AI Summary
McKinsey's recent findings reveal that while the cost of AI models is decreasing significantly, enterprise spending on AI continues to rise. The firm emphasizes the need for companies to understand the economics of agentic AI and manage costs effectively to maximize returns.
- McKinsey's senior partners discussed the paradox of decreasing AI model costs alongside rising enterprise AI expenses during a virtual session.
- AI models, such as GPT-4, have become significantly cheaper, with some now costing less than a dollar per million output tokens.
- Despite lower costs per unit, businesses are consuming more AI capabilities, leading to increased overall spending.
- The cost of executing tasks with AI agents can vary greatly, with some tasks costing up to 30 times more depending on the approach taken.
- Companies are advised to evaluate AI agents based on task execution costs and success rates rather than just cost per token.
- McKinsey identified three key areas for managing AI spending: gaining visibility into spending drivers, optimizing workflows, and enforcing sourcing discipline.
- The firm cautioned against indiscriminate cost-cutting, urging companies to focus on maximizing returns from their AI investments.
ai modelscost managementbusiness strategymckinseyai expenses