Demo · Kansas City · AI strategy
Where does AI pay off first? Count the months to proof.
A use case pays off only after someone can show it worked. That takes longer when the decision is rare, when the world is slow to grade it, and when the outcome is noisy. This calculator runs the time-to-proof model from Eight Industries, One Question for sixteen use cases across eight Kansas City industries. Click any use case to change its inputs.
Months to prove it worked
Bar length uses a log scale so a 1-month and a 70-year proof fit on one chart. Dark: the fast use case in each industry; grey: the slow one.
Time saved is not money saved
Payback from a proven use case depends on how much of the gain the business actually captures: absorbing growth without hiring, cutting overtime, clearing a backlog. P = 12 × F / (k × g × W − C).
The model
Time to proof T = A + L + max(0, N − H) / r. A is setup time (a week), L is the feedback lag, N the units the test needs, H history already on hand for a fair backtest, and r the units produced a month. For outcomes on a scale, N per arm = 2 × 7.85 × (CV / δ)², where CV is the coefficient of variation and δ the relative improvement; paired backtests drop the 2 and use one arm. For event rates, N per arm = 7.85 × [p₀(1 − p₀) + p₁(1 − p₁)] / (p₀ − p₁)². 7.85 is (z₀.₉₇₅ + z₀.₈₀)², for a 5% two-sided test with 80% power.
Proposed model with illustrative inputs: volumes describe one mid-sized Kansas City organization per industry, effects come from published field studies where one exists, and the rest are stated assumptions. Treat every result as a lower bound. Full sources and discussion are in the article.