The Jevons Paradox of Tokens

By Sam M. Sweilem. Cheaper AI does not simply reduce cost. It unlocks new classes of work.

Efficiency breeds consumption

When a useful resource becomes dramatically cheaper, organizations usually consume more of it, not less. In AI, lower inference cost makes previously impractical workflows economically viable.

The enterprise mistake

The wrong budgeting question is, "How much will AI save?" The better question is, "What new operating capability becomes possible when reasoning, summarization, classification, and software generation can run continuously?"

The strategic implication

AI leaders should plan for capability expansion. The most important returns may come from workflows that were never attempted under the old cost structure: continuous QA, deeper personalization, faster engineering loops, and more complete compliance evidence.