Case Study: Visión Banco cuts model build time to under a week and doubles customer buy-rate with H2O.ai

AI Improves Profitability at Paraguay Bank

Visión Banco, a Paraguay-based bank serving small and micro businesses, needed faster, more accurate predictive analytics to expand services and assess credit risk at scale. The data science team had tried IBM SPSS and open-source tools (R, H2O, Openscoring.io) but model development and deployment could take months, slowing initiatives like churn, default prediction, and targeted offers — so they turned to H2O.ai’s Driverless AI.

H2O.ai deployed Driverless AI on IBM Power System AC922, automating feature engineering, model selection/tuning, interpretability, and MOJO-based deployment. The result: model build time fell from about six months to less than a week, customer propensity-to-buy doubled, model accuracy improved (translating to millions in potential additional revenue), and Visión Banco sped and scaled use cases such as propensity-to-buy, default prediction, and credit scoring.

Visión Banco

Ruben Diaz

Data Scientist