Stop Guessing. Start Calculating.
Most AI business cases are built on guesses dressed up as projections. A consultant says "AI could reduce your processing time by 40%". A vendor promises "customers who use AI see 3x ROI". An executive sponsor tells the board "this is a strategic investment." None of these are business cases. They are opinions.
Why Most AI Business Cases Fail
- ◆Vanity metrics — measuring what's easy to count (model accuracy, hours of AI usage) rather than what matters (business outcomes, cost per unit, revenue impact)
- ◆No baseline — projecting improvement without documenting the current state. If you don't know where you started, you can't know if you improved.
- ◆Cost amnesia — including the headline savings figure while omitting implementation costs, data preparation costs, ongoing maintenance, and staff retraining
- ◆Optimism bias — using vendor case studies or industry averages as proxies for your own situation, without adjustment for your actual data maturity, team capability, or operational context
- ◆No time-to-value realism — assuming benefits arrive on day one, when most AI projects take 3–6 months before measurable impact begins
A credible AI business case starts with your current-state costs, applies realistic automation multipliers based on your actual operational context, subtracts full implementation costs, and projects returns on a realistic time horizon. It produces a range — not a single number — and it names the assumptions so they can be challenged. That's what the ROI Engine in the next lesson will do for you.
What Makes a Number Credible?
A credible ROI projection has four components: a documented baseline, a realistic automation rate, a full cost model, and a time-to-value curve. The Engine applies all four — adjusted for your industry and your specific operational situation. The headline number it produces is the mid-estimate: a conservative realisation factor applied to the maximum theoretical gain.
