At the Eighth China Fintech Forum, held as part of the China International Fair for Trade in Services on September 9, industry leaders gathered to discuss technology-driven digital innovation in financial services. The chief information officer of East Asia Bank (China) took the stage to deliver insights on modern software productivity approaches in commercial banking.

Opening his presentation, Wang Yue acknowledged the excellent perspectives shared by previous speakers and pivoted toward a fundamental question for financial institutions in the AI era: What should our research and development frameworks prioritize, and how should they operate? Over the past eighteen months, his team has conducted extensive experimentation, confirming that artificial intelligence can drive substantial efficiency gains. However, for the banking sector, the critical challenge lies in balancing cost, iteration speed, governance, and boundaries simultaneously. The ultimate objective is establishing a software productivity methodology uniquely suited to banking environments while creating practical, actionable implementations.

Artificial intelligence has evolved from being an auxiliary plugin to becoming a core participant throughout the software lifecycle. Across the globe, major software enterprises are rebranding traditional development processes with new terminology. Though naming conventions differ, the underlying principles converge on three key concepts. First, AI coding must function as an essential component embedded within the software lifecycle rather than serving merely as an external enhancement tool. Second, AI integration should extend beyond the coding phase to encompass requirements gathering, design, testing, delivery, and ongoing operations. Third, rather than simply selecting an off-the-shelf AI solution, banks must identify approaches aligned with their unique corporate DNA and institutional characteristics.

The banking sector faces distinctive obstacles in human-machine interaction and governance limits. Traditional bank projects progress through requirements, design, development, testing, and delivery phases. While technology companies have embraced comprehensive AI adoption, banks recognize that complete delegation to artificial intelligence is neither prudent nor feasible. Consequently, banking institutions embed human oversight checkpoints throughout their workflows. People remain the paramount element in this equation. The emphasis centers not on AI replacing human workers but rather on productive human-machine collaboration. While AI substantially enhances execution capabilities, accountability boundaries, validation frameworks, and governance mechanisms must advance in tandem. Bank project participants frequently cite tedious documentation, difficult requirement articulation, and cumbersome approval processes as their primary frustrations. Creating technical requirements proves straightforward; the genuine bottleneck emerges from signature chains across documents, audit trail requirements, and compliance closure at each procedural stage.

The practical implementation pathway encompasses engineering discipline, scenario-based approaches, and organizational cultural transformation. Banking institutions are actively merging AI with traditional software development lifecycle processes, yielding three core value propositions. The engineering dimension addresses inconsistent AI coding outcomes among team members. Disparities arise from varying prompt formulations, divergent contextual understanding, and insufficient system familiarity. Standardization at the engineering level becomes imperative, creating a unified entry point spanning requirements through delivery. By codifying prompts and institutional code assets into structured specifications, AI coding transforms from ad-hoc experimentation into governable, replicable, and scalable productivity infrastructure capable of autonomous operation. The scenario-based dimension recognizes that business requirements frequently arrive as brief statements such as "I need an information system." However, non-functional requirements, including monitoring anomalies and vulnerability scanning defects, equally demand attention. Capturing these signals within a scenario awareness framework, AI converts them into comprehensive specifications navigable across the entire delivery chain. Once this cycle functions effectively, scenario-based implementation transcends simple requirement discovery, enabling self-iteration of micro-functionalities. The organizational culture dimension emerges naturally once engineering and scenario-based approaches mature. Within most banks, only a handful of business unit members serve as requirement originators; the majority remain focused on sales or direct customer engagement. When institutional learnings become visible and accessible across the entire organization, every employee transforms into both a participant in and beneficiary of technological innovation. This cultural evolution carries greater significance than any technological advancement.

To integrate development scenarios thoroughly with the software development lifecycle main chain, banking institutions have established five points of unification. The scenario entry point consolidates business requirements, regulatory mandates, security defects, and operational alerts through a single portal. The requirement contract ensures all documentation derives from AI-generated templates with consistent contextual framing. The development phase encompasses code generation, unit testing, and system integration and user acceptance testing within the primary chain. Human governance maintains standardized audit checkpoints with uniform permissions and traceability mechanisms at critical milestones. The feedback loop converts individual efficiency gains into organizationally reusable productivity improvements. When new scenarios emerge, whether originating from business units or self-identified technological needs, this framework enables rapid integration into the main chain. Historical knowledge repositories provide contextual structure, allowing AI to generate requirement texts and design documents before routing through native AI and agent layers where specialized intelligent agents decompose and translate specifications for final delivery. Post-launch operational data feeds back as fresh input, sustaining continuous improvement cycles.

A cross-border points redemption collaboration with the parent Hong Kong institution illustrates this methodology in practice. Previous agent-based AI initiatives over two years had failed to resolve fundamental problems. Business departments continued flooding technology teams with requirements while technology staff remained overwhelmed. Recalibrating the approach using the five unification principles, the project addressed scenario classification in conjunction with the main lifecycle chain. The core scenario involved Hong Kong credit card customers redeeming rewards through mainland retail banking platforms. This undertaking spanned multiple systems, cross-border compliance considerations, and data flow complexities. Implementing the complete loop of scenario perception, requirement generation, system design, coding and testing, delivery, and feedback collection demonstrated that even modest requirements could achieve efficient execution when scenario-based methodology aligns with the primary chain. Customer feedback, operational metrics, and performance indicators generated inputs for subsequent iteration cycles. This represents an initial experiment rather than a finished solution, validating the concept while acknowledging the aspiration that this methodology extends beyond single business lines, individual systems, or isolated scenarios to function as bank-wide capability infrastructure.

Three principles summarize the vision for banking transformation through artificial intelligence. Embedding intelligence into workflows means AI participates meaningfully throughout every software development lifecycle phase rather than existing as ancillary tools. Governance maintains boundaries through accountability, validation, and audit systems that evolve alongside human-machine collaboration. Value creation forms feedback loops where each response and metric becomes the starting point for subsequent iterative improvements. This exploration promises not merely to reshape how bank technology teams operate but to elevate organizational efficiency and innovation culture comprehensively.

These insights represent one institution's journey toward redefining software productivity in the banking sector, offering a blueprint for peers navigating similar digital transformation trajectories. By addressing engineering standardization, scenario awareness, and cultural readiness as interconnected priorities, the presentation illuminated a practical path forward. The integration of AI throughout the software lifecycle, governed responsibly and designed for continuous refinement, positions commercial banks to deliver technological excellence while maintaining the trust and compliance standards essential to banking operations.