From Cumbersome Processes to Automated Control

Having worked in financial operations for many years, I have seen several waves of technology promised to transform the back-office. Systems have improved, automation has expanded, and transaction volumes have grown significantly. Yet one thing has remained surprisingly consistent: exceptions still consume a large part of operations teams’ time. In many ways, the daily work of operations analysts today still looks very similar to what it did years ago.

Last week at the Asian Banking & Finance and Insurance Asia Summit in Manila, I had the opportunity to discuss this topic with operations practitioners from banks and insurance firms in the Philippines. During the coffee roundtable session, a common theme emerged. Many institutions have invested heavily in automation, but analysts still spend hours investigating breaks, gathering data from multiple systems and resolving issues manually. In fast growing markets across Southeast Asia, including the Philippines, these pressures are becoming more visible. Digital payments, securities processing and insurance activity are expanding quickly. At the same time, regulatory expectations around transparency and operational resilience continue to increase. Operations teams are therefore under pressure from both sides. Volumes are rising, but many investigation processes remain heavily manual. This is where a new shift is beginning to appear. Agentic AI is starting to change how investigation workflows can be handled. Unlike traditional automation, which relies on fixed rules, agentic AI systems can analyse data across systems, identify likely causes of exceptions and take action while maintaining transparency and governance.

During the roundtable discussion titled “AI in Operations: From Cumbersome Processes to Automated Control”, several themes emerged.

Across financial institutions, the same pressure points continue to appear. Transaction volumes continue to increase across payments, securities processing and reconciliations. At the same time, regulatory expectations around reporting, transparency and operational resilience are expanding. Yet many operational environments remain fragmented. Data is distributed across multiple systems, processes rely on manual intervention, and exceptions require detailed investigation. As a result, operations teams spend a significant portion of their time gathering data, checking inconsistencies and resolving breaks rather than managing the process itself. These exception-heavy workflows create bottlenecks that traditional systems struggle to address.

Traditional automation has delivered real improvements. Workflow engines, reconciliation platforms and rule-based automation have streamlined many high-volume processes. However, these tools were designed for structured workflows where data conditions are predictable. When exceptions occur, investigations often require interpretation of data across multiple systems and decisions based on incomplete information. This is where rule-based automation reaches its limits. Processes often fall back to manual work supported by spreadsheets, emails and individual expertise. While this keeps operations moving, it introduces inefficiencies and operational risk.

AI can deliver the most immediate value in exception management and investigation workflows. Rather than replacing existing systems, AI can act as an intelligence layer across them. Agentic AI systems can gather relevant data automatically, analyse relationships between data points, determine the likely root cause of an exception and propose or execute remediation steps. This reduces the time required to investigate operational breaks. In practice, operations teams receive structured information instead of searching for it. Investigations can be prioritised, and routine exceptions can be resolved automatically. The role of operations teams begins to shift from manual investigation to oversight and control.

Across Southeast Asia, institutions are increasingly exploring practical uses of AI within operations. Payments investigations, reconciliation exceptions and corporate actions processing generate large volumes of operational breaks that require manual analysis. AI can help identify patterns, prioritise investigations and automate resolution. Insurance firms face similar challenges in claims reconciliation, policy administration and payment verification, where data is often fragmented across systems. Platforms such as Smartstream’s Air bring together data automation, reconciliation intelligence and AI-driven workflows to help institutions manage these processes more effectively. When combined with Smart Agents (our agentic AI solution), organisations can deploy autonomous agents that assist with investigation and resolution workflows while maintaining transparency and governance. In markets such as the Philippines, where digital financial activity continues to grow quickly, these capabilities are becoming increasingly relevant.

While the potential of AI is clear, operational leaders remain focused on governance and control. Several principles are essential. Transparency is critical. AI systems must be able to explain how decisions are reached and provide clear audit trails. Human oversight must remain possible, with defined escalation paths when judgement is required. Integration is also important. AI should enhance existing systems rather than require a complete redesign of operations. Finally, solutions must scale to enterprise volumes without adding complexity. When these conditions are in place, AI strengthens operational control rather than weakening it.

Financial operations are entering a new phase. For years, automation focused on accelerating existing processes. The next phase is about managing those processes more intelligently. Agentic AI allows operations to move from reactive investigation to proactive control. Exceptions can be resolved faster, and teams gain better visibility across the entire process. For banks and insurance firms, the operational challenges are increasingly similar. Both deal with growing transaction volumes, fragmented data environments and increasing regulatory expectations. AI offers a practical way to move operations away from manual investigation and towards automated, controlled processes.