Companion page for ctaio.dev/en/ai-roi. Not an official CTAIO product. No cookies. No analytics. Numbers stay in this tab.

Kill the agent rollout before week 2 if these five numbers do not clear.

CTAIO already says killing projects is the highest-leverage ROI move, and that productivity is not ROI. This card makes that rule operational for one agent workflow: a Cursor / Claude Code / similar coding agent that a team wants to fund as if it were a junior.

I built this as one HTML file because a CAIO does not need another SaaS login to fail a business case. Type the numbers. Read the formula. If it still loses money after a 25% adoption haircut, do not buy more seats.

Fully loaded junior engineer, USD / hour Agent cost per finished task, USD Tasks you actually want done per week Human review minutes per finished task Reviewer fully loaded, USD / hour Adoption haircut, percent (not 80)

Defaults are labeled guesses, not CTAIO field data. Change them. 75 and 95 are US-shaped loaded rates, not Penang ones. 25% adoption is the failure-pattern number from the ROI guide, not a promise.

…

Why this and not a chatbot wrapper

The live CTAIO site is a CAIO newsletter. The hiring card still describes an agentic-coding audience. The overlap that is actually missing on /en/ai-roi/ is a one-screen kill test for the unit that burns money in 2026: one agent task plus the human who still has to review it.

I left Gartner and MIT citations on their page. This file does not repeat them. If the formula is wrong, you can see it. If the adoption box is 80, you are lying to yourself on purpose.

Jian Hong Oo, Penang. Built for the Flywheel Motion Agentic Engineer intake, 23 Aug 2026. Formula is arithmetic, not a model.