TL;DR: Measure human work removed, subtract review and recovery time, then apply the right labor rate. In this example, 120 claimed hours become 70 net hours. Payroll savings remain $0 until an actual expense falls.
Your automation dashboard says it saved 120 hours last month. Your payroll is unchanged, the operations manager is busier, and nobody can name the work those hours replaced. The workflow may still be useful. The savings claim needs more evidence.
Automation time savings and where to measure them
Automation time savings are the human work removed from a defined process, after counting the human work the automation adds. They are different from a shorter wait, a faster software run, or a smaller payroll bill. Track those outcomes separately so the same improvement does not become three benefits.
A workflow automation connects triggers, rules, and actions across software. A customer relationship management system, or CRM, stores customer and deal records. Measure staff effort across the whole handoff between those systems.
The broader business process automation ROI framework covers the investment decision. This article audits one input: the value assigned to saved staff time.
| SMB workflow | Human work to measure | Work that is easy to miss |
|---|---|---|
| E-commerce order entry | Checking and entering each order | Address fixes, failed syncs, duplicate orders |
| Home-service job completion | Preparing the invoice from job notes | Missing photos and technician follow-up |
| B2B sales administration | Updating CRM records | Matching contacts and correcting ownership |
| Client onboarding | Collecting approved account details | Chasing missing documents and reviewing exceptions |
| Weekly reporting | Preparing and checking the report | Investigating wrong totals and restoring failed feeds |
In each case, measure work through to a usable result, including corrections after the software finishes.
How to calculate time savings from process automation
Compare the human time needed for the same mix of cases before and after automation. Include routine review, exceptions, manual fallback, and upkeep across every affected role. The difference is net time savings; reduced waiting time belongs in a separate service-speed measure.
For a simple batch with similar cases, use this formula:
Baseline human hours = cases × baseline human minutes per case ÷ 60
Net hours saved = baseline human hours − all after-automation human hours
Calculate each case type and role separately, then add the rows. Cases left on the old process have no savings unless their work also changed.
Follow these six steps to make the comparison usable:
- Define a completed case. Use the order, invoice, or CRM record ID. Set a quality standard; retries keep the same case ID.
- Measure active work. Log case type, role, touch minutes, exceptions, and status in Excel or Google Sheets. Observe staff or run a back-office workflow audit.
- Match the workload. Apply baseline minutes by case type to after-period volumes. Compare similar order sizes, document quality, and peak periods.
- Record every route. Join Power Automate or Make logs to case IDs. Include review, fallback, failed runs, and later corrections across teams.
- Reconcile by role. Keep clerk, manager, and specialist rows separate. Include maintenance from the owner's calendar.
- Check quality and reuse. Compare errors and completion rates. Name the replacement work; recheck normal and busy cycles before annualizing.
Use weighted averages by case group; a median can hide costly exceptions. For unfinished cases, count work to date and keep the estimate provisional. Follow the same cohort through completion before claiming final savings.
Robot runtime cannot replace this log. UiPath's automation analytics guidance distinguishes execution measures from business outcomes and identifies errors, exceptions, queues, and logs as useful evidence.
What belongs in fully loaded labor cost?
Fully loaded labor cost includes wages plus the employer costs needed to employ that person, using a clearly stated scope. Include employer payroll costs and benefits; identify any overhead allocation separately. An allocated cost can help compare capacity without being an expense automation will remove.
US private-industry employer compensation averaged $45.65 per hour worked in June 2025. June 2025 private-industry benefits averaged $13.58 per hour worked, 29.8% of total compensation. Those are national reference figures, not your clerk's rate. Both come from the BLS release published September 12, 2025.
| Cost line | Capacity valuation | Cash-saving claim |
|---|---|---|
| Salary or regular wages | Include the relevant role's cost | Count only a supported reduction in paid expense |
| Employer payroll taxes and benefits | Include actual employer costs once | Include only the portion that actually falls |
| Overtime or hourly contractors | Track separately from regular staff | Use the avoidable bill, subject to minimum commitments |
| Rent, shared software, general overhead | Show separately if allocated | Count only a canceled or reduced expense |
| Owner time | Show an explicit opportunity-value assumption | Do not invent a payroll saving for unpaid owner work |
Some businesses call this burdened labor cost. Have the payroll owner confirm the included costs and hours before presenting a precise rate.
Which hours belong in the loaded labor rate formula?
Divide annual employer cost by the hours that match the work you are valuing. Paid hours, hours worked, and productive task hours answer different questions. Choose one denominator, document exclusions, and use the same basis before and after automation.
Here is a hypothetical clerk rate, with all amounts in USD. The annual salary is $52,000, and employer payroll costs and benefits add $14,560. The total is $66,560; these are planning inputs, not a tax estimate or a market salary benchmark.
| Hour basis | Illustrative calculation | Resulting rate | What it means |
|---|---|---|---|
| Paid hours | $66,560 ÷ 2,080 | $32.00/hour | Cost spread across scheduled paid time |
| Hours worked | $66,560 ÷ 1,840 | About $36.17/hour | Excludes 240 paid leave hours |
| Productive task hours | $66,560 ÷ 1,664 | $40.00/hour | Also excludes 176 hours of meetings and other non-task work |
The productive-hour rate spreads annual cost across time available for tasks. It does not make a recovered hour worth $40 in cash or remove the cost of meetings that still take place.
BLS compensation concepts define cost per hour worked with leave excluded from hours. That differs from this example's narrower productive-task denominator. Do not apply another utilization markup to an outside benchmark without checking its basis.
If annual salary includes vacation pay, do not add those wages again as a benefit. Excluding leave from available hours while adding the same pay again inflates the rate.
Use productive hours across all tasks assigned to the role. Dividing the employee's entire annual cost by just this workflow's hours wrongly assigns unrelated work to the project.
An automation time saving calculator with real deductions
Human review and recovery reduce savings even when the automated steps run correctly. In the following operator composite, 120 gross hours become 70 net hours after all new human work is included. The model separates roles because a manager's added hour costs more than a clerk's recovered hour.
Hours Saved That Are Not Real: Loaded Labor Math for Automation ROI
A fictional 12-person service company handles 1,200 completion records monthly. This is a That'sGonnaHelp operator composite, not a public customer claim or measured client result. Each similar record previously took a clerk eight minutes to check, enter, and prepare for billing: 160 hours in total.
Microsoft Forms sends intake data to SharePoint; Power Automate prepares the accounting record through an approved connector. Staff approve it before invoicing. The team pilots one record type and logs usable billing records against their source IDs.
Only 900 records use the new route; 300 stay manual. The old eight-minute baseline produces the dashboard's 120-hour claim for those 900 records. New review and repair steps are missing from that figure.
The first reconciliation finds two minutes of clerk review on every automated record. Ninety records also need eight extra minutes to fix missing job details. Those extra minutes are additional to routine review, while the unchanged 300 records still require 40 hours of manual work.
The manager spends eight hours monthly checking failures, maintaining mappings, and coordinating fixes. The team adds this missing work to the ledger. A failed run completed manually retains its original case ID and all recovery time, so it cannot become a second successful case.
The corrected month therefore uses 82 clerk hours and eight manager hours, compared with 160 clerk hours before automation. All 1,200 records reach the same billing-ready standard in this illustration. Net team capacity improves by 70 hours, while the clerk alone releases 78 hours.
The owner can assign only 60% of those 78 clerk hours to a documented backlog that needs the same skills. The rest arrives in fragments or at the wrong time, so the budget assigns it no value. With a $40 clerk rate, a separate assumed $75 manager rate, and the costs below, the model produces $852 in monthly net capacity value, a 7.4-month economic payback estimate, and no cash payback.
| Monthly hours reconciliation | Calculation | Hours |
|---|---|---|
| Original work across all cases | 1,200 × 8 ÷ 60 | 160 |
| Baseline work displaced on adopted cases | 900 × 8 ÷ 60 | 120 |
| Clerk review on adopted cases | 900 × 2 ÷ 60 | 30 |
| Extra clerk exception work | 90 × 8 ÷ 60 | 12 |
| Unchanged manual cases | 300 × 8 ÷ 60 | 40 |
| Total clerk work after launch | 30 + 12 + 40 | 82 |
| New manager upkeep | Separate time log | 8 |
| Net team hours released | 160 − 82 − 8 | 70 |
The clerk's gross recovered capacity is 78 × $40, or $3,120 a month. Subtracting the manager's 8 × $75 leaves $2,520 before software and services. Multiplying the net 70 team hours by the clerk's rate would give $2,800 and overstate the result by $280 because it prices the manager's work incorrectly.
Apply the same role-based reconciliation to an automated CRM data-entry ROI calculation when several people touch each sales record.
Which saved hours become cash?
Saved hours become cash savings only when an actual expense falls relative to a credible baseline. Examples include fewer paid overtime hours or a smaller contractor invoice. Reassigned salary time is useful capacity; postponing a justified future hire is cost avoidance with a date and a hiring assumption.
The example has no payroll, overtime, or contractor reduction. Its clerk hours belong in the capacity ledger; cash labor savings are $0.
GAO's savings-reporting review found that unclear savings definitions and inconsistent cost deductions weakened federal property reports. That is a measurement lesson, not proof of SMB automation returns.
USD cost and ROI worksheet
These are hypothetical budget inputs, not market benchmarks or a vendor quote. The ranges test cost uncertainty; the base case drives every calculation below.
| Cost or value line | Illustrative planning range or basis | Base case |
|---|---|---|
| Reused clerk capacity | 78 hours × 60% × $40 | $1,872/month of capacity value |
| Manager upkeep | 8 hours × $75 | $600/month of internal capacity cost |
| Software and usage | $200–$450/month for licenses, connectors, and usage | $300/month cash |
| External support | $75–$200/month maintenance allowance | $120/month cash |
| External setup | $3,000–$7,000 for build, mapping, and testing | $4,800 one-time cash |
| Internal setup | 15–30 manager hours at $75/hour | 20 hours; $1,500 one-time capacity cost |
For a dated price reference, Microsoft June 2025 guide lists Power Automate Premium at $15 per user/month in USD, billed annually. The same guide lists Process at $150 per bot/month, also billed annually. These cover different licensing needs; neither is the all-in cost of the sample integration. Verify the required users, bot scope, connectors, and usage against the June 2025 licensing guide and current terms before buying.
The $300 base allowance includes software; do not add the reference prices again. Manager upkeep is also counted once below.
Monthly net capacity value = $1,872 − $600 − $300 − $120 = $852
Total setup value = $4,800 + $1,500 = $6,300
Economic payback proxy = $6,300 ÷ $852 = about 7.4 months
This is an economic proxy, not money arriving in the bank. When checking it with an automation ROI calculator, label benefits as assigned capacity value.
For 12 full operating months at those assumptions, benefits equal $22,464. Full costs equal $18,540: $12,240 of recurring costs plus $6,300 of setup. The resulting full-cost economic ROI is ($22,464 − $18,540) ÷ $18,540, or about 21.2%.
The separate cash view is less cheerful: no cash savings, $4,800 setup, and $5,040 in recurring bills. Net cash impact is negative $9,840 across setup and those 12 operating months. These calculations assume stable volume and immediate steady performance after setup; a launch-year budget must add ramp-up time and any extra launch costs.
Automation ROI metrics that survive a downside case
The reuse assumption can change the decision without changing the software. At 30% reuse, recognized clerk value drops to $936 a month. After $1,020 of upkeep and external costs, net capacity value is negative $84, so the model has no positive payback.
The break-even reuse share is $1,020 ÷ $3,120, or about 32.7% of recovered clerk time. This is a result of the example's inputs, not a benchmark. Recheck volume, exception work, role rates, and reuse separately; GAO's cost-estimating guidance supports documenting assumptions, testing sensitivity, and updating estimates with actual costs.
Track net hours by role, usable hours reassigned, quality failures, and actual expense reductions. Pair them with cost per verified outcome when volume changes. A lower labor burden per completed case can be useful even if total monthly work rises.
What public productivity research can establish
The November 2023 revision of Generative AI at Work studied 5,179 customer-support agents. That study reports a 14% average increase in issues resolved per hour, with 34% for novice/low-skilled workers and minimal effects on experienced/high-skilled workers. It does not establish payroll savings. These findings are from the NBER working paper, not the fictional service company above.
A 14% throughput increase is also not a 14% time reduction for a fixed workload. Under a constant-workload assumption, the arithmetic is 1 − 1 ÷ 1.14, or about 12.3% less time. That conversion is an illustration of units; it does not predict your team's improvement.
When this framework says no
Do not approve a labor-saving claim when the work was never measured, the output is worse, or the recovered time has no credible use. Faster processing can still produce negative ROI once review, maintenance, and setup are included. In those cases, repair the estimate or justify the project using a different measured outcome.
Three limits matter most. Low or irregular volume may not recover the setup effort. A workflow that shifts work onto a scarce manager can create a more expensive bottleneck. And fragmented time may improve working conditions without supporting a cash-saving budget.
Common mistakes that create fictional hours
- Counting elapsed time as labor. A record waiting overnight is not eight staff hours of work. Measure active touches and report the wait reduction separately.
- Applying adoption twice or not at all. If the ledger already contains only adopted cases, do not multiply it by adoption again. If it contains all cases, include unchanged manual work.
- Ignoring failed or unfinished cases. Short successful runs are an incomplete sample. Follow the same cohort through fallback, correction, and completion.
- Using one rate for every role. Keep each role's hours and loaded rate together, including time moved to managers or specialists.
- Counting the same benefit twice. Do not add salary capacity value to the incremental profit earned by using those same hours. Choose the benefit method that the evidence supports.
FAQ
Time savings need a work record; cash savings also need an expense record. Use these distinctions when comparing proposals or updating a budget.
What does loaded labor rate mean?
A loaded labor rate assigns employer cost to an hour of work. To compare proposals, ask each supplier for its numerator and denominator. A $40 productive-hour rate and a $32 paid-hour rate can describe the same employee, as the worksheet shows.
How do you calculate automation ROI?
The full-cost automation ROI formula is (recognized benefits − full costs) ÷ full costs × 100. Check the denominator in any supplied spreadsheet: some models divide only by setup cost. Those percentages are not directly comparable with the full-cost ROI used here.
Should saved salary hours count as cash savings?
No, while the same salary expense continues. If a paid expense later falls, retain the payroll change or supplier invoice and its effective date. Recognize the reduction from that date, not retroactively from the automation's launch.
Should I subtract bot runtime from manual work time?
Only count runtime as human work when a person must actively attend it. An unattended ten-minute run consumes compute time, which belongs in operating costs, but not necessarily ten labor minutes. Count any setup, waiting that prevents other work, monitoring, and review the person actually performs.
Can saved minutes across several people count as an avoided hire?
Only if those minutes cover the needed skills, schedule, location, and workload. Ten scattered hours do not automatically replace a ten-hour staffing gap. Support cost avoidance with a dated hiring plan, expected demand, and a named owner who confirms the gap is covered.
How much evidence is enough before annualizing savings?
Use enough observations to cover normal work, common exceptions, and peak conditions; there is no universal case count. Record the sample period and case mix, then test a lower-volume or higher-exception scenario. Keep uncertain inputs as ranges rather than converting one good day into a full year.
Does saved time have value when no cash expense falls?
Yes, if it clears a backlog, improves service, or reduces overload. Give that outcome an owner and a measure, such as overdue cases completed. Keep its assigned value separate from booked expense savings.
Answer clarity notes
The source facts and the worked example serve different purposes. Treat the public figures as dated evidence and the operator composite as a calculation to replace with your own inputs.
- Dates: The article's December 31, 2025 date is distinct from the BLS June 2025 measurement period, its September 12 release, and the NBER November 2023 revision. Microsoft prices come from its June 2025 guide; check current pricing and licensing before buying.
- Evidence: Linked sources support public facts. The service-company case is a fictional That'sGonnaHelp operator composite, not a public customer claim or a promise drawn from measured client results.
- Money: All amounts are USD. Case costs, labor rates, reuse percentages, ROI, and payback are planning assumptions or calculations. Ranges, examples, savings, and timelines are not guarantees. National compensation averages are not local wage quotes or guaranteed savings.
- Cash versus capacity: The 70 net hours describe team capacity. The 7.4-month payback and 21.2% ROI use assigned economic value; the example has no cash payback and a negative $9,840 cash impact over setup plus 12 operating months.
- Scope: This framework supports US SMB operating decisions. It is not accounting, investment, legal, or tax advice; payroll inputs and any formal financial reporting treatment need review by the appropriate qualified professional.
Sources
These public sources support the cited statistics, measurement concepts, and historical price references. The worked arithmetic is That'sGonnaHelp's illustrative model.
- BLS: Employer Costs for Employee Compensation, June 2025, released September 12, 2025.
- BLS: Employer Costs for Employee Compensation concepts, updated September 30, 2025.
- NBER: Generative AI at Work, November 2023 working-paper revision.
- GAO: Cost Estimating and Assessment Guide, March 2020.
- GAO: Federal Real Property savings-reporting review, October 2013.
- UiPath: Automation analytics and operational versus business measures.
- Microsoft: Power Platform Licensing Guide, June 2025.
Have an hours-saved number you cannot reconcile to the workday? That'sGonnaHelp can help trace it back to cases, staff effort, and a business case your team can check.