Previously, Intesa Sanpaolo’s risk management team used an on-premise data analytics lab to prototype machine-learning (ML) solutions and risk models that would allow it to monitor and analyze risk and assess its potential impact. However, this lab environment was separate from the production environment, so when it came to preparing these solutions for industrialization, a different team had to spend significant time rewriting them in the production environment, delaying their release dates and making it harder for the bank to react quickly to rapidly changing markets. 

At the same time, unable to scale up its on-premise architecture on demand, Intesa Sanpaolo was restricted to developing models in sequence, which again slowed its time to market. 

“To scale up our capabilities, we needed a cloud-based solution and cutting-edge AI technology to quickly spot patterns in our data,” recalls Domenico Fileppo, Executive Director - Operational, IT and Cyber Risk Management at Intesa Sanpaolo. “The flexibility, ease of use, and advanced capabilities of Vertex AI, Gemini, BigQuery, and Looker convinced us Google Cloud was the right home for our Democratic Data Lab.”

With Intesa Sanpaolo already in the process of migrating its core banking application to Google Cloud, the Group Technology & Services and Data & Artificial Intelligence Office teams began building its Democratic Data Lab for risk management in 2021.

Running its Democratic Data Lab on Google Cloud means Intesa Sanpaolo’s risk management team, supported by technical teams, can use Google Kubernetes Engine to scale its resources up and down on demand, rather than being constrained by on-premise hardware. This allows the team to develop models in parallel, rather than in sequence, helping the bank to release risk management solutions more quickly. And with managed services such as Vertex AI and BigQuery, the team doesn’t need to spend time managing infrastructure and can focus instead on developing the most effective models. 

“With our on-premise environment, developing ML models involved spending time adapting our lab environment before we began. Google Cloud makes us more agile because we don’t have to manage our infrastructure. We’ve saved up to 30% of our time developing models. Those time savings are crucial for the work we do.”

As the name suggests, the Democratic Data Lab is allowing the Intesa Sanpaolo risk management team to democratize access to risk management data insights across the bank. Business team members no longer need to rely on data analysts for insights, and instead can use BigQuery to easily navigate and query the data they need, at a more granular level than before. With Looker, meanwhile, the risk management team can quickly produce dashboards for the chief risk officer and other board members across the organization, helping them to visualize key risk indicators in real time.

The Intesa Sanpaolo ICT risk and security team has also begun using generative AI to improve the efficiency of its Democratic Data Lab by speeding up the way it extracts and stores data to build and train its risk management models. Previously, this was a time-consuming process, as this data could be incomplete or improperly structured. Now, the team uses Gemini 1.5 through Vertex AI to speed things up.

“Previously, every time we received the data, we needed someone to manually extract the correct information,” explains Giorgio Bella, Head of IT & Cyber Risk Analysis and Controls at Intesa Sanpaolo. “Now we can use Gemini to automatically extract and load the right information, saving our analysts hours of repetitive work, and allowing them to do more interesting and useful tasks instead. Gemini is outperforming our expectations and making a tremendous difference for our team.”

As Intesa Sanpaolo continues to expand its data-driven approach to risk management, with increased use of Looker dashboards and more generative AI solutions to improve efficiency, Fileppo is confident the bank is keeping pace with a fast-changing risk landscape: “Financial risk is evolving rapidly and we need to continue becoming more data driven. It’s a long journey. Rome wasn’t built in a day–they were laying bricks every hour–and that’s just what we’re doing with Google Cloud: building towards our target together.”