American Family Insurance Group, founded in 1927, began as Farmers Mutual Insurance Company in Madison, Wisconsin, targeting farmers with auto insurance. Over the decades, it expanded its offerings and changed its name to American Family Mutual Insurance Company in 1963. As per the financials published by the company, it made $17.1 billion in revenue in 2023, up from $14.4 billion in 2022.
In 2017, the American Family Insurance Group acquired Networked Insights to bolster its data analytics and AI capabilities, aiming to improve customer interactions and operational efficiency. In a recent podcast with Emerj, Ryann Foelker, Strategy Director at American Family Insurance Group, shared insights on designing impactful AI use cases. Foelker explained that her company is focusing on targeting persistent problems that lack effective AI solutions.
This article explores two compelling use cases that illustrate how Family Insurance Group’s AI initiatives are actively supporting its strategic business objectives:
- Streamlining subrogation with computer vision and machine learning: Automating subrogation claims using computer vision for damage assessment and machine learning to identify recovery opportunities, boosting speed and accuracy.
- Optimizing policy updates and business rules with machine learning: Using machine learning to automate business logic extraction and centralize decision rules, enhancing product delivery speed and minimizing IT dependence.
Streamlining Subrogation with Computer Vision and Machine Learning
The Washington State Office of Insurance Commissioner describes subrogation in insurance as the legal process where an insurer, after paying a claim to its policyholder, seeks reimbursement from the party responsible for the loss. This mechanism allows the insurer to recover costs.
However, the traditional subrogation process poses challenges for insurance companies. As per Tata Consultancy Services (TCS), these challenges include inefficient operational models and a heavy reliance on manual processes, which ultimately lead to missed recovery opportunities and human biases in decision-making. Insurers often need help with competing priorities that distract adjusters from focusing on subrogation tasks, resulting in errors and delays.
The TCS Source also highlights that AI and machine learning help address these challenges by automating routine tasks, identifying subrogation opportunities throughout the claims lifecycle, and intelligently triaging cases to the fitting adjusters at the right time. It enhances efficiency, reduces leakage in recoveries, and ultimately improves the overall claims process for insurers.
Per a press release published by Tractable, the insurance group leveraged Tractable’s AI subrogation solution across all of American Family Insurance Claims Services (AFICS) with the objective of improving speed, efficiency, and accuracy. In 2021, American Family Insurance Group launched an Innovation Lab to explore new technologies that could enhance its claims processes. According to the aforementioned press release, the innovation lab initiative led to the collaboration with Tractable, focusing on integrating AI into their operations.
Tractable AI is a technology company that develops AI solutions for assessing damage to property and vehicles, utilizing computer vision and deep learning. The press release also cited the objectives of this partnership, which include:
- Increased Efficiency: Automating the review process for subrogation claims to reduce processing times.
- Enhanced Accuracy: Utilizing AI to provide consistent evaluations of claims based on photographic evidence.
- Improved Customer Experience: Allowing claims adjusters more time to focus on customer interactions rather than administrative tasks.
The same press release also stated that American Family Insurance Group leveraged Tractable’s AI solution, AI Subro solution, to its then-inbound subrogation operations.
Per the company’s website, Tractable’s AI Subro solution is for the subrogation process and is a part of Tractable Auto Reviewer product reviewer, and it automated the third-party claim settlement.
According to the company’s promotional content, this is how this AI tool works:
- Data Input: The process begins when a claim is submitted, including relevant documentation such as photographs of the damaged vehicle or property.
- Image Analysis: Tractable’s AI algorithms analyze the submitted images using computer vision techniques to assess the extent of the damage. This analysis identifies specific areas of concern and estimates repair costs based on historical data.
- Automated Evaluation: The tool automatically evaluates the claim against predefined standards and guidelines set by the insurance company. It determines whether the claim meets the criteria for subrogation and calculates potential payouts.
- Recommendation Generation: Based on the analysis, the system generates a comprehensive report that includes recommended actions, estimated costs, and any necessary documentation to support the subrogation process.
- Claims Handling: The claims team reviews the AI-generated recommendations and insights, allowing them to make informed decisions quickly. This streamlined approach reduces processing time and enhances overall efficiency in handling subrogation claims.
Screenshot from Tractable Subro. (Source: Tractable)
Tractable claims that its AI-powered tool only takes 15 seconds to assess the claim and file the report.
While the American Family Insurance Group has yet to publish any results of this partnership publicly, there are worthwhile conclusions to extrapolate from looking at its financials. Per the company’s website, it saw strong customer satisfaction and retention rates, with 14 million policies in force at year-end, reflecting a 3.8% increase from 2022. The use of AI tools like Tractable can enhance customer service by speeding up claims processing, thereby improving customer experiences and loyalty.
Additionally, American Family Insurance Group reported a revenue increase to $17.1 billion in 2023, up from $14.4 billion in 2022, driven by premium growth and investment income. The efficiency gained through AI in the claims process could contribute to this revenue growth by enabling faster claims processing and potentially increasing customer retention.
Optimizing Policy Updates and Business Rules with Machine Learning
The insurance industry faces numerous challenges in updating policies, not the least of which is the compounding effect of climate change over time.
In research from Deloitte co-authored by ‘AI in Business’ podcast guest Karl Hersch, global economic losses from natural catastrophes totaled US$357 billion in 2023. However, only 35% of these losses were insured, resulting in a protection gap of 65%, or US$234 billion. This gap is especially significant in regions such as the Middle East, Africa, and Asia.
Per the press release issued by Sapiens Americas, the American Family Insurance Group adopted decision management software to improve product offerings and ultimately deliver better value.
According to the company website, Sapiens provides a decision automation platform to enable businesses to translate policy into code and manage complex business rules. Its decision platform for insurers automates and streamlines decision-making processes, allowing organizations to manage business rules centrally, enhance operational efficiency, and accelerate product delivery without heavy reliance on IT resources.
The company website also indicates Sapiens Decision extracts the logic in the policy administration system to be managed centrally with easy-to-use tools. It claims that there is no need for businesses to rip and replace their policy administration systems or rely on IT for policy updates.
Per the product documentation for the Sapiens Decision Management Platform, the software operates through a structured process that automates and manages insurers’ business logic. It begins with Automated Logic Extraction, decision manager, and, lastly, decision execution.
- Automated Logic Extraction: It uses machine learning to extract existing business logic from legacy systems and convert it into technology-independent decision models. This step claims to eliminate the complexities and risks associated with outdated code.
- Decision Manager: It allows business analysts to model, validate, and test these decision rules using a no-code visual interface, making it accessible without technical expertise. It ensures clarity and consistency in the rules that govern operations.
- Decision Execution: It offers flexible integration options with existing systems and full traceability of decisions made. This approach enables organizations to optimize their decision-making processes, reduce operational risks, and adapt to future technological changes seamlessly.
Sapiens provides a tour of the product through this digital interface available through their website:
Screenshot of the Sapiens Decision Management Platform tour interface. (Source: Sapiens)
While neither Sapiens nor the American Family Insurance Group has published any specific results from the above partnership, Sapiens published the below results for one of their customers, Hiscox:
- Increase in customer satisfaction scores from 83% (Oct 2019) to 93% (September 2020)
- Increase of Net Promoter Score (NPS) from 68.3 (October 2019) to 82.0 (September 2020
- 90% of changes can be developed and deployed with zero IT involvement.
Based on the financial results of American Family Insurance Group for 2023, it is evident that the company faced significant challenges, including a record $3.5 billion in catastrophe claims and a net underwriting loss of $1.7 billion. Although AFI mentioned efforts to manage expenses, their combined ratio of 110.8% shows they are spending more on claims and expenses than they earn in premiums.
Tracking these expenses indicates that while they are trying to control costs and improve claims management, the financial results suggest they have yet to succeed. Therefore, even though using decision management software like Sapiens could help improve their processes, the current data does not clearly show that these improvements have been achieved so far.
