AI-Powered Digital Workers Reshape Banking Operations
The integration of artificial intelligence into financial services is no longer a speculative trend—it is a structural transformation with measurable impact. Among the most compelling developments in this domain is the rise of digital workers, or “digital employees,” which are rapidly becoming core components of banking operations across China and beyond. These AI-driven agents are not mere automation scripts; they represent a new class of labor capable of executing tasks that range from simple data entry to complex cross-system workflows. Yet despite their growing adoption, significant challenges—including strategic misalignment, regulatory ambiguity, and industry fragmentation—threaten to undermine their full potential.
Digital workers are redefining how banks operate by embedding themselves into front-, middle-, and back-office functions. From credit assessments and account management to tax reporting and performance evaluations, these AI agents perform duties with speed, consistency, and precision unmatched by human counterparts. In some institutions, digital employees have already assumed 30% to 70% of routine operational tasks. At Sumitomo Mitsui Banking Corporation, for instance, over 1,600 digital workers have been deployed since 2017, collectively saving 3.5 million work hours and freeing up nearly 1,750 full-time human staff for higher-value roles. This case is not anomalous but indicative of a broader shift toward AI-augmented workforce models.
What distinguishes today’s digital workers from legacy robotic process automation (RPA) tools is their cognitive capacity. By integrating optical character recognition (OCR), natural language processing (NLP), knowledge graphs, and computer vision, modern digital employees can “read” unstructured documents, “understand” contextual meaning in client communications, and “reason” through rule-based decision trees. They no longer merely mimic keystrokes; they interpret, validate, and act. A loan contract generation bot, for example, can extract business rules from internal policy documents, populate dynamic fields based on borrower profiles, and output a legally compliant agreement ready for e-signature—all without human intervention.
This technological leap has catalyzed a quiet revolution in human resource strategy within financial institutions. The traditional tripartite division of labor—frontline service, risk control, and administrative support—is dissolving as digital workers assume roles that blur these boundaries. A single digital employee might simultaneously manage customer onboarding, conduct anti-money laundering checks, and file regulatory reports to central bank systems. This convergence enables what some industry experts call “process banking”: a modular, end-to-end operational model where value flows seamlessly across formerly siloed departments.
Yet the promise of such transformation is tempered by persistent friction points. Chief among these is strategic underestimation. Many banks still treat digital worker initiatives as IT department projects rather than enterprise-wide strategic imperatives. This narrow framing leads to “patchwork” implementations that optimize isolated workflows without reimagining the broader operating model. The result is a fragmented digital workforce that fails to deliver systemic efficiency gains. True transformation requires that digital employees be recognized not as tools but as strategic assets—integral to the bank’s talent portfolio alongside human staff.
Compounding this misalignment is a lack of institutional frameworks to govern digital labor. Traditional banking operations rely heavily on dual-verification protocols and clearly defined accountability chains—conventions that break down when a machine executes one step and a human reviews the next. Who bears responsibility when a digital employee misclassifies a transaction as low-risk, leading to a compliance breach? Current governance structures offer no clear answer. Moreover, digital workers routinely span departmental boundaries, executing tasks that cut across credit, compliance, finance, and HR. This fluidity clashes with rigid, hierarchical organizational charts that define roles by department rather than by function.
Technical debt further constrains scalability. Many banks operate on legacy core systems that were never designed for high-frequency, real-time interactions with external AI agents. Digital employees frequently encounter unstable APIs, unresponsive interfaces, or incompatible data formats—obstacles that cause workflow interruptions or even system crashes. The irony is stark: institutions investing in cutting-edge AI are often shackled by decades-old infrastructure that cannot support its full capabilities. Until these foundational systems are modernized, the deployment of digital workers will remain partial and precarious.
Perhaps the most insidious barrier is cultural resistance. At the executive level, some leaders view digital labor as a threat to managerial authority, fearing that workforce automation will erode their span of control. In the technology division, engineers accustomed to waterfall development cycles struggle to adapt to the agile, iterative nature of digital worker deployment. And among frontline staff, anxiety about job displacement fuels passive resistance or outright sabotage—such as deliberately introducing errors into digital workflows to “prove” their unreliability. Without a deliberate effort to cultivate a culture of digital co-creation, these human factors can stifle innovation regardless of technical prowess.
Regulatory uncertainty adds another layer of complexity. While China’s central bank has pioneered a “regulatory sandbox” approach to fintech innovation, digital workers occupy a gray zone. Regulators appreciate their potential to enhance compliance through consistent, auditable execution—but they remain wary of systemic risks arising from opaque algorithms or overreliance on automation. Current guidelines offer little clarity on how digital employees should be monitored, validated, or held accountable. This ambiguity discourages banks from scaling pilots into production, especially in high-stakes domains like credit adjudication or fraud detection.
The market itself is rife with disorder. The burgeoning digital worker sector has attracted a flood of vendors, many offering superficially similar but technically disparate solutions. Without standardized benchmarks for performance, security, or interoperability, banks face a buyer’s dilemma: how to distinguish robust platforms from overhyped demos. Price wars have further degraded quality, as some vendors slash costs to gain market share, compromising on support, training, and long-term maintainability. This “race to the bottom” risks tarnishing the entire category and delaying mainstream adoption.
Addressing these challenges demands a coordinated, multi-stakeholder response. First, banks must elevate digital worker strategy to the C-suite. This entails developing a holistic roadmap that aligns AI deployment with talent development, process reengineering, and customer experience goals. Human resources departments—not just IT—should lead this effort, treating digital employees as part of an integrated “dual-talent” workforce that includes both humans and machines.
Second, institutions need new governance mechanisms. This includes creating a Center of Excellence (CoE) for digital labor, tasked with onboarding, scheduling, monitoring, and retiring digital workers—much as HR manages human hires. Clear protocols must define handoff points between human and machine, establish audit trails for every action, and assign liability in case of error. Cross-functional task forces can help redesign workflows to leverage digital strengths while preserving human oversight where judgment is critical.
Third, systemic modernization cannot be deferred. Banks must accelerate the retirement of legacy platforms and invest in API-driven, cloud-native architectures that support real-time AI integration. This is not merely a technology upgrade but a strategic enabler—without it, digital workers will remain confined to the periphery of operations.
Fourth, cultural transformation must be intentional. Leadership must champion digital co-creation, not as a cost-cutting measure but as a productivity multiplier that liberates human talent for creative, empathetic, and strategic work. Training programs should equip staff to “manage” digital colleagues, while incentive structures should reward collaboration across human-machine teams.
Fifth, regulators should expand sandbox programs to explicitly include digital worker pilots. These testbeds can generate empirical data on risk profiles, failure modes, and consumer impacts—informing the development of nuanced, risk-proportionate rules. Rather than imposing blanket restrictions, supervisors should define “safe harbors” for well-governed AI deployments while flagging high-risk use cases for closer scrutiny.
Finally, industry bodies must establish technical and ethical standards. A unified framework for digital worker implementation—covering design principles, security controls, performance metrics, and interoperability requirements—would reduce buyer confusion and elevate market quality. Such standards could be developed through consortia involving banks, vendors, academics, and regulators, ensuring they reflect both technical rigor and operational reality.
The trajectory is clear: digital workers are not a passing fad but a foundational element of next-generation banking. Their ability to deliver 24/7, error-free execution at scale offers a path to unprecedented efficiency, compliance, and customer satisfaction. But realizing this potential requires more than algorithmic sophistication—it demands strategic vision, institutional courage, and collaborative governance. Financial institutions that treat digital labor as a mere operational tweak will fall behind. Those that embrace it as a catalyst for holistic transformation will define the future of finance.
Lu Minfeng, Nanjing Tech University Internet Finance Innovation Development Research Center
Wang Zugang, Nanjing DeepThink Data Technology Co., Ltd.
Journal of Financial Innovation and Technology
DOI: 10.12345/jfit.2025.0401