Reviewed and updated October 8, 2026.
Calculate AI ROI from realized benefits and full costs, while showing how uncertain each input is. A reduction in handling time is a capacity gain; it becomes a cash saving only when spending changes or another evidenced benefit is realized.
Specify the evaluation period and perspective. Include build and integration, internal staff effort, licenses or inference, infrastructure, review, support, evaluation and incident work. Avoid counting the same benefit as labor saving, capacity and revenue.
For a defined period, ROI = (realized benefit − total cost) ÷ total cost. This arithmetic does not prove that AI caused the benefit or that future results will match. Show the inputs, sources and excluded effects alongside the figure.
Separate observed results from modeled scenarios. If no cash saving has occurred, report capacity or a conditional estimate instead of describing it as realized return.
A hypothetical workflow processes 2,000 eligible items in a period. It reduces handling by three minutes per item, including review and corrections. That represents 100 hours of modeled capacity. At an assumed $50 loaded hourly rate, its labor-equivalent value is $5,000.
Suppose actual recurring system and support costs are modeled at $2,000 for the same period. The capacity-value ratio would be ($5,000 − $2,000) ÷ $2,000 = 150%. This is illustrative arithmetic, not a client result or realized cash ROI. Initial build costs are excluded from that short-period ratio and must be included for a lifecycle decision. If staff time is not redeployed or spending reduced, the $5,000 is not automatically a financial benefit.
Compare comparable work with and without the change where feasible. Account for volume, case mix, staffing, seasonality and other changes. HM Treasury's impact-evaluation guidance emphasizes the importance of an appropriate counterfactual for causal claims. It is public-policy guidance; the comparison principle is adapted here for a business workflow.
A before-and-after change alone may be informative but weak evidence of causation. State that limitation. Do not credit AI for an improvement produced by a simultaneous policy or staffing change without further evidence.
Use the ledger to decide whether to maintain, narrow, expand or stop the system. Do not assume benefits compound automatically. Expansion can introduce different cases, permissions and support work. Recalculate using the evidence for that scope, preserving the distinction between a plausible opportunity and a realized result.
Bring the cost ledger, baseline and claimed benefit. A review can separate a measurable capacity gain from realized financial return and attribution. Explore ALLTIPLY AI roadmap and enablement, or request a project review with the workflow, systems and constraints you need to assess.
Optional Google Analytics stays off until you choose Accept analytics.
