We recently hosted a roundtable focused on optimizing risk and exposure management with data insights. For financial institutions and insurers, risk and exposure management has always been a fundamental tenet of the business. Now, risk management has become exponentially complicated in multiple dimensions.
In this session we explored what firms are doing to approach the uncertainty with more predictability. Attendees included senior risk managers and analytics experts from financial institutions and insurance companies. Some of the key points raised during this session included:
- Pandemic Resiliency and Opportunities to Improve
- Low Probability, High Impact Events Readiness
- AI and ML’s current State of Play
Pandemic “Pressure” Testing
The participants were generally pleased and proud with how quickly their organizations adjusted and maintained operations, especially in the earliest days of the pandemic. However, through this real-time “pressure test”, they identified areas of weakness, dependencies, and opportunities. Area such as:
- Capacity planning requires greater attention, specifically for anomaly events
- Partner dependencies and concentrations (both geographical and vendor) need scrutiny as they are often assumed to be always available
- Regulatory examinations performed via digital means were a welcome advancement
- Transition of paper only and wet signatures to digital is more widely accepted
- Mental health of employees is a critical area to monitor
Low Probability, High Impact Events Readiness
Across the industry, the pandemic caused a huge breakdown in model performance due to the change in macroeconomic conditions and government stimulus packages. Modeling that was previously well established – for both commercial and consumer lines – became less reliable. Are organizations truly prepared for events that are unlikely but do happen?
Throughout this discussion, the group offered insights into the risk model impact:
- Model Robustness – it’s important to extract information that is available from the models by adjusting the assessment of the output of the models. Observe what the model has to offer even if not the intended output.
- Manage the extreme risk data points – As part of the analysis of the model output, it is critical to assess whether the shocks are realistic or true outliers.
- Contagion Impacts – interactions of different risks and related impacts were brought into focus by the pandemic and it highlighted the area that need attention. Here it was agreed to be open to the model providing unexpected outcomes, i.e. expecting an insurance exposure event when an asset risk impact was a bigger issue but not expected.
To enhance the risk models, one participant identified the need to look at additional data sources, how to do data exchange and how to link data sets to best manage the models.To integrate such extreme events, Machine Learning and AI can be utilized to help take a fresh look at models and how they can be scaled.
Challenges of implementing ML and AI at scale
The group was unanimous in the potential of Machine Learning and AI – “it is the way of the future”. ML and AI can react quickly and handle mass amounts of data to give leading indicators. It also enables agility to assess more data points quickly and assess KPIs.
Examples of successes shared include fraud detection and consumer behavior. It was also highlighted that workforce efficiency has seen success where AI and ML are enabling people to perform jobs faster and more efficiently and where it can be used to fix legacy issues and create efficiency.
The group then discussed some of the challenges in implementing ML and AI models at scale and to its potential. Good POC work is happening and there are advances in the use of Natural Language Processing (NLP). However, it was highlighted that none of this is easy to fully automate, especially within giant firms that are regulated and operate across multiple countries. The group concurred that they must be pragmatic about the approach, especially given the regulatory scrutiny related to bias and accuracy.
Quality of data
The regulatory oversight coupled with potential AI applications launched a discussion about the quality of the data – the classic “garbage-in, garbage-out” challenge. One participant emphasized their firm’s focus on the foundational aspects of data first before applying AI recognizing if data quality is not good the application of AI/ML won’t be applied with accuracy. This was not challenged, but the firms participating are in various states of adoption on data usage and governance. Some were actively seeking to harnessing additional data sources to inform their models, others are augmenting classic data, and others are assessing what are the most valuable data sources to use as there are so many to choose.
Using the right real-time data
There was enthusiasm overall for the use of AI and it’s potential. One participant called out the opportunity for AI as “phenomenal” – emphasizing that Twitter, court filings, geo location, ESG data, sentiment/social media – all have a profound story to tell and cannot be ignored, but it needs to be relevant by the time we derive value from it. Another attendee called out the untapped opportunity citing that the real advantage and opportunity is in areas not yet touched by analytics or AI.
At Cloudera, we believe in the untapped opportunity presented by data and AI, too. The way we say it is that data can make what is impossible today, possible tomorrow. Financial institutions and insurers need the ability to analyze and act on massive volumes of data to monitor, model, and manage risk across the enterprise. Using the most up to date market data and AI can help an organization gain better visibility to act as market events unfold. The Cloudera Data Platform enables organizations to gain control of their enterprise data and leverage AI to drive the business forward.
To learn more, join us for our upcoming virtual event, Transform Innovative Ideas into Data-driven Insights. We’ll be joined by Nvidia, Accenture and Forrester and share information on mastering complexity in dynamic markets, embedding AI into every aspect of the business and using hybrid cloud to drive efficiency.