AI Agent Costs Could Reach 20% of Labor Expenses as Banks Deploy New Cost Control Teams

Banks Confront Unprecedented AI Agent Costs

Banks worldwide are grappling with rapidly escalating AI agent costs as token consumption threatens to consume 20% of total labor expenses, according to McKinsey's Global Banking Annual Review.1

1

This financial pressure has prompted institutions to establish specialized teams called AI FinOps to monitor and control technology spending.3

3

The challenge extends beyond simple cost management—banks must now reconsider technology budgets while maintaining competitive advantage in an AI-driven marketplace.

Traditional Cost Controls Fail Against Agentic AI Systems

Initial attempts to manage AI in banking expenses proved inadequate when dealing with agentic AI systems. Banks first deployed models that directed employees to different AI platforms based on query complexity. "But those models are ineffective when AI agents are working, because they communicate with each other and tokens quickly add up," explained Vik Sohoni, senior partner at McKinsey.1

1

This inter-agent communication creates exponential token consumption that traditional monitoring systems cannot effectively track or limit, forcing institutions to develop entirely new approaches to managing these expenses.

Strategic Deployment Replaces Blanket AI Adoption

Financial institutions are becoming more selective about where they deploy generative AI, focusing on applications with measurable productivity benefits and revenue benefits. Rather than implementing AI across every department, banks now prioritize areas offering clear returns on investment.3

3

Productivity-related savings remain relatively straightforward to measure, but institutions increasingly examine AI-driven revenue growth to justify higher spending in specific areas, according to the McKinsey report.1

1

Customer Adoption Outpaces Previous Technology Waves

Retail banking customers are embracing AI faster than any previous technology cycle. McKinsey expects AI agents to achieve widespread adoption within three years, compared to the 5 to 10 years mobile internet required.1

1

A survey of 46,000 retail banking customers across 13 countries revealed that 23% frequently use generative AI for financial tasks including comparing services, finding higher-yield accounts, and managing debt.2

2

Among AI users, 72% indicated willingness to allow banking agents to complete at least one financial task autonomously, though most would cap approvals at $500.1

1

Third-Party Platforms Capture Banking Interactions

A concerning trend for traditional institutions: 74% of customers using AI for banking rely on third-party platforms like ChatGPT, Claude, and Gemini, compared to just 37% using their bank's chatbot.2

2

This shift means consumer choices increasingly happen outside bank ecosystems, with large language models becoming a new intermediary layer between institutions and customers. Despite this migration, 62% of consumers using AI for banking would prefer receiving AI-powered services directly from their primary bank, suggesting opportunity remains for institutions that move quickly.2

2

AI Disruption Dividend Promises 20% Cost Reduction

While managing AI agent costs presents immediate challenges, the Global Banking Annual Review projects AI could cut banking industry's cost base by 20% or more for institutions that fundamentally rewire their operations.2

2

Most banks have thus far only deployed AI to boost individual employee productivity, creating negligible bottom-line impact. Institutions investing significantly in AI and restructuring processes stand to capture what McKinsey calls the AI disruption dividend—creating new organizational models with strong controls, institutional intelligence, and rapid change management.2

2

Profitability Risks Emerge Alongside Opportunities

AI agents present a double-edged sword for bank profitability. While they promise to reduce labor expenses and boost revenue, these systems could pressure margins by shifting customer deposits from non-interest-bearing accounts into higher-yielding options.1

1

Currently, this risk remains contained because most banks block external AI agents from authorizing transactions within their systems. However, as customer experience expectations evolve and competitive pressures mount, institutions may need to grant broader access while managing the financial implications.

Early Movers Gain Potentially Insurmountable Lead

The speed of customer adoption creates urgency that differs from previous technology cycles. "Early movers will have the advantage of being able to invest in better designed and incentivized products, harvest productivity, and offer better rates and experiences to their customers," noted Sohoni. "Unlike in past technology cycles, a late start may not offer chances to catch up."2

2

Banks must position themselves strategically within the industry ecosystem, match customer adoption rates, and build infrastructure for controls while maintaining unprecedented speed. The institutions that successfully balance AI agent costs with strategic deployment may establish market positions that competitors cannot easily challenge.