences, and increase profitability while effectively managing the inherent risks and pitfalls associated with this revolutionary technology. As GenAI continues to advance with more accurate open-source Large Language Models (LLMs), understanding how to integrate this technology effectively is becoming ever more crucial for staying competitive and gaining an advantage. In this white paper, we explore the operationalization of GenAI within the insurance industry, providing an overview of the operating model you’ll need for scalability and compliance. We also outline the process of implementing GenAI across the value chain of an insurance enterprise.

No doubt, the hype around ChatGPT has raised awareness of the potential of AI. Rather than replacing people, it enables insurance professionals to provide a better customer experience.

How insurers are using GenAI Insurers are currently in the understanding and experimentation phase of GenAI use. Two of their biggest concerns are data security and hallucination (a tendency to create outputs that seem correct when compared to factual data but are simply made up). While open-source LLMs can drastically reduce data security risks, GenAI’s ability to confidently provide incorrect or biased outputs must be mitigated through validation before this data is used to take informed decisions. Even so, in its current form, GenAI currently works well as a knowledge assistant to insurance professionals across the value chain and can help to improve efficiencies internally (i.e., in low-risk, high-impact use cases). For instance, Zurich is currently conducting experiments involving applications that aim to extract data from claims descriptions and various other documents. To enhance its underwriting process, the company is inputting the most recent six years' worth of claims data. In another example from RGA Data Science, VP Jeff Heaton has said, “RGA is currently in the early stages of evaluating ChatGPT primarily to assist insurance professionals with routine tasks.”

Although insurance providers might not experience an instantaneous impact on their financial performance or customer interaction, a phased value measurement framework assists us in charting the path toward profitable expansion.

The framework below (Fig. 1) is helpful for measuring value in implementing GenAI within the insurance and helps us evaluate the advantages it delivers to customers and the business alike.

GenAI for enhanced outcomes, while protecting the downside

VALUE TO CUSTOMERS

Parallelly, key risks need to be guarded against/ mitigated

Speed to outcome

Convenience, ease of business

Superior experience

• Faster claims

• Intuitive process • Information availability • Reduced consumer • complaints, load on customer service.

• Increased trust index • Higher Member NPS

VALUE TO ENTERPRISE

Risk Mitigation • Data privacy & security • Compliance/ legal hassles • Reputation risk

Day-to-day efficiencies

Process Redesign

Better financials • Combined ratio • Lower loss ratio (More accurate risk assessment )

• Standardization • Improved humans accuracy • Lower Attrition

• Higher productivity (# Quotes/UW,

reduction (Expense cost)

#Claims/Adjustor, etc.)

• Reduced time load

Fig 1. Value assessment framework

Setting up an operating model for scaled success A well-defined and scalable operating model is essential to successfully integrate GenAI into insurance operations. The model below (Fig. 2) outlines how the collaboration between the GenAI Center of Excellence (CoE) , GenAI core team, and GenAI governance council ensures consistency and efficiency while maintaining strong governance. This operating model focuses on three key areas, namely:

1. Foundation:

Assessing the enterprise's readiness for GenAI initiatives and establishing the necessary infrastructure.

2. Operational:

Offering oversight and strategic direction for GenAI projects, which involves using expansive language models and agreement on GenAI tools and technologies. The CoE and the core team foster collaboration among data science teams, helping to spread the technology more widely and implement solutions efficiently.

3. Functional:

Aligning GenAI efforts with organizational goals, assessing risks, and upholding responsible AI standards.

This trio of teams collaborates with the legal/compliance department to ensure compliance with existing regulations. As the GenAI initiatives mature to the level of other AI technologies, the governance and CoE for GenAI can be integrated with the broader AI governance and CoE. Setting up for success GenAI governance

• Head of business functions (underwriting, claims marketing) • Head of the compliance team

Governance Objectives :

GAI Governance Council

• • •

Alignment with organizational goals Enterprise risk assessment Alignment with ethical standards

• Head of analytics • GenAI evangelists • NLP/NLG SMEs • Business SMEs (from different business teams)

• • • •

Oversight of GenAI initiatives Establishing the use of LLM models Alignment on GenAI tools and technologies Alignment between AI/ML/data science teams across the organization to democratize

GAI Center Of Excellence

•

Figuring out deliberate zones

• •

Set up the required infrastructure Set up the GAI data science team, COE & Governance team

• Head of Analytics • GenAI experts • NLP/NLG data scientists • Data engineers • Designers

GAI Core team

Fig 2. Operating model

Operationalizing GenAI The potential of GenAI has become increasingly evident, calling for a comprehensive and structured approach to adoption. To fully leverage the potential of GenAI, businesses must first identify their objectives and determine how GenAI can provide improvements. This involves identifying relevant functions to kick-start training and adoption of GenAI, as well as prioritizing use cases using the 4E framework, which consists of: Economics: The strategy’s financial aspect involves analyzing and understanding the potential return on investment and profitability. Exposure: The level of vulnerability an organization has to reputation risk and compliance issues. Edge: The competitive advantage that the tech provides, which involves identifying and leveraging unique selling propositions or critical differentiators. Extensibility: The ability of a system to be easily extended or modified without the need for significant changes to its underlying structure, allowing for adding new features, functionalities, or modules to an existing system without disrupting its core functionality.

Once the use cases have been identified, businesses must select the appropriate GenAI tools and models for implementation, embed the GenAI team across functions, and provide training on prompt engineering. Finally, businesses must define value measurement metrics, set up monitoring tools, review, and audit, and collect feedback to improve continuously. How do we unlock full potential of GenAI to deliver across insurance enterprise?

Incubation and Identification

Monitoring and adoption

Impact Generation

Tool integration

Orchestration

• Identifying business objectives where GenAI can provide improvements • Identify relevant functions to kick start training and adoption of GenAI • Use cases prioritization leveraging 4E (Economics, Exposure, Edge and Extensibility) framework: • Economics -> Revenue, Productivity, Costs • Exposure -> Reputation risk, compliance risk, Privacy & Security • Edge -> Cool factor, Brand impact, Novelty • Extensibility -> Concept and tech extension

• Selection of GAI

• Infrastructure check: (Open AI, Azure Open AI, Open source LLMs ) • Maturity of LLMs from a commercial standpoint - Licensing, Vendor selection, Compliance check • Customized inferencing

• Define metrics such as the number of AI models deployed, the amount of data processed by AI models • Set up monitoring tools to track and collect data on the defined metrics • Review and audit metrics • Conduct audits of the models to ensure that they are being used ethically and effectively • Collect feedback to continuously improve

• Value to

Customers

tools and models for implementation – API integration and orchestration implement GenAI tools for UW, Claims, etc. owners to embed use of tools

• Value to

Enterprise

• Embed GenAI team across functions to

• Functional

using client data

• Transfer learning/

• Training on Prompt engineering • Monitoring,

Fine-tuning of models based on custom data

implementing data privacy measures

GAI Governance Council + Ethics Committee

GenAI Centre of Excellence GenAI Core team

Fig 3. Operating model

Use cases From underwriting and claims management to customer service and risk assessment, GenAI has the potential to transform a wide selection of facets within the insurance industry. In Fig. 4, we have mapped multiple use cases of GenAI. Its potential impact on the insurance value chain is far-reaching, enabling insurers to streamline operations, mitigate risks, optimize pricing, and deliver superior value to policyholders and enterprises.

Functions

Sales & Distribution

Risk Management • Identifying at risk customers for proactive outreach and support.

Asset Management

Marketing

UW & Pricing

Claims

 Draft policy documents.  Synthetic data for better risk assessment.  KYC & data validation using public, 3PD & policy data.  Digitized policy buying experience.  24x7 knowledge assistant for UWs (Regulator changes, public info on customers, initial risk assessment, etc)

 Personlized targeting & campaign content generation.  Educating & creating.

 Interactions & guidance chatbot for agents.  Cross-sell and Upsell insurance policies with GenAI-assisted agents.  Empathetic AI-powered chatbots to improve customer experience. • Reduce digital friction by customer sentiment analysis on social media.

• 24x7 knowledge assistant to adjustors-doc summarization, letter of intent generation, etc

• Optimizing investment portfolios by analyzing different combinations of assets and identifying the most efficient portfolio structure. • Forecast market trends, identify investment opportunities. • Real-time monitoring of investment portfolios.

• Detecting & addressing bias in

• Detect fraud

using synthetic data.

insurance practices to ensure equitable emotional treatment of all customers.

 Improved customer

 Improved efficiency  Reduced

• Increased customer retention • Reduced claim cost

 Accurate severity estimation & reserving  Faster claims processing  Lower claims severity  Reduced LAE

 Higher response rate  Higher customer engagement

 Higher return on investments

experience

 More

expense cost  Increased risk assessment accuracy

productive & effective agents

Fig 4: Use cases

Example: GenAI in an underwriter’s journey The integration of GenAI into the underwriting process delivers automation, efficiency, accuracy, and improved decision-making, transforming the traditional journey of underwriters by allowing them to concentrate on intricate cases, consistently follow guidelines, optimize operations, manage risks, and enhance customer experience, all while delivering superior underwriting services. Figures 5 and 6 provide a comparative analysis of an underwriter's journey within a specific line of business, in this instance, the small commercial property sector, using the example of a city-based café chain's insurance application.

Typical UW journey for small commercial property UW Journey

Underwriter logs in to the Underwriting Portal

Sees a list of quotes which have been declined, accepted and referred for Underwriter’s review

The ones that have been declined just have a one liner explanation as the reason of decline - no further details provided

Same scenario holds true for the auto accepted ones - plus, there is no documentary of how the risk impacts the overall portfolio performance of the BOB

For the referred quotes, no priority order exists as well as specifics to be checked for each quote

Sees a list of quotes which have been declined, accepted and referred for Underwriter’s review

3 Underwriter picks up a sample of declined & accepted quotes and does manual reviews to validate the decisions-notices a few gaps in some of the quotes Underwriter reviews all the submitted documents regarding property location, building type, age, construction materials, other occupancies details, fire safety certificates, etc. 5

Determine if they meet the insurers risk appetite & underwriting guidelines.

Keys in pertinent information on the pricing tool to get premium & coverage limit/ deductible guidance

In case of quote acceptance, drafts policy papers & sends directly to insured or relevant stakeholders in the organisation to take it forward.

Logs into the portfolio management system to check how the risks impact the overall BOB performance

Documents the underwriting decision made for each policy & maintain a record of the policy details.

Pain Points

Documents are currently unstructured - Incomplete data Unstructured quote requests - from agents/ brokers via emails Extraction using OCRs -*** Manual ways

Multiple Data sources - needs for data harmonisation

Misrepresentation of data - Fraud detection Not able to understand fraud behaviour

Which UW to assign the quote to?

Fraud detection - Detecting the right KYC is a challenge - streamlining

Risk assessment - needs assistance Not integrated - not consolidation - easy choose & pick for tailored products

Decisions are not data driven rather geography - What is driving this decision?

Ingestion period - incompleteness of data - agent / brokers - manual to email

Non accuracy

Manual assessment of regulatory documents

Pricing models are not accurate - need to be explainable

Separate model for coverage, premium offering

Portfolio optimisation

Feedback loop to Gen Al Initial policy draft takes time

Binding is not real time

Lack of empathy during rejection

Fig 5. Process without GenAI help and STP workflows in place

GenAI enabling underwriters to make better decisions with higher accuracy at speed

Gen Al assistant talks to customer/agents - fetches all information

Creation of knowledge Assistant Bots

Extracts relevant data to create documents in a structured format

GenAI

Ai Flucin

Premium estimation

UW guidelines

Other 3rd party data

Public data scraping

GenAI integrated with risk assessment model GenAI integrates with other existing solutions to accept/reject/refer quotes with a priority order to the UW GenAI interacts with multiple tools in silo to do initial risk assessment and quote generation for simple tasks

GenAI integrates with other existing solutions to prioritize the list of quotes

GenAI

Sees a list of prioritised quotes which have been declined, accepted & referred for review

Underwriter logs into the Underwriting portal

UW chats with Gen Al to understand reasons of referral & impact on overall BOB

UW sees One Stop View of all possible risk attributes: An integrated view of summary documents, insurers risk appetite, Extract premium limit & deductible guidance, information extracted from aerial image of property (overhanging, tree, square footage, etc.), etc.

Picks highest priority quote sent for UW referral

UW picks accepted/ rejected quotes from prioritised list provided by GenAI assistant

GenAI

Premium estimation UW guidelines Other 3rd party data Public data scraping

Gen Al evaluates inputs on errors & documents it

UW asks GEN Al about auto-decision making for D/A/R? - If any discrepancy, asks to document it & learn from errors

GenAI

Gen Al interacts with multiple tools in silo to help UWs do better & faster risk assessment

GenAI

Gen Al integrated with risk assessment model

Gen Al creates a policy draft in case of a complex policy

Gen Al will fetch policy from automated system case of standard policy

In case of quote acceptance, asks GenAI to draft policy papers and Validate by UW before sending it to insured / stakeholders

GenAI

Policy draft

Gen Al documents each policy decision & maintains a record of it

GenAI

GenAI Team

Fig 6. Process without GenAI help and STP workflows in place

Using the power of GenAI, the underwriter journey undergoes a transformation: data analysis is streamlined, processes are automated, and risk assessment is enhanced, leading to more informed decision-making. GenAI empowers underwriters to handle complex commercial risks efficiently, optimize pricing, and deliver superior underwriting services in the competitive commercial insurance market.

Conclusion GenAI presents immense potential for the insurance sector, opening avenues for creative solutions, streamlined operations, and improved client interactions. Insurers aiming to harness these advantages should adopt a well-rounded, systematic strategy. This includes establishing a scalable operating model, aligning initiatives with business objectives, and choosing the most relevant use cases for integration. Essential aspects of successful implementation include the judicious selection of tools, training in prompt engineering, and an ongoing commitment to improvement via monitoring and feedback. By wholeheartedly adopting GenAI, insurers can maintain a competitive edge, foster profitability, and deliver superior customer experiences in the ever-evolving digital landscape.

Author

Rashid Khan Engagement Manager, Consulting Insurance

Angkeeta Goswami Senior Design Consultant, Strategic Center

About Fractal

Fractal is one of the most prominent providers of Artificial Intelligence to Fortune 500® companies. Fractal's vision is to power every human decision in the enterprise, and bring AI, engineering, and design to help the world's most admired companies. Fractal's businesses include Crux Intelligence (AI driven business intelligence), Eugenie.ai (AI for sustainability), Asper.ai (AI for revenue growth management) and Senseforth.ai (conversational AI for sales and customer service). Fractal incubated Qure.ai, a leading player in healthcare AI for detecting Tuberculosis and Lung cancer. Fractal currently has 4000+ employees across 16 global locations, including the United States, UK, Ukraine, India, Singapore, and Australia. Fractal has been recognized as 'Great Workplace' and 'India's Best Workplaces for Women' in the top 100 (large) category by The Great Place to Work® Institute; featured as a leader in Customer Analytics Service Providers Wave™ 2021, Computer Vision Consultancies Wave™ 2020 & Specialized Insights Service Providers Wave™ 2020 by Forrester Research Inc., a leader in Analytics & AI Services Specialists Peak Matrix 2022 by Everest Group and recognized as an 'Honorable Vendor' in 2022 Magic Quadrant™ for data & analytics by Gartner Inc.

For more information, visit fractal.ai

Corporate Headquarters Suite 76J, One World Trade Center, New York, NY 10007

Get in touch

Page 1 Page 2 Page 3 Page 4 Page 5 Page 6 Page 7 Page 8 Page 9 Page 10 Page 11

Made with FlippingBook - PDF hosting