US companies begin limiting AI usage amid soaring costs

As companies face higher-than-expected cost issues in the process of adopting generative AI, they are entering a 'rationing' phase in which they adjust or limit AI usage.

The main response strategies are as follows:

• Usage control: Instead of giving unlimited access to all employees, AI usage is allocated based on job role and necessity.
• Model switching: Instead of costly large-scale models, cheaper small language models (SLMs) optimized for specific tasks are used to reduce costs.
• Efficiency improvement: Focus is placed on prompt optimization and workflow redesign to reduce the number of API calls.

This shows that companies have moved past the technical experimentation stage of AI and entered an operational stage where they must demonstrate real ROI (return on investment).