How Absa Group uses AI

Absa put an OCR and AI gateway into its Debt Review team, which supports over indebted South African consumers. The gateway extracts and captures information from standard document types that were previously keyed in by hand, and the bank reports a one day turnaround for all new documents and emails since the late June 2026 go live. It is a document intake job for customers in financial distress, not a collections conversation agent.

Absa, one of South Africa's five largest banks, integrated FICO's Customer Communication Services with its existing FICO Falcon Fraud Manager to become the first of the country's big five banks to use WhatsApp for two way, real time fraud verification. When suspicious activity is flagged, the system sends an interactive WhatsApp message within milliseconds, the customer confirms whether the transaction is genuine or fraudulent, and where fraud is confirmed a fraud representative is automatically brought into the conversation. FICO gave Absa its 2026 Decision Award for the results.

Absa tested SymphonyAI financial crime AI against its existing transaction monitoring system, using two years of masked data from one country and then data from a second country to build AI models for alert efficiency and new risk detection. SymphonyAI reports that the tests cut false positive alerts by 77% while still capturing all suspicious activity the existing system had identified. An Absa compliance executive said the bank decided to roll out the SymphonyAI model to three further countries in Africa. The record is stage pilot because the sources describe proofs of concept, not live production figures.