Pierre Schurmann
Framework · Return on Tokens

ROT — is your AI actually paying?

Return on Tokens (ROT) is the unit economics of AI in production: value created per dollar of model/inference spend. Cost-side ROT measures hours saved and waste avoided against token budgets. Value-side ROT measures product velocity and revenue capture that have no floor of zero. If ROT > 0, ship; if ROT > 1, multiply deployment. Savings have a floor. Value creation doesn’t.

The flagship ROT essay — reworking “The Hidden Cost of AI Adoption” — lands here first. Mirrors on LinkedIn and elsewhere point to this page as canonical.

FAQ
What is Return on Tokens (ROT)?
The unit economics of AI in production: value created per dollar of model/inference spend — cost side (hours saved, waste avoided) and value side (product velocity, revenue capture) measured separately.
How is ROT different from AI ROI?
Most AI ROI framings stop at cost takeout. ROT splits the measurement: savings have a floor; value creation doesn’t. Reporting AI only as cost reduction understates the winning deployments and hides the losing ones.
When should you ship?
If ROT > 0, ship. If ROT > 1, multiply deployment. Measured in production — demos don’t count.
Related

Home · Writing · X thread pin: @Xurmann