Frontier Radar #3: How agentic AI is turning tokens into a business metric

Editor's take

A new analysis highlights how autonomous agentic AI workflows, which can run for hours and consume significantly more tokens than traditional chat interfaces, are posing a challenge to current subscription-based business models for generative AI providers.

This shift is critical because it directly impacts the economic viability of AI services. Companies like OpenAI and Anthropic, which rely on per-token pricing or tiered subscriptions, may find their current revenue streams unsustainable as agents necessitate a move towards flat-rate, high-consumption models. The implications are broad, affecting developers building these agents and users who will experience potentially higher or different cost structures.

Future developments to monitor include the emergence of new pricing structures that accommodate agentic AI's higher token throughput, potentially through time-based billing or a tiered system that accounts for sustained, autonomous operation. The success of this transition will hinge on whether providers can offer predictable, affordable pricing that encourages widespread adoption of these more powerful AI capabilities.