Digital bank Tonik has completed its transition to an artificial intelligence (AI)-driven credit decisioning system to strengthen its risk management and support more inclusive lending.
The new system uses AI scorecards for real-time and batch decision-making. It combines more than 1,000 data points, such as device behavior, network data, and repayment patterns, to better assess risk and pricing compared with traditional models. Tonik said this approach has more than twice the predictive accuracy of older scorecards, improving returns and enabling responsible lending at scale.
“We rebuilt the decisioning engine around data and actionability,” said Biswa Banik, chief data officer of Tonik. “By fusing diverse signals into our AI scorecards, we’re making faster, fairer, and more profitable credit decisions at scale, and generating highly customized offers by customer profile. That lets us responsibly serve a broader share of the population with the right product, price, and limit.”
Tonik has also applied AI to other areas of its operations, including customer chat, voice services, and back-office automation, to improve efficiency and service quality. Over the past three years, the company said its loan portfolio grew tenfold while it reduced its operations staff by more than 20%. The bank expects its AI program to save about $20 million in operating costs over the next three years.
“AI is giving us tangible savings, not hype,” said Tomasz Borowski, chief operations officer of Tonik. “It lets us scale profitably without adding staff, so every new peso of lending drops more to margin.”
Tonik reported a more than threefold annual growth in both loan portfolio and revenue. The company aims to continue expanding through broader use of AI tools, new partnerships, and further automation in lending and collections.
Tonik said its AI-driven model supports responsible growth in the Philippines’ estimated $40 billion unsecured consumer lending market by improving access to credit while maintaining sound risk management practices.