PortfolioC3-5

Business of AI · Gallery

Business Case / ROI Builder

SIMULATEDVerified Jul 2, 2026

Payback, NPV, and IRR are table stakes. The tornado is the difference — single-point ROI gets challenged in the room; a range that stays positive gets funded. Adoption ramp ties to EL-01.

Same instrument · three industries — pick a use-case to reconfigure the run

$600k
$1.40M
9 mo
$180k
12%
NPV · 3yr
$1.91M

@ 12% discount

IRR
143%

Break-even discount rate

Payback
0.8 yr

Undiscounted

Tornado · NPV sensitivity (±30%)

Annual value$1.03M … $2.80M
Discount rate$1.75M … $2.09M
Adoption ramp$1.77M … $2.05M
Run cost$1.78M … $2.04M

Dashed line = base NPV $1.91M. Widest bar = the driver that most moves the case.

Steering pre-read · business case

Fund

AI initiative — 3-year business case

NPV (range)

$1.03M – $2.80M

IRR

143%

Payback

0.8 yr

Recommendation: Fund. The case stays NPV-positive across the full ±30% sensitivity band. Largest lever: annual value.

Present the range, not the point

A single NPV invites a fight about the assumption behind it. A tornado shows you already stress-tested it — and names the one driver leadership should actually govern. That's what moves a case from "interesting" to "funded."

Steering-committee takeaway: I present the range, not the point. Points get challenged; ranges get funded.

How this is built

Cash flows: year 0 = −investment; year t = annual value × average adoption (linear ramp) − run cost, over 3 years. NPV discounts at the chosen rate; IRR solved by bisection; payback interpolated on undiscounted cumulative flow.

Tornado varies each driver ±30% and re-computes NPV; bars are sorted by swing and centered on the base NPV. Stack: Next.js (static) + shared design system; client-side.

Limitations: a 3-year horizon and a linear adoption ramp are simplifications; real cases model per-year ramps, taxes, and terminal value. It frames the decision and its sensitivity, not an audited model.