The Customer

A leading insurance carrier and Fortune 500 company previously partnered with the Roots' team to introduce automation into their claims department. The previous AI Agents functioned so successfully that the executive team began looking for automation expansion opportunities in other departments. For this partnership, the team decided to focus on the premium audit process within their underwriting department.

This property & casualty carrier is U.S. based, focusing on auto insurance and workers compensation, with approximately $3 billion in revenue. They have roughly 1,000 employees, many of whom work remotely.

The Situation

The Customer’s premium audit team faced significant pressure to effectively and efficiently complete premium assessments. This pressure intensified during audit season as the number of assessments typically doubled Month on Month.

During audit season, the Customer would spend a lot of time hiring and training temporary staff and the incumbent team would need to work heavy overtime to deal with the additional workload. Finally, the work itself required deep concentration and attention to detail – which was often affected by the long hours, resulting in errors – particularly in areas that involved heavy data entry.

In addition to overwhelming the full-time team with overtime, the company would be fined if audits were not completed within the regulatory period. The Customer decided to pursue an automation solutions to alleviate the workload and the danger of acquired fines.

The Challenge

The Customer was challenged by the significant seasonal cost burden, as a result of necessary overtime payroll as well as the need to oversubscribe the team to complete the workload. The heavy reliance on temporary staff increased the pressure on the existing team to effectively and efficiently source, onboard and train new team members. Finally, the Customer faced potential fines for incomplete, incorrect audits or failure to meet deadlines.

As 2021 rolled around, the Customer was under significant pressure to complete double the number of audits as compared to the year prior. Additionally, the expectation was that the premium audit team would not make any mistakes on premium assessments. Relying exclusively on a human team would have increased the need to significantly increase spend for hiring and training temporary staff.

The Solution

Roots worked closely with the Customer to build an ‘Underwriting Assistant" AI Agent – which the Customer’s team voted to name Stuart. Stuart leverages an advanced Machine Learning model to differentiate between 7 different data sources and the unstructured payroll data.

We started by trying to learn from the customer, with the intention of understanding the requirements for this process. Then, we collaborated on the future state process and established team protocol, to encourage efficiency as the team works alongside the AI Agents.

In addition to the Machine Learning, we had to work through systems integrations as the unstructured data flowed in from 7 different sources. Typically, it would be desirable to start a partnership like this one with somewhere between 5,000 to 10,000 datapoints. However, the Customer only had about 500 datapoints, despite all the data and information from the audits provided. Datapoints are very important for Machine learning models to be accurate. To circumvent this, the Roots' team employed its creativity with the existing samples and, with assumptions, this obstacle was overcome. 

Today, Stuart leverages Document Vision, Process Automation & the advanced Machine Learning to review financial reports, validate employee/vehicle classifications, analyze payroll, redact PII, and prepare “pre-audit” assessments for the human team. The human team is then able to rapidly pick-up work from Stuart to then continue with their full assessment.

Following the successful implementation and adoption of Stuart, turnaround times for the premium assessments improved by 25%. Additionally, Stuart completed 25% of the audit team’s work in 2021, and expected to double to 50% in the 2022 audit season. This output enabled the Customer’s team to complete all premium audits with high quality results without having to increase their headcount or payroll. Furthermore, the team morale improved when both administrative data entry work and required overtime were eliminated.

The AI Agent was deployed within 8 months, with an ROI of 113% so far. After deployment, the Roots Automation team and the Customer immediately began to discuss potential enhancements and scaling for the coming year, with an emphasis on expanding out into new areas of the team and the business.