Maveric executive says banks need operating model shift for AI

Banks should treat artificial intelligence as a change to their operating model rather than as another digital channel, according to Venkatesh P, co-founder and director of Maveric Systems. In a Finextra community opinion piece, he said the shift could affect onboarding, financial-crime controls, customer service, personalisation and software engineering.

Venkatesh argued that banking technology over the past two decades has focused largely on making existing processes faster and cheaper through core modernisation, internet banking, mobile journeys and omnichannel servicing. He contrasted that with an AI-led model in which systems use context and data to infer a likely need or risk before a customer or employee completes a conventional workflow.

From self-service to fewer customer handoffs

In customer servicing, Venkatesh said digital banking often transferred operational effort from branches and contact centres to customers, who still had to complete forms, upload documents, correct errors and wait for updates. He said AI could reduce that burden by drawing on customer relationship management systems, registries, third-party data providers, behavioural information and existing bank records.

Onboarding is one area where he sees immediate relevance. Venkatesh said banks face a persistent tension between regulatory checks and customer expectations for speed. In his view, AI could help by sending low-risk applicants through straight-through processing while directing higher-risk cases to compliance staff, with explanations for why a case was escalated.

He linked that approach to know-your-customer, anti-money-laundering and financial-crime programmes. Venkatesh said these functions are often run through separate tools and data sets, creating duplicated information and fragmented alerts. He argued that AI architecture could allow banks to combine structured and unstructured information in a single risk infrastructure, with potential benefits including fewer false positives, lower fraud losses and reduced integration complexity over time.

Customer engagement and product development

Venkatesh also drew a distinction between multichannel banking and what he called multimodal intelligence. Rule-based chatbots, he said, can fail when a customer falls outside a predefined decision tree. A multimodal AI system could, in his account, move between text, voice, images and video, adjust an explanation when a customer appears confused, and transfer context, sentiment and possible resolution steps to a human agent.

On personalisation, Venkatesh said banks have long pointed to fragmented data as a barrier, but he argued that static analytical models are also a constraint. He said AI systems can detect signals across salary patterns, rental changes, location context, product usage and digital behaviour to infer life-stage needs before a customer states them directly. He described this as a move from reactive cross-selling to anticipatory engagement.

Inside banks, he said AI should be considered across the software development lifecycle, rather than limited to code assistance. He cited possible uses in backlog creation, user stories, code generation, unit tests, integration mapping, documentation, regression testing and self-repairing test scripts. Venkatesh said banks need faster release cycles while maintaining engineering discipline in a regulated sector.

Venkatesh concluded that chief information officers will need to measure AI programmes beyond uptime, cost and delivery speed. He said banks should also track turnaround time, revenue productivity, fraud reduction, explainability, fairness, reliability and compliance as they assess whether AI-led systems can preserve trust while changing how banking processes are run.

This story draws on original reporting from Finextra Research.