Artificial Intelligence in Finance

Navigating the Complexities of Risk Data Aggregation and AI Integration in Modern Banking

Financial institutions globally continue to grapple with deeply entrenched operational inefficiencies, characterized by fragmented data architectures and a heavy reliance on recurring manual corrections. According to supervisory assessments released by the European Central Bank (ECB) in February 2025, these persistent vulnerabilities stem largely from foundational weaknesses in internal data controls and legacy infrastructure. Such structural shortcomings severely impair a bank’s capacity to respond swiftly, accurately, and decisively during periods of systemic financial stress or sudden market volatility.

These regulatory warnings underscore a broader, systemic compliance challenge that has plagued the banking sector for over a decade. The origins of contemporary data aggregation mandates trace back to the aftermath of the 2007–2008 global financial crisis, which exposed severe blind spots in how major financial institutions monitored and reported their risk exposures. In response, the Basel Committee on Banking Supervision (BCBS) established formal guidelines known as BCBS 239, designed to strengthen risk data aggregation and risk reporting practices.

Despite years of institutional upgrades, adherence remains critically low. A comprehensive November 2023 review published by the Basel Committee revealed that out of 31 global systemically important banks (G-SIBs) assessed—using baseline data compiled as of June 2022—only two institutions were found to be fully compliant with all risk data aggregation and reporting principles. This stark compliance gap illustrates that data fragmentation is not merely an isolated technical annoyance, but an industry-wide structural hazard that continues to invite intense regulatory scrutiny across international jurisdictions.

The Operational Dilemma: Scaling Automation Amid Fragmentation

Against this challenging backdrop of heightened oversight and structural friction, banking executives are under mounting pressure to deploy artificial intelligence and advanced automation to streamline operations. However, integrating sophisticated technology into legacy environments often introduces new hazards if underlying processes remain flawed. This operational paradox was recently explored on the Emerj AI in Business Podcast, where Nathaniel Bell, Data Management Director at Wells Fargo, joined host Yolandi de Weerdt to examine the multifaceted challenges of scaling automation across fragmented banking operations.

During the discussion, Bell and de Weerdt analyzed how inconsistent data interpretations, execution evidence gaps, and ballooning regulatory demands collectively complicate executive decision-making. Specifically, the conversation focused on how leadership teams evaluate where to deploy capital for AI investments and how to accurately measure their operational value. Rather than viewing artificial intelligence as a universal remedy for institutional inefficiency, Bell cautioned that deploying automation without first addressing foundational data hygiene risks "industrializing bad data quality."

The dialogue highlights a critical evaluation framework for banking leaders navigating digital transformation: before introducing advanced machine learning or robotic process automation, institutions must standardize workflows, close control gaps, and embed monitoring directly into data products to reduce the heavy compliance tax currently borne by human operators.

Standardize Workflows Before Scaling Automation

Large financial institutions rarely operate on unified, greenfield technology stacks. Instead, decades of corporate mergers, product diversification, regulatory updates, and organic growth leave banks with inherited processes and platforms distributed across disparate business units. As Bell points out during the podcast, individual teams frequently operate under the assumption that they are executing tasks correctly within their own silos, even as the broader institution struggles to produce consistent, harmonized outcomes across the complete end-to-end workflow.

This disconnect is driven as much by a lack of shared operational context as it is by data accessibility constraints. A typical banking process—such as commercial loan origination, anti-money laundering (AML) screening, or customer onboarding—often traverses multiple departments and software systems without ever establishing a standardized, enterprise-wide interpretation of the underlying data attributes being transferred. Furthermore, evolving regulatory requirements introduce another layer of operational variance, as compliance teams must constantly interpret written supervisory guidance and translate it into day-to-day execution rules.

When automation is injected into a fractured environment, it does not resolve these underlying contradictions; instead, it accelerates them. To mitigate this risk, Bell advocates for a disciplined three-step approach:

  1. Map and understand the complete end-to-end workflow before determining automation points.
  2. Resolve conflicting definitions and business rules where input disagreements trigger repeated manual reconciliation.
  3. Evaluate whether standardizing the workflow eliminates the need for certain manual steps entirely, rather than merely speeding up a broken process.

By way of illustration, consider a standard customer-document review process. If one risk team considers a corporate tax document current for 90 days, while another division applies a 30-day validity threshold, introducing a high-speed document routing tool does not solve the underlying dispute. The receiving team will still pause the workflow to resolve the conflicting interpretation. True operational improvement requires establishing a unified enterprise acceptance rule before attempting to automate the transfer of the document.

Finding Control Gaps and Reducing Manual Verification

Software vendor demonstrations often project an illusion of seamless efficiency, showcasing AI tools that appear to solve complex operational hurdles instantly. However, these demonstrations frequently fail to address how a bank will independently verify, monitor, and audit the tool’s performance in a live production environment. Bell emphasizes that financial leaders must anchor their technology investments in concrete business objectives, ensuring that automated outcomes can be rigorously measured, monitored, and understood well enough to facilitate continuous improvement.

Establishing this level of operational transparency requires robust execution evidence. While standard procedure documents outline the intended sequence of actions, reconstructing what actually happened during a complex transaction becomes increasingly difficult as data flows across multiple legacy systems. Bell identifies this challenge as reconciling "what actually happened" with the expected workflow, proposing the strategic use of automated system logs to prove that mandatory operational steps occurred and to flag exceptions in real time.

For instance, consider a regulatory mandate requiring customer identity verification documents to be fully reviewed before a credit application advances. In a fragmented architecture utilizing separate banking applications, staff are often forced to manually hunt down records across different databases to prove compliance. Implementing structured automated logging allows institutions to programmatically confirm that a case advanced only after the requisite review was successfully completed.

To manage this transition effectively, Bell’s framework for evidence review involves:

  • Defining the precise compliance and operational metrics required to substantiate a transaction.
  • Identifying system logs and digital audit trails that capture actual execution steps without manual intervention.
  • Determining whether workflows should retain exhaustive operational logs or focus selectively on exception reporting to minimize data storage and review burdens.

This evidentiary burden carries a substantial cost. Bell characterizes the generation, organization, and maintenance of manual compliance evidence as a "heavy tax" that drains institutional resources long after the initial transaction has closed. When evaluating an AI or automation investment, banking leaders must calculate whether the tool merely accelerates transaction processing while leaving the evidence-collection tax intact, or whether it actively reduces the auxiliary labor required to prove regulatory compliance.

Minimizing Manual Oversight While Preserving Accountability

The continuous expansion of regulatory complexity has forced financial institutions to layer on additional controls, audit checkpoints, and supervisory reviews. Yet, as regulatory expectations intensify, banks can no longer meet this compliance burden simply by adding headcount and manual oversight. Bell argues that the industry must transition away from reactive, document-heavy reviews and instead embed automated controls directly into the underlying workflows and data products.

This shift fundamentally alters the daily responsibilities of risk and governance professionals. Rather than spending valuable hours manually reviewing static documents and coordinating cross-departmental sign-offs, governance staff can pivot toward systemic monitoring—analyzing how operating systems function and identifying early behavioral signals that indicate a potential risk event.

In this modernized operating model, a deviation from historical data patterns acts as an automated signal requiring targeted human investigation. Interpreting these signals, however, demands deep institutional knowledge: analysts must understand what the data represents, its original source lineage, how it has been transformed, and which specific business outcomes or regulatory controls it impacts.

To execute this shift successfully, institutions must connect three core responsibilities:

  1. Automated Detection: System-level controls that continuously monitor data streams and flag operational anomalies or control breaches.
  2. Contextual Interpretation: Human risk professionals evaluating flagged signals within the broader context of the business process and regulatory framework.
  3. Accountable Resolution: Clear ownership assigned to investigate the root cause of an anomaly and update underlying system parameters to prevent recurrence.

Crucially, technology adoption in banking mirrors physical fitness regimens: purchasing sophisticated equipment yields no health benefits without consistent lifestyle changes. Similarly, deploying advanced AI or automated controls requires banks to actively dismantle legacy habits. For example, if a compliance team continues to maintain a shadow spreadsheet tracking approvals manually even after an automated workflow generates certified digital audit trails, the institution has failed to realize the efficiency gains of the technology and remains burdened by redundant work.

Broader Implications and Future Outlook for Banking Operations

The convergence of strict regulatory enforcement by institutions like the ECB, persistent data aggregation deficiencies highlighted by the Basel Committee, and practical operational insights from industry leaders like Wells Fargo’s Nathaniel Bell paints a clear picture of the banking sector’s current inflection point. Financial institutions are discovering that digital transformation cannot succeed as a superficial overlay on broken foundations.

As banks evaluate capital allocation for artificial intelligence and data modernization initiatives throughout 2025 and beyond, executive leadership must adopt a holistic, disciplined approach. By prioritizing end-to-end workflow standardization, eliminating ambiguity in data definitions, automating execution evidence capture, and embedding controls directly into data products, banks can successfully alleviate the heavy administrative tax of manual oversight.

Ultimately, this structural evolution does not diminish the need for human expertise; rather, it elevates it. By delegating repetitive reconciliation and verification tasks to properly configured automation, financial institutions can empower risk professionals to focus on strategic monitoring, anomaly interpretation, and proactive problem-solving—ensuring both regulatory compliance and operational resilience in an increasingly complex financial landscape.

Related Articles

Back to top button