If someone tells you their AI project delivered 60% ROI, what two numbers are missing from that claim?
Financial ROI is measurable, defensible, and backward-looking. But the decisions that determine whether an AI investment was actually worth making are often taken three years before any ROI number is calculated. Return on Efficiency (ROE) measures the operational changes AI enables — time saved, errors reduced, capacity freed. Return on Future (ROF) measures the strategic position you didn't have before — new capabilities, new markets, reduced dependence on processes that won't scale. Most AI business cases present only ROI. The ROE and ROF are why the project was actually worth doing.
Financial return: cost savings, revenue uplift, measurable in dollars. Calculated as (net gain ÷ cost) × 100. The only lens that appears in the board presentation — which means it's the only one that's defended and stress-tested.
Risk of using alone: a project with 60% ROI but zero ROF may be optimising a process that AI will make irrelevant in three years.
Operational return: how AI changes the way work gets done — speed, accuracy, headcount freed for higher-value tasks. Harder to put a single number on, but it's where most of the day-to-day value is actually realised.
Risk of ignoring: if ROE isn't tracked, the team never knows whether the AI is actually being used — or whether it's running in the background while people work around it.
Strategic return: the capabilities and options the investment creates. New data assets, new market positioning, reduced dependence on legacy processes, the ability to move faster than competitors when the next model drops.
Risk of ignoring: ROF is the main reason to invest early — and the hardest to defend in a budget meeting. Which is why most organisations invest late.
A regional insurance company deploys an AI platform to process and triage claims. Budget: $250,000. The platform reduces manual processing time by 40%, flags potential fraud at 3x the rate of human reviewers, and creates a real-time data asset that the business had never had before.
| Lens | What you measure | What this project shows | The number |
|---|---|---|---|
| ROI | Net financial gain ÷ investment cost | $150K in labour savings + $90K in fraud prevention = $240K year 1 gain on $250K spend | ≈ −4% year 1 (negative — project looks bad) |
| ROE | Operational change: speed, accuracy, capacity | Claims processed in 2 days instead of 6. Fraud detection rate tripled. 4 FTEs redeployed to complex cases. | Significant — measurable even in year 1 |
| ROF | New capabilities and strategic position | First insurer in region with real-time claims data. Ability to price risk dynamically. Platform scales to 10x volume without headcount. | Unquantifiable — but the reason competitors are now behind |
ROI is measurable, auditable, and comparable across projects. It wins in budget processes because finance teams can put it next to every other capital allocation request and rank them. The problem is that AI investments often have a different return profile than traditional IT projects:
Non-financial returns — customer satisfaction, brand perception, sustainability positioning — are also legitimate returns from AI. They're harder to measure, not less real.
Know which return you're maximising before you build the business case. A 60% financial ROI from a project that built zero strategic capability may be worth less than a −4% ROI from one that fundamentally changed what the organisation can do.
What do ROI, ROE, and ROF each measure — and why does using only ROI create a blind spot?
Why are AI returns described as "probabilistic rather than deterministic" — and what does that mean for how you evaluate them?
The insurance company example shows a year-1 ROI of roughly −4%. Should the project have been approved? Justify using all three lenses.