Claude Code Creator: Measure AI ROI by Human Hours Saved, Not Tokens

Boris Cherny, creator of the Claude Code tool, has proposed a shift in how enterprises measure artificial intelligence success, moving away from tracking “token burn” toward calculating human hours saved. This shift follows widespread concern among corporate executives regarding the rapid escalation of AI operational costs and the difficulty of quantifying return on investment (ROI) in software development.

The Shift from Token Consumption to Efficiency Metrics

The prevailing “tokenmaxxing” trend—a shorthand for maximizing model interaction to boost output—has hit a fiscal wall. Companies are now scrutinizing their AI budgets. According to Boris Cherny, the focus is pivoting toward the tangible economic impact of AI-assisted engineering.

Cherny’s framework suggests that usage dashboards often reflect activity, not utility. Instead, he argues that organizations should isolate specific tasks, determine if an engineer would have performed them manually, and estimate the associated labor costs.

Market Sentiment and the Cost of Innovation

The push for rationalized AI spending is not limited to software developers. JPMorgan CEO Jamie Dimon signaled a broader corporate trend during a recent CNBC interview, noting that artificial intelligence costs are rising rapidly. Dimon emphasized that companies must treat AI as a finite resource, demanding the same fiscal discipline expected of any other capital investment.

This sentiment was echoed at the Allen & Co. Sun Valley Conference, where OpenAI CEO Sam Altman confirmed that AI ROI has become a hot topic of conversation among institutional leaders.

Framework for Measuring AI-Driven Productivity

Cherny’s model for assessing success relies on a four-step framework, specifically targeting the “return” generated once teams integrate AI into their daily workflows. The core of his argument lies in the distinction between active coding and background maintenance. When AI handles routine maintenance, it frees engineering talent to focus on high-value initiatives—tasks that were previously out of reach due to resource constraints.

Rather than measuring the quantity of code generated, firms are beginning to optimize for the speed of the development lifecycle.

The Path Toward Fiscal Rationality

Companies like Coinbase and Vercel have already begun exploring cost-reduction strategies, including the use of cheaper Chinese models, to maintain output without inflating their quarterly operating expenses.