AI Agent Costs Create New Challenges for Banks

Banks worldwide are reassessing how they deploy artificial intelligence as the cost of operating AI agents continues to increase.

Token expenses, which cover the computational units used by AI models to process information and generate responses, could reach 20% of total labor costs, according to McKinsey’s Global Banking Annual Review.

To manage the growing expense, some financial institutions have established dedicated teams known as AI FinOps. These groups oversee AI-related spending and help determine where additional investment is economically justified.

Earlier cost-control strategies relied on directing employees toward different AI models depending on the complexity of their queries. However, this approach has become less effective with agentic AI systems, which communicate with one another and can rapidly accumulate token usage.

The challenge is shifting from managing individual AI interactions to controlling the broader costs generated by interconnected AI agents.

Banks Prioritize AI Applications With Measurable Returns

Financial institutions are becoming more selective about where they deploy agentic AI, concentrating spending on applications with the strongest expected returns.

Productivity improvements and labor-related savings remain relatively straightforward to measure. However, banks are increasingly evaluating whether AI can generate additional revenue to justify higher operating expenses.

This reflects a broader change in how financial institutions assess AI investments. Rather than focusing exclusively on operational efficiency, banks are looking for applications that can contribute directly to business growth.

The potential cost burden also creates pressure to ensure that AI deployment produces measurable financial benefits rather than simply expanding technology expenditure.

Banking Customers Are Adopting AI at a Faster Pace

Customer adoption could accelerate the transformation of retail banking. McKinsey expects AI agents to become widely used within three years, considerably faster than the five to 10 years it took mobile internet to achieve broad adoption.

A survey of 46,000 retail banking customers worldwide found that many already use AI to compare financial services, identify higher-yield savings accounts and refinance or consolidate debt.

Among respondents already using AI, 72% said they would allow a banking agent to complete one financial task on their behalf. However, they generally preferred to limit autonomous approvals to $500.

The rapid adoption of AI-powered services suggests that customers may increasingly rely on automated tools to make financial comparisons and manage routine banking activities.

Transaction Restrictions Limit AI’s Impact on Banks

Despite the potential benefits, AI agents could create new financial pressures for banks.

One concern involves deposit migration. If AI agents consistently identify higher-yield savings products, they could encourage customers to move funds away from non-interest-bearing accounts into accounts that offer better returns.

Such changes could increase funding costs and put pressure on bank profitability.

For now, the impact remains limited because most banks prevent external AI agents from authorizing transactions within their systems. Potential liability remains a significant obstacle to granting autonomous systems greater financial authority.

This creates a gap between customer willingness to use AI agents and the level of access banks are currently prepared to provide.

What Investors Should Watch Next

Investors will be watching how financial institutions balance rising AI infrastructure costs against potential productivity gains and new revenue opportunities.

The development of dedicated AI FinOps teams, the expansion of customer-facing agents and the policies governing autonomous transactions will help determine how quickly the technology becomes integrated into banking operations.

The longer-term financial impact will depend not only on adoption rates but also on whether AI-driven efficiency and revenue growth can outweigh higher token expenses and potential pressure on deposit costs.