A McKinsey report published this week finds that across 52 agentic AI use cases in banking, the average investment payback period was approximately 10 months, while the estimated three-year return on investment was around 2.9 times but that gap between early movers and those still running pilots is likely to widen quickly.
The advantage for banks will increasingly depend on how quickly they move beyond pilot projects and integrate AI into their operations, products and customer services, the report says. Institutions that redesign entire business processes, rather than simply automate individual tasks, will be better positioned to capture the benefits of AI.
Yet the headline return figures come with a significant caveat. The report cautioned that productivity improvements do not automatically translate into higher profits. While 80 per cent of respondents in McKinsey's 2026 State of AI survey reported individual productivity gains from AI, only 6 per cent attributed a significant share of their earnings before interest and taxes (EBIT) to the technology.
Customer behaviour is also shifting the pressure on banks. A survey of around 46,000 consumers across 13 countries found that 23 per cent were already using generative AI for banking-related tasks, and among those users, 74 per cent relied on platforms such as ChatGPT, Claude or Gemini rather than their banks' own tools. Despite the growing use of third-party platforms, 62 per cent of consumers using AI for banking said they would prefer to receive AI-powered services and advice from their primary bank.
The report also flags rising costs and security risks. Spending on AI tokens, computing and operations could reach as much as 20 per cent of labour costs. It cited around USD 900 million in reported losses from AI-enabled crime recorded by the FBI in 2025.
On the workforce side, McKinsey said banks would need to build institutional intelligence by combining reliable data, customer information, internal policies and the experience of senior employees. Data readiness remains a major challenge, with 48 per cent of banking executives surveyed identifying it as their biggest obstacle to AI adoption. The report also said banks would need clear permissions, continuous monitoring and human oversight to ensure AI agents operate safely.
Source : McKinsey & Company






