Technology and Innovation Community

Recent post by the Team @ McKinsey focusing on AI agents pay off.

Link to analysis here.

Key details:

1. Where are the most significant economic opportunities emerging?           

• Agents can change the unit economics of personalized service in areas such as mass-affluent banking, for example, where tailored relationship support has historically been difficult to justify at scale.
• For AI to generate attractive economics, workflows need both sufficient scale and enough automation potential for value to compound over time - some of the strongest opportunities sit in high-value workflows with large pools of repeatable work where small improvements compound over thousands or millions of transactions.

      2. What drives the unit economics of AI-based workflows?           

• Variable costs. Token costs can vary significantly based on the task. High-compute tasks, like generating code, have significant token costs. But for an agent performing a customer service task in banking, token costs frequently represent just 20 to 25 percent of the variable run costs of an AI agent. Human oversight, on the other hand, accounts for 70 to 75 percent of the variable costs. Functional and risk experts often tend to perform these oversight tasks. In banking customer onboarding, for example, we would expect 10 to 20 percent of agentic runs to be reviewed by risk and functional experts.
• Fixed costs. AI infrastructure and agent orchestration costs are fixed per agent and make up the rest of the annual run costs of an agent. AI infrastructure costs include public cloud containers, memory, management, and analytics, while agent orchestration includes the data scientist capacity required to maintain and enhance the agent in production.

      3. A new wave of implications and questions:          

• When should AI costs be passed through to customers or absorbed by the organization?
• How much pricing flexibility do business leaders need to build into their budgets?
• How do current budgeting, governance, and operating disciplines (internal chargebacks, cost coding) need to evolve to account for AI economics?
• How should cost and pricing decisions be made as AI agents work increasingly across functional boundaries?