How to Avoid Double-Counting AI Value

By Paige Gilmore, Founder, NetLift · Published 2026-07-28 · Updated 2026-07-28

Avoid double-counting AI value by strictly separating realised savings from future projections and subtracting the full cost of adoption, including rework and training, from gross labour value.

Double-counting AI value typically happens when organisations conflate theoretical productivity gains with realised time savings, or fail to account for the hidden costs of rework and training. To maintain finance-grade accuracy, value must be measured as the delta between a verified baseline and the actual time taken to complete work with AI tools.

By focusing on current net value—labour value minus the total cost of implementation—finance leaders can isolate true ROI. This approach ensures that future recurring savings are stated separately, preventing the inflation of current year performance with speculative projections.

Why is AI value frequently overstated?

Value inflation often occurs when teams apply a flat productivity percentage across an entire department without verifying specific work outputs. If an organisation counts the 'potential' time saved by every licence holder rather than the actual time saved on tracked work, the resulting figure is usually an estimate rather than a financial fact. Overlap between different AI tools also leads to counting the same saved hour multiple times across different software budgets.

How do you separate realised value from future projections?

To avoid double-counting, you must distinguish between what has happened and what is expected. Realised value represents the labour value of time already saved. Future value is a projection based on expected recurring time savings multiplied by expected volume. By reporting these as two distinct figures, finance teams prevent 'paper gains' from being treated as actual budget relief. This distinction allows the business to see exactly when an investment has reached its payback point.

What costs must be deducted to find net value?

Gross labour savings do not represent true value until all adoption costs are subtracted. A complete net value calculation must deduct licence fees, usage costs, implementation time, and initial training. Critically, it must also include the time spent on review and rework. If an AI generates a draft in seconds but requires an hour of human correction, the net time saved is the only figure that should enter the value model.

How does Evidence Quality impact value reporting?

Not all data points are equal. Value calculations based on self-reported estimates are less reliable than those derived from objective baselines, such as historical data or cohort comparisons. Assigning an Evidence Quality grade to each measured area helps stakeholders understand the confidence level of the ROI. Verified data from large sample sizes provides a more stable foundation for expansion decisions than small-scale, anecdotal evidence.

NetLift ensures accuracy by applying a deterministic value model to tracked work, comparing time taken against objective baselines to calculate true time saved. By grading Evidence Quality from 'Estimate Only' to 'Verified,' NetLift provides CFOs with a transparent view of net value that accounts for all costs, including rework. This focus on work-level measurement avoids the pitfalls of employee surveillance while providing the data needed to decide whether to Expand, Continue, or Stop specific AI initiatives.

Frequently asked questions