Kofi Ndaikate has spent his career at the intersection of traditional finance and cutting-edge technology, helping institutions navigate the complex shift toward automation. As an expert in Fintech policy and blockchain, he understands that the true value of artificial intelligence lies not in its raw power, but in how seamlessly it integrates with human expertise. Today, we explore how the concept of a “dual workforce” is reshaping banking, moving beyond simple automation to a more sophisticated, orchestrated model of digital partnership.
Financial institutions often find that managing dozens of specialized, isolated AI tools creates significant coordination hurdles. How does implementing an orchestration layer resolve these integration challenges, and what specific steps are required to ensure these disparate agents align with existing banking regulations?
The primary friction in modern banking isn’t a lack of AI capability, but the “silo effect” where narrow tools operate in total isolation. By implementing an orchestration layer, or an Agentic Operating System, we create a centralized command center that assigns specific roles to AI agents rather than letting them run as disconnected features. This layer acts as a traffic controller, ensuring every automated action is mapped to a permission framework and a specific internal governance standard. To stay compliant, the system must be designed to follow pre-set banking regulations, meaning every output is validated against current policy before it reaches a human or a client. It effectively turns a chaotic collection of tools into a unified, auditable team that speaks the same language as the bank’s compliance department.
Many banking analysts currently spend the majority of their time on financial spreading and covenant testing. When transitioning these data-heavy tasks to digital partners, what metrics should leadership track to measure success, and how does this shift fundamentally change the analyst’s daily decision-making process?
When we shift financial spreading and covenant testing to digital partners, the most critical metrics move from “volume processed” to “time to insight” and “accuracy of early warnings.” Leadership should track how quickly a digital partner can flag a covenant breach compared to manual monitoring, as well as the reduction in administrative hours per loan file. For the analyst, the daily grind shifts from the sensory fatigue of manual data entry to the high-level work of strategic interpretation. Instead of hunting for numbers across hundreds of pages of documentation, the analyst spends their energy evaluating the risk profile presented by the AI. This allows them to focus on complex credit assessments that require human intuition rather than just calculating debt-service coverage ratios.
Generative AI projects frequently stall in the pilot phase because they are treated as standalone features rather than integrated components. What are the practical requirements for embedding AI directly into a bank’s broader operating model, and can you share an example of a workflow that successfully bridged this gap?
To move past the pilot phase, AI must be treated as a colleague, not a gadget; this requires embedding it directly into the bank’s core operating model. Practically, this means the AI needs to be role-based—acting as a “digital partner” that mirrors existing functions like loan origination or transaction management. A successful workflow involves the AI handling document validation and financial data extraction in the background, then surfacing those results directly within the existing software the human team uses. For instance, in a commercial lending workflow, the AI might automatically process thousands of documents to fulfill closing conditions. Because the AI is an integrated part of the deal team rather than a separate tab on a browser, it eliminates the “integration tax” that usually kills new technology initiatives.
A dual workforce model relies on human employees and AI systems working alongside each other in a coordinated environment. How can organizations define the boundaries between human intuition and automated analysis, and what strategies prevent the automation of administrative tasks from creating new bottlenecks in loan origination?
In a dual workforce, the boundary is defined by the difference between data processing and relationship stewardship. We assign the repetitive, analytical heavy lifting—like monitoring activities and condition clearing—to the digital partner, while reserving strategic decision-making for the human professional. To prevent new bottlenecks, organizations must ensure that the “hand-off” between AI and human is frictionless and real-time. If an AI flags a compliance issue during loan origination, it shouldn’t just send an email; it should update the workflow dashboard with an explainable reason for the flag. This ensures that the speed of automation isn’t lost when a human needs to step in to provide the final sign-off or manage a sensitive client interaction.
Regulatory expectations for transparency require that all automated banking actions remain auditable and explainable. How do you design a system that maintains these governance standards at scale, and what role does an “Agentic Operating System” play in documenting the logic behind a digital partner’s specific decision?
Maintaining governance at scale requires that every single action taken by a digital partner leaves a “digital breadcrumb” that can be reviewed months or years later. An Agentic Operating System serves as the historical record-keeper, documenting the specific logic and data points the AI used to reach a conclusion. This means if a loan is flagged or a transaction is questioned, the system can provide an auditable outcome that explains the decision-making process in plain English. By embedding these compliance checks directly into the automated workflow, we ensure that transparency isn’t an afterthought or a manual reporting burden. It creates a “compliance-by-design” environment where the bank can prove to regulators exactly why and how a digital partner acted.
Client-facing teams are increasingly using real-time insights to inform their conversations with customers. How do these digital tools change the preparation process for relationship managers, and what specific types of customer data provide the most value during live interactions?
Relationship managers are moving away from spending hours prepping for meetings and toward a model of “just-in-time” intelligence. Digital tools can surface relevant customer information—such as recent transaction patterns or potential product needs—right when the conversation is happening. This shift allows the manager to be more present and responsive, as they don’t have to memorize a client’s entire history or dig through multiple screens during a call. The most valuable data during these live interactions includes early warning signals on credit health and real-time insights into the client’s current operational volume. It transforms the relationship manager from a salesperson into a proactive consultant who can offer solutions before the client even realizes they have a problem.
What is your forecast for the evolution of digital partners in modern banking?
My forecast is that the “standalone AI” era is ending, and we are entering the era of the permanent dual workforce. Within the next few years, I expect digital partners to become a standard, non-negotiable component of every high-performing banking team, much like how Excel became standard decades ago. We will see these agents move beyond back-office support to become deeply specialized experts in areas like complex document processing and predictive risk modeling. Ultimately, banks that successfully embed these orchestrated systems will manage higher volumes with less operational overhead, while their human employees find more fulfillment by focusing on the strategic, high-touch interactions that technology cannot replicate.
