Data Extraction for Better Risk Insights

Intelligent Document Automation
NLP
Data Science
Insurance
CASE STUDY
CLIENT | ZURICH
INDUSTRY | INSURANCE
DATE | 2023

A Solution for a Leading Insurance Company
THE CHALLENGE
In today's rapidly changing insurance landscape, effective risk management is essential for organizations to thrive and remain competitive. However, many risk engineering teams face significant challenges in assessing and prioritizing risks due to the lack of a holistic view across their portfolio of risk reports. This limitation compromises their ability to identify and analyze interconnected risks, resulting in suboptimal decisions and inefficient risk management.
Zurich Insurance Portugal, a leading multi-line insurer serving customers in global and local markets, was aware of this challenge and decided to overcome it by improving their risk assessment and portfolio management. To address this matter, Zurich partnered with Two Impulse for the development of a document automation solution that would provide comprehensive risk insights while freeing up their risk engineering teams to focus on value-enhancing tasks.

Harnessing the Power of NLP and Document Automation
THE SOLUTION
Over several months, our team of Data Scientists worked alongside the Pricing team from Zurich to develop and implement a dedicated artificial intelligence solution to automate the risk engineering processes, making sure the best technology was applied and fine-tuned for each particular case.
In a nutshell, Two Impulse developed a Natural Language Processing (NLP) solution that automatically extracts entities from risk and claims reports, and classifies documents according to their type.
This was the document automation process flow:
- Identify the providers and types of reports to be targeted
- Load and pre-process PDF reports
- Extract the information split by topics
- Save it into a format to be digested by a data warehouse.
In particular cases, through NLP technology, we are able to detect if specific predetermined words were present in the documents and then classify the respective sentences.
Obviously, creating a solution able to work with all types of reports is quite a challenge. Therefore, in order to achieve the intended data extraction results, the team decided to assess each specific report type and tweak the code accordingly.

Enhancing Efficiency with Document Data Extraction
THE RESULTS
Through Two Impulse’s AI-driven data extraction process, Zurich Insurance Portugal was able to capture risk insights across their portfolio of reports in a structured manner.
- Portuguese language documents (majority of all claims and reports) were automated with an impressive 87% success rate.
- English language documents achieved a 73% success rate of relevant information extracted (smaller number of documents), having also successfully extracted the tables in 99% of cases.
Furthermore, by automating the data extraction process, the risk engineering teams were alleviated from time-consuming manual tasks and could focus on high-value activities instead.
Empowering Zurich Risk Management
In conclusion, our collaboration with Zurich Insurance Portugal for the development of a solution for data extraction and process automation has yielded clear benefits:
- Improved efficiency of risk management practices
- Risk engineering teams are now able to focus on high-value activities
- Better insights for making informed decisions and mitigating potential threats
Through document automation and NLP, Two Impulse has helped Zurich Insurance Portugal to leverage the power of AI by transforming risk engineering processes and embracing a more proactive approach to risk management.
“At Zurich we aim for AI to drive remarkable efficiencies and improvements in our risk engineering process”
_Mariana Santos, Head of Pricing at Zurich Insurance Portugal



