At the International Conference on Sustainable Computing and Communication Technologies (ICSCCT 2026), hosted by the Faculty of Information & Communication Technology at the University of Malta and published in partnership with Springer, Subhankar Panda took the stage as keynote speaker to make a case that few in the audience disputed but many are still slow to act on: banking’s next decade will be decided by how well institutions absorb AI and machine learning into their core planning, not just their back-office tools.Subhankar’s talk moved past the familiar narrative of AI as a cost-cutting layer bolted onto legacy systems. Instead, he framed AI and ML as instruments that should sit inside a bank’s financial planning function itself — shaping how risk is priced, how liquidity is forecast, and how decisions that once took committees weeks now take model minutes.“Banks that treat AI as an add-on will keep optimizing yesterday’s processes. Banks that treat it as a planning partner will start asking better questions about tomorrow’s balance sheet.”From automation to anticipationA recurring thread in the address was the shift from AI that automates known tasks to AI that anticipates unknown ones. Subhankar pointed to machine learning models now capable of stress-testing portfolios against scenarios that human analysts would rarely think to construct on their own, and of surfacing early signals in transaction data long before they would show up in a quarterly report.He argued that this shift changes the job of the financial planner as much as it changes the job of the engineer: less time spent assembling numbers, more time spent deciding what the numbers mean.“The value isn’t in the model producing an answer. The value is in the banker knowing which question was worth asking.”Trust, governance, and the limits of the modelSubhankar was careful not to present AI adoption in banking as a purely technical problem. He spent a portion of the keynote on governance — the need for explainability in credit and risk decisions, the regulatory weight carried by financial institutions, and the reputational cost of deploying models that cannot account for their own reasoning. In a sector where a single miscalibrated model can ripple through customer trust and compliance exposure alike, he suggested that responsible deployment matters as much as capability.This emphasis connects to the broader argument running through Panda’s recent work: that reliability engineering and AI adoption are not separate conversations. His writing on AI-driven test automation in enterprise delivery has made a related point in the software world — that as systems grow more autonomous, the discipline of verifying them has to grow just as fast, or the speed AI promises turns into risk instead of advantage. Applied to banking, the same logic holds: an AI-driven planning system is only as trustworthy as the testing and governance built around it.A call to institutional patiencePanda closed by cautioning against treating AI transformation in banking as a single project with an end date. He described it instead as a standing capability that has to be funded, staffed, and revisited continuously — closer to how institutions treat risk management than how they treat a software rollout.“The banks that get this right in five years are the ones treating it as infrastructure now, not the ones waiting for a finished product to buy.”ICSCCT 2026 drew researchers and practitioners from computing, sustainability, and applied technology fields across two days of sessions at the University of Malta, with proceedings to be published through Springer.
