Royal Bank of Canada is using its own proprietary AI foundation model to improve credit adjudication capabilities and loyalty programme personalisation.
Created by RBC's AI research unit Borealis, Atom (Asynchronous Temporal Model) was trained using large-scale financial datasets including billions of client transactions which the bank says gives it "deep financial expertise" that can be used across a variety of tasks.
The model is seen by RBC as central to its stated ambition of realising $700 million to $1 billion in enterprise value generated from AI-driven benefits by 2027.
It is already being used for credit adjudication to make the process more accurate, consistent, and insightful and enables the use of large volumes of complex data including transaction histories, and non-traditional sources. This, says RBC, helps it extend credit to clients such as newcomers who may be disadvantaged by traditional models.
"Applying AI to credit adjudication has provided an opportunity to evolve decisioning to ensure the best customer outcomes and increased levels of personalization in a way that wouldn't be possible with traditional adjudication models and processes," says Gopala Narayanan, SVP, chief risk officer, personal banking, RBC.
Atom is also being used for the Avion Rewards programme to provide more personalised recommendations, leading to a "significant lift" in redemption conversion as well as cost savings and increased benefit adoption.
"Atom represents the future of banking at RBC," says Foteini Agrafioti, SVP and chief science officer, RBC. "It helps to personalize products and services at an individual level and enables us to more deeply understand our clients' individual circumstances."