The following is a guest post from Yunita Anwar, Ph.D., the Lillian Cook Braun Endowed Chair in Accounting and assistant professor of accounting at Shenandoah University. Opinions are the author’s own.
In managerial accounting, I teach students that a favorable direct labor efficiency variance means fewer hours were used than the standard allowed for the output produced. It does not automatically mean the company paid less cash. The same distinction matters when a finance team reports savings from artificial intelligence.
Grant Thornton’s Q3 2026 survey of nearly 230 U.S. finance leaders reports that 84% said AI returns were meeting or exceeding expectations. That is encouraging, but it does not show whether the return came from lower spending, avoided hiring, additional capacity or better service. Satisfaction with a project is not the same as a reconciliation to the accounts.
Benefits without an immediate cash effect can still justify an investment. Faster service, more analytical capacity and fewer errors may be valuable. The point is to name the benefit accurately, define how it will be measured and avoid calling every improvement a saving.
An accounts payable team processing 1,000 invoices a month provides a simple example. If average handling time falls from 12 minutes to 9 minutes, including review and rework, the team releases 50 hours per month. In managerial accounting terms, the team has improved its labor efficiency. The financial result depends on what happens next.
If salaries, overtime and contractor bills remain unchanged, the immediate benefit is time available for other work. The team may use it to resolve supplier disputes, analyze exceptions or absorb growth. Those uses may justify the investment, but they need an owner, a target and evidence that the capacity was actually redeployed.
Research supports the need for this distinction. The published study “Generative AI at Work” examined 5,172 customer-support agents at one firm and found a 15% average increase in issues resolved per hour. The researchers could not observe wages or overall labor demand. The study therefore shows productivity improvement, not a measured reduction in payroll, and it does not test accounting workflows.
Create a benefits schedule
Before an efficiency estimate is used to reduce a budget, CFOs should ask for a short benefits schedule that separates four different claims. The schedule should also distinguish forecasts from realized results. Revenue projects need evidence of incremental contribution rather than a time-savings proxy.
1. Realized spending reduction. Which overtime payment, contractor invoice or other cash outflow actually declined? Compare similar workloads and explain other changes. Make sure the work did not simply move to another team. A smaller departmental budget is not a saving if the cost appears elsewhere.
2. Avoided future spending. A deferred hire may have economic value, but the claim needs an approved staffing plan, a credible workload forecast and a date when the cost would otherwise have arisen. Fifty hours spread across several employees do not automatically eliminate one position. The business case should show which responsibilities no longer require additional capacity.
3. Released capacity is different. State the hours made available and the work assigned to them. A lower backlog or faster supplier response can be measured directly. If demand has not grown and no useful work has been reassigned, the capacity remains available, but it is not yet a delivered outcome.
4. Quality should also stand on its own. Define errors, rework or unresolved exceptions before the pilot starts and keep the measures stable. Faster invoice handling is not an improvement if another team inherits more mistakes. Quality belongs beside the time measure, supported by its own evidence, rather than buried inside a salary-based estimate.
These categories should not count the same hour twice. If released hours lead to lower overtime, do not also count their full salary value as a second saving. If only some of the hours support the reduction, identify the remainder and how it was used. One operational improvement may support several outcomes, but the same economic benefit cannot be added more than once.
Costs need the same discipline
Record subscription and usage charges, integration, training, monitoring, review and rework. Do not count review time twice if it is already included in measured handling time. Separate initial spending from recurring costs, and cash outlays from the opportunity cost of staff time. Both matter, but they answer different questions.
If the pilot releases time but adds a software bill, near-term cash spending may rise. That does not automatically make the project a poor investment. It means the case rests on capacity, quality or a credible future benefit. When overtime or contractor spending genuinely falls, deduct the incremental running costs before reporting a net cash benefit, and assess implementation costs over the investment’s full period.
Measurement should begin before deployment. Save the starting workload and processing-time sample, name an operational owner and a finance reviewer and compare similar work and, where feasible, retain a comparison group. Seasonal demand, simpler cases or another process change should not receive an AI label by default.
Keep the original forecast beside the result. If the case assumes growing invoice volumes, test flat demand. If software charges rise with usage, test the higher-volume bill. Update assumptions openly rather than rewriting the baseline when a pilot disappoints.
Before expanding the pilot, reconcile forecast benefits with observed results. Which spending actually changed? Where did the time go? Did quality hold? Which assumption failed? A result below target should lead to a change in deployment, workflow or expectations, not a new definition of return.
Some investments are intended to build capability rather than cut cash immediately. That is a valid objective. Approve the project on that basis, with a defined outcome and review date. Do not require a team to invent a cash-saving story to obtain funding, and do not cut its budget on the strength of one.
CFOs can support ambitious AI investment without accepting imprecise benefits claims. Fund capacity when capacity is the objective. Recognize a spending reduction when the accounts show it. A saved hour is useful. The business case should explain what happens to that hour next.
