McKinsey tells banks that AI agents can get the cost back in 10 months
McKinsey found in 52 uses of AI agents in banks a payback time of about 10 months and a return on investment of 2.9 times.
McKinsey, a consulting firm, tells that a bank that uses AI agents can get the cost back in about 10 months. It examined 52 uses of AI agents in banks. The return on investment in 3 years is about 2.9 times the first cost. McKinsey tells that a bank must move fast from pilot projects.
The difference in profit
In the 2026 State of AI survey of McKinsey, 80% tell that AI increases the productivity of their work. But only 6% tell that AI gives an important part of their earnings before interest and tax. McKinsey tells that more productivity does not make profit.
The costs and risks
- AI tokens, computing power and operations can be 20% of labor costs.
- 48% of bank executives tell that data readiness is the primary obstacle.
- The US Federal Bureau of Investigation recorded about USD 900 million of losses from crime that uses AI in 2025.
What is not known
The report names no bank that got these results.
Insight: Banks want agents, but not yours
McKinsey tells that 23% of 46,000 consumers in 13 countries use generative AI for tasks with their bank. Of these, 74% use external AI platforms. On 17 October 2026, Citigroup told that it does not let AI agents of other companies use customer accounts. Banks want agents in the bank and external agents out, which reads as a move to keep the customer.
Sources
Posted