.jpeg "Bendigo and Adelaide Bank uses GenAI, MongoDB to refactor application")
Bendigo and Adelaide Bank has utilized creative AI and MongoDB to transform the code for a retail banking application, paving a fresh path to upgrade and migrate some of its older applications to the cloud.
.jpeg&h=420&w=748&c=0&s=0)
Bendigo and Adelaide Bank CIO Andrew Cresp.
The enhancement of its Agent Delivery System (ADS) – teller software used by pharmacies and other non-bank businesses (“agents”) to provide banking services in their communities – is being highlighted as a pioneering initiative, executed in partnership with MongoDB.
Similar to numerous Australian financial institutions, Bendigo and Adelaide Bank is currently undergoing a cloud transition but has realized that not all applications and workloads can be effortlessly modernized or moved.
“Outdated applications can significantly impede a cloud migration,” explained the bank’s chief information officer Andrew Cresp to iTnews.
“I’ve encountered this in two organizations now, where you migrate about 40 to 60 percent of your workloads, and then you scrutinize the remaining applications and think, ‘this will be very costly. How can we justify this from a business perspective?’
“ADS was primed to be a prime illustration of this.”
The remedy was a blend of MongoDB Atlas and generative AI utilities developed by MongoDB’s professional services branch.
“The primary focus of the MongoDB team was to aid us in transitioning the [underlying] database [for ADB] to their platform, which is tailored for the cloud. We managed to execute this quite swiftly, in… three weeks,” expressed Cresp.
“This was where the dynamics really shifted. We convened with all stakeholders and they addressed that moving the database was intriguing, but streamlining the application posed the real challenge. So, why not tackle that?
“The joint effort entailed revamping the application based on its database access methodology. This unparalleled initiative is crucial as we required the MongoDB [team] to align their thought process with MongoDB principles.
“They contemplated how the application ‘communicates’ [with the database and other systems] and reverse-engineered that, which, I believe, sets it apart from other AI programs.”
For Bendigo and Adelaide Bank, an important upshot is the assurance that generative AI can immensely facilitate application modernization.
“It wasn’t merely a ‘copilot’ that improved our developers’ efficiency by 30 percent,” Cresp emphasized.
“It essentially revamped the application, crafted documentation… and automated testing functions for the application.
“This application modernization blueprint, in my opinion, is revolutionary for the industry at large.”
Despite facing initial doubt and “healthy skepticism” regarding assigning generative AI with rewriting an application, Cresp stated that the output was nearly 90 percent accurate.
“We were astounded by the quality of the 90 percent,” he added. “It makes some misinterpretations, for which human intervention is still necessary.”
Owing to the banking nature and regulatory constraints, the generative AI was confined to Bendigo and Adelaide Bank.
“The extensive language model we employed here is proprietary,” disclosed Cresp.
“We instructed it to study how to create robust APIs from our existing APIs. We refrained from sharing any data beyond our confines.”
The entire modernization project spanned three months.
MongoDB verified that the usual engagement scheme covers 14 weeks, consisting of seven two-week sprints, with the purpose of migrating the modernized application to pre-production within that timeframe.
Cresp conveyed that an extensive deliberation was held internally to pick the inaugural application for this unconventional modernization process. “We avoided tackling something excessively daunting,” he remarked.
Nevertheless, a core business-critical application was chosen – and the outcome has provided Bendigo and Adelaide Bank with the confidence to employ the same strategy for larger targets.
“We are now progressing to revamp our branch teller system,” stated Cresp.
“We are currently immersed in upgrading our entire branch teller system and payments infrastructure, and we are presently following this modernization approach.”
The ADS modernization also garnered considerable internal interest regarding its scalability across the application spectrum.
“Presently, our primary challenge resides in the heightened interest from various entities to leverage this capability for their applications, and we are endeavoring to regulate this enthusiasm,” noted Cresp.
“We have an extensive roster of subsequent applications that we wish to subject to this process, and the team involved in this initiative exhibits immense enthusiasm.”
Cresp observed that the bank evaluated other unidentified generative AI tools during the ADS modernization efforts before proceeding with MongoDB.
Besides facilitating rapid modernization, the bank affirmed that this method was also cost-efficient compared to more laborious alternative routes.
The objective is to further exploit this application modernization tactic while keeping an eye on the rapidly changing landscape of generative AI.
“The challenge with generative AI tools is the continuous enhancements, they relentlessly outdo each other. Hence, it’s prudent to avoid becoming fixated on a single solution as it may evolve rapidly,” Cresp remarked.
About Author
