KPMG Report: Nearly Half of Executives Scale Back AI Agents Over High Costs
A KPMG report reveals that nearly half of business executives have scaled back or paused their AI agent deployments. The primary driver behind this decision is the unexpectedly high operational cost associated with running these autonomous systems. This trend highlights a shift from initial generative AI hype to economic reality, forcing companies to seek more cost-effective solutions. It could accelerate the adoption of open-source, local Large Language Models (LLMs) that offer lower operational costs than proprietary APIs. While AI agents can automate complex workflows, their reliance on continuous LLM queries and token-based pricing models often leads to unpredictable and unsustainable expenses. Consequently, organizations are re-evaluating their ROI metrics before committing to full-scale production.
## BACKGROUND
AI agents are software systems designed to autonomously pursue goals and complete multi-step tasks on behalf of users by leveraging generative AI. However, running these agents requires frequent calls to Large Language Models (LLMs), which are typically priced based on the volume of input and output tokens. This architecture can quickly accumulate high operational costs, especially when using proprietary models like GPT-4 or Claude.