Bendigo and Adelaide Bank has employed generative AI and MongoDB to update the code for a retail banking application, paving the way to modernize and move some of its older applications to the cloud.
The modernization of the Agent Delivery System (ADS), teller software used by pharmacies and other non-bank businesses to offer banking services, is being highlighted as a global-first, in collaboration with MongoDB.
Like many Australian financial institutions, Bendigo and Adelaide Bank is in the midst of a cloud migration but has found not all applications and workloads can be easily moved or modernized.
Older applications can stifle a cloud migration, said Andrew Cresp, the bank’s Chief Information Officer, in an interview with iTnews. Cresp noted that he had experienced this challenge at two organizations, where after migrating 40-60% of workloads, the remaining applications seemed too costly to move.
The ADS application was initially seen as one such costly challenge. However, MongoDB Atlas and generative AI tools provided a solution. MongoDB’s team aimed to move the underlying database for ADS to their cloud-friendly platform, which they accomplished quickly in about three weeks.
The project then took a new direction. The joint team decided to refactor the application based on its database interactions. This approach was innovative and necessary, as it required MongoDB’s expertise in how the application communicates with the database and other systems.
A key outcome for Bendigo and Adelaide Bank was the realization that generative AI could handle much of the heavy-lifting in application modernization. The AI not only rewrote the app but also created documentation and automated testing capabilities.
Although there was initial skepticism, the AI’s output was surprisingly accurate, reaching about 90% correctness. Human intervention was needed only for minor corrections. For regulatory reasons, the generative AI used was private to Bendigo and Adelaide Bank, learning from internal APIs without sending data outside the bank’s boundaries.
The entire modernization took three months. MongoDB typically engages over 14 weeks, aiming to get the modernized application to pre-production within that timeframe.
Cresp mentioned that selecting the right application to test this modernization approach was a significant internal discussion. They opted for a business-critical core application, and the success has given the bank confidence to tackle bigger targets next.
The next project is the branch teller system, and the bank is currently working on modernizing its payments platforms using the same approach. The success of the ADS modernization has generated considerable internal interest, with many eager to apply this method to other applications.
Cresp noted that the bank had considered other generative AI tools for the ADS modernization before choosing MongoDB. The method proved quick and cost-effective compared to other intensive approaches.
The bank intends to continue using this application modernization method while staying alert to the rapidly evolving generative AI space, acknowledging that new advancements could emerge in the near future.
