Commercial banking analysts have long split their time between data assembly and credit judgement, but research from nCino suggests the industry may be misreading where the real value of the role actually sits.
Agentic AI is shifting this balance by taking on the manual, multi-system work that dominates credit operations, allowing analysts to focus on evaluation, interpretation and decision-making. It represents a new layer of automation that sits above traditional workflows, changing both how work is done and what work remains for humans.
McKinsey estimates banking operations staff spend around 80% of their time on coordination and rule-based tasks, including gathering data, reconciling systems and preparing credit documentation, leaving limited time for judgement-led analysis.
Agentic AI differs from earlier automation tools by executing multi-step workflows independently. Unlike robotic process automation (RPA), which relies on fixed scripts and often fails when inputs vary, agentic systems can pull data across platforms, adapt to inconsistencies, structure outputs and flag exceptions without constant human input. Early deployments are already reducing manual workloads by 30% to 50%, according to McKinsey.
The impact is most visible across three commercial lending workflows: credit review preparation, covenant monitoring and portfolio risk tracking, all of which rely on fragmented data and repetitive manual processes. Credit review preparation is often the most time-intensive. Analysts typically pull data from multiple systems before analysis can begin. Agentic AI compresses this by consolidating and standardising inputs, resolving inconsistencies and producing structured credit packs ready for review.
McKinsey’s QuantumBlack team estimates AI-assisted credit memo workflows could improve productivity by more than 60%, delivering annual savings of over $3m, though gains depend heavily on data architecture. Fragmented systems reduce efficiency by forcing AI to replicate manual reconciliation work.
Covenant monitoring is shifting from periodic checks to continuous oversight, with agentic AI flagging potential breaches as new data becomes available. Portfolio risk management is also moving toward real-time visibility, surfacing concentration risks and early warning signals as they develop.
However, the effectiveness of agentic AI remains constrained by data fragmentation. Institutions with unified commercial data can enable end-to-end automation, while those with siloed systems often see manual work redistributed rather than removed.
This makes data architecture a key determinant of value. The difference between automation that eliminates work and automation that reshapes it is often set before deployment.
As adoption grows, the analyst role is being redefined rather than replaced. Agentic AI can assemble data, monitor portfolios and surface exceptions, but judgement, interpretation and accountability remain human responsibilities, particularly in regulated environments.
Over time, analysts are moving closer to the point of decision. As data assembly is automated, expertise is increasingly concentrated in interpretation and judgement — the core of credit work.
The Monday morning credit review still begins the same way. The difference is that the package now arrives pre-assembled, with exceptions already flagged, and analysts spend more time interpreting signals than building datasets.
For more insights, read the full story here.
Copyright © 2026 FinTech Global
