Banks Must Redesign Workflows to Unlock AI Returns

Banks that deploy AI mainly as a writing, summarization or customer-service aid may capture scattered productivity gains but struggle to translate them into earnings. McKinsey says the stronger competitive advantage comes from combining proprietary transaction data and banking expertise with end-to-end workflow redesign. As revenue and loan growth slow and fintechs, neobanks and nonbank lenders contest profit pools, embedding AI across underwriting, risk, sales and operations is becoming a strategic profitability issue rather than an IT upgrade.

In an analysis published on Dec. 9, 2024, McKinsey said a typical bank has about 25 subdomains but fewer than 10 should be the first AI-transformation targets; those areas could generate 70% to 80% of total incremental value. A regional-bank trial lifted developer productivity about 40%, while multiagent systems could raise credit-analyst productivity 20% to 60% and speed decisions roughly 30%. Another bank projected a 10% revenue increase from an analytics platform supporting more than 50 models and 150 use cases. No aggregate investment amount was disclosed.