Agentic AI Goes Mainstream, Fueling Growth in the Token Economy

AI is no longer limited to conversation. It has started doing the work.

For years, artificial intelligence was essentially a tool that produced polished answers. Humans asked questions, machines responded, and people still had to take action themselves. AI was a fast, fluent reference tool, but execution remained in human hands.

That framework changed in February this year, when so-called agentic AI entered the commercial mainstream. Once given a goal, agentic systems can break work into smaller tasks, select the appropriate tool at each stage, retry when errors occur, and deliver a final result. In response to commands such as "book me a flight," "handle this customer complaint," or "write the code and test it," AI is no longer just an assistant. It has become the entity carrying out the task.

Usage data shows the shift. AI token consumption has risen sharply since February. Agentic tasks consume five to 30 times more tokens than standard conversations because they execute an entire workflow rather than generate a single paragraph. Goldman Sachs projects global token consumption will be 24 times higher by 2030 than in 2026, with nearly all of that growth driven by agentic use cases. Salesforce disclosed that it processed 2.4 billion agentic tasks last quarter, up 57% from the previous quarter. Processed tokens totaled 19 trillion, five times the level a year earlier. Those are live operating figures, not pilot-program data.

Corporate results are also reflecting the shift. Anthropic's annual recurring revenue was revised higher to more than $40 billion from about $30 billion in just one month. OpenAI is also gaining ground, with its autonomous coding service Codex ranking near the top in both developer benchmarks and field evaluations. The two companies' combined revenue is projected to reach $100 billion soon. The token economy, long treated as a theoretical concept, has begun generating measurable, recurring revenue. For companies with large foundation models, demand for agentic AI is translating directly into sales growth and increasing their appeal to investors.

Meaningful changes are also emerging in Asia. AI companies in the region are rising quickly as they develop competitive in-house models and capture sovereign AI demand from governments. In robotics and autonomous driving, commercial services have moved beyond the demonstration stage. This is no longer a period defined by flashy announcements. It is a period in which working systems are being built. In this environment, new investment opportunities are opening up for component suppliers and systems integrators.

Woo Gun, manager at Manulife Asset Management