New data techniques to turbocharge risk management

Risk management and data management have become central to the broader digitalisation efforts of financial institutions. A robust data strategy has an important role in supporting and enhancing risk management efforts
 

The panel

• Irene Galperin, Senior adviser for financial services, InterSystems
• Nicole Königstein, Chief artificial intelligence (AI) officer and head of quantitative research, quantmate
• Peter Quell, Head of the portfolio analytics team for market and credit risk, DZ Bank
• Biljana Vujovic, Chief risk officer, Addiko Bank Montenegro

In a Risk.net webinar, a panel of experts discussed best practices on how risk management can be improved using advanced modelling techniques and methodologies, and how to optimise data-driven operating models, eliminate errors and provide consistent, reliable data for informed decision-making.

They also shared insights on how risk managers can use big data to improve risk modelling.

 

Advancing big data and risk modelling

Today, firms employ big data to gain various perspectives on risk. But stress-testing and credit risk management still underutilise it, and improvements in these areas would be beneficial.

Nicole Königstein, quantmate

The panel discussed the use of AI in risk management, especially when combined with large language models (LLMs) and agents to sift through vast amounts of data. Nicole Königstein, chief AI officer and head of quantitative research at quantmate, explained that “you can use an agent guided by an LLM, like ChatGPT, to interact and provide summarised information. The agent can access different tools and application programming interfaces [APIs], gathering data from financial reports, sentiment analysis, social media and alternative sources such as satellite images.”

Königstein emphasised that these agents can perform reasoning, handle specific tasks and provide controlled information, enhancing the ability to detect patterns in large datasets. Human oversight remains essential to evaluating this data and making informed decisions.

Irene Galperin, senior adviser at InterSystems, highlighted the ultimate use of big data in risk management: “pulling data from various sources, including social media and internal sources, and transforming it into a consistent format for high-quality output.” She mentioned the evolution of data management, particularly the adoption of a ‘smart data fabric’. This approach dynamically integrates and unifies data, embedding machine learning and natural language processing to run models directly on the fabric.

Irene Galperin, InterSystems

Modernising risk management – and, simultaneously data infrastructure – is paramount for financial institutions. This enables organisations to respond flexibly to market events and capture real-time insights. Analysing historical datasets is also crucial for model development, ensuring models perform effectively before moving into production.

These techniques are being increasingly adopted, with firms leveraging smart data fabrics for various use cases, especially in risk management. Ensuring the risk department has a resilient infrastructure is essential for advancing risk management within the organisation.

 

Resilient data infrastructure

A resilient data infrastructure is crucial, the panel agreed, with a sound data governance framework for the entire organisation, ensuring clear data ownership.

Königstein highlighted the importance of a resilient data structure, recommending a cloud or hybrid cloud strategy for balancing scalability and security. “It should enable seamless data access and storage, and adopting a micro-service architecture can enhance system flexibility and fault tolerance,” she said.

She also emphasised the need for data quality, suggesting a framework to control the data that is entering models and validate data types and API access.

Biljana Vujovic, Addiko Bank Montenegro

Data governance is key to maintaining data standards and quality. Biljana Vujovic, chief risk officer at Addiko Bank Montenegro, outlined four pillars: proper data governance; a lean data architecture ensuring accessible and well-documented data; high data quality without complex lineages; and robust data capabilities. “This requires support from the entire organisation and understanding from the top,” she said, noting the challenge of integrating legacy systems.

Vujovic stressed the need for clear data ownership, comprehensive documentation and a process for managing data incidents and quality. Machine learning methods can help identify data quality errors, ensuring transparency and proper usage across the organisation.

Once a governance framework is in place, technology can implement it, allowing business users to access data without requiring domain expert permission. This frees up high-value resources to focus on critical tasks, especially given the constantly changing regulations. “Tracking data changes over time is vital for AI applications and overall data management,” Galperin noted.

Systematic data tracking and lineage improve data quality by reducing manual tasks. Implementing technology solutions can be done nimbly and non-disruptively. Galperin emphasised that democratising data access within a firm can advance and modernise risk management.

 

AI and advanced modelling techniques

There are many new capabilities in the market today, but financial services firms often struggle to attract and recruit the talent needed to implement and take advantage of them, the panel said.

“There’s a problem with hiring talent,” Königstein noted. “To push boundaries and set up things properly, you need good talent and the right infrastructure to fully utilise big data.”

Galperin agreed, emphasising the need for ongoing education and training for teams to take advantage of modern methodologies and techniques. “You can accomplish a lot by choosing a partner that can help fill gaps and conduct necessary training,” she said.

Peter Quell, DZ Bank

Importantly, firms need a proper AI strategy across the organisation, and a business case for the models they plan to put into production. In risk management, working in a regulated area requires models to gain regulatory approval, which demands a level of maturity in these techniques for organisations and regulators to keep up with progress.

“If we have strong IT foundations, we do not have to reinvent the wheel and can experiment with benchmark models. This would help gain initial experience with new technologies,” noted Peter Quell, head of the portfolio analytics team for market and credit risk at DZ Bank.

 

In summary

It is vital for firms to adopt modern data management and infrastructure technologies alongside contemporary modelling techniques. By leveraging real-time data sources, companies can enhance risk management, conduct ad hoc analyses and ensure they work with timely, consistent, accurate and usable data.

Establishing a single source of truth and a unified data view is achievable without significant disruption by using a smart data fabric architecture. This approach empowers risk organisations with self-service access to data, future-proofing the team’s – and organisation’s – risk management capabilities.

Cross-collaboration among teams is increasingly important, with data governance a key focus. Ensuring all teams have secure, appropriate access to the data they need is crucial. Additionally, organisations must scale their infrastructure and technology while retaining and hiring talent to maintain a competitive edge.
 

View the webinar

Advancing risk management using new data techniques

Sponsored content