Banks have a lot of data on their customers, but most of it sits unused. The customer gets flooded the same generic offer everyone else gets, and the bank misses a warning sign of fraud its own systems could have caught earlier, simply because nothing connects the dots fast enough.

Discovery Bank, the South African digital bank owned by Discovery Limited, built an artificial intelligence (AI) system specifically to close that gap, reading customers’ actual spending, saving, credit and rewards behavior to figure out what they need next, before they ask. The result has been a return on investment of more than 500%, driven by much faster data work and faster AI rollout, according to a case study Databricks published about the partnership. Databricks is the bank’s data platform vendor.

Pipeline development now runs 20 times faster, and teams build new data products five times faster. The bank can create more than 300 AI models a day. “With the Databricks platform and the Azure OpenAI-powered assistant, we’ve seen a 500% ROI,” Stuart Emslie, Discovery Bank’s head of actuarial and data science, said in a Microsoft customer story that carries the same figure.

The bank’s system decides what to show each customer next based on their own behavior: an offer, a reminder, a financial tip. That next-best action model lifted client engagement impact by 40%, according to Databricks.

The Same System Also Fights Fraud and Assesses Credit Risk

Preferred Source

We’d love to be your preferred source for news.

Please add us to your preferred sources list so our news, data and interviews show up in your feed. Thanks!

Discovery Bank uses the same system for fraud detection, credit decisions and general customer service, giving the bank one AI system doing several of its biggest jobs instead of building a separate tool for each one. On fraud, the bank said in a May 14 press release that the system has prevented an estimated 100 million rand (about $6 million) in potential losses since late 2025.

Customer service runs on the same system, which fully resolves about 55% of client queries on the first interaction, according to the same release. Discovery Bank framed the AI push as one piece of a wider consolidation. “We are entering the next phase in Discovery Bank as we deepen the integration of banking, protection, rewards and investments into a single digital experience,” CEO Hylton Kallner said in the release.

Deutsche Bank and Others Put Numbers on AI Payback

Discovery Bank isn’t the only bank putting a specific number on what AI is worth. Deutsche Bank is using AI to cut the completion time of some tasks from two years down to as little as three months, Denis Roux, chief information officer for the bank’s investment bank, told Reuters in June, PYMNTS reported.

Eighty-five percent of financial services firms with at least $1 billion in annual revenue plan to increase AI budgets over the next 12 months, according to PYMNTS Intelligence’s May edition of the Enterprise AI Benchmark Report, and 65% cite productivity gains as the reason, tied with competitive positioning.

Ninety percent to 100% of enterprises surveyed by PYMNTS Intelligence for the August 2026 Enterprise Benchmark Report said they’re now getting some kind of return on AI investments. Full payback is rarer, with only 5% to 10% who said they’re currently getting a full return. At least half of firms in every industry group projected full payback at five to six years. Against that backdrop, a bank claiming more than 500% is claiming something almost nobody else has.

For all PYMNTS AI and digital transformation coverage, subscribe to the daily AI and Digital Transformation Newsletters.