Calculate automation savings after allowing for human review

Measure the current process before pricing the future one

Choose a specific workflow, a clear start and a clear finish. “Admin” is too broad. “Receive an enquiry, check required details, create the CRM record and assign the right owner” is something you can observe.

Sample ordinary cases and awkward ones. Record hands-on time separately from waiting time. A request may spend two days waiting for approval while requiring only twelve minutes of staff effort. Automation might shorten the wait, reduce the effort or do both. Those are different benefits.

Include correction and rework in the manual baseline. Otherwise the comparison gives the existing process a cleaner record than it deserves. Use the same completion standard for both versions, including accuracy and customer follow-up.

A short time study will usually produce a range rather than a perfect number. Keep the range visible. Five carefully observed cases can inform a pilot, but they do not establish a dependable annual forecast for a highly variable process.

Build the calculation around residual work

Start with monthly volume multiplied by average manual handling time. Subtract the human work that remains after automation: routine review, exception handling, corrections, monitoring and maintenance.

For a simple spreadsheet, use these inputs:

1. Number of eligible cases per month.
2. Average manual minutes per case, including normal rework.
3. Share of cases that can use the automated route.
4. Review minutes for each automated case.
5. Exception rate and additional minutes per exception.
6. Monthly time spent monitoring and maintaining the process.
7. Software, usage and support costs.
8. One-off setup, training and implementation costs.

Define whether exception time is additional to routine review or includes it. Otherwise it is easy to count the same minutes twice. Keep non-eligible cases in the manual workload rather than quietly assuming the entire volume can be automated.

Also specify who does the remaining work. Ten minutes saved for a coordinator and ten minutes added to a senior specialist may not have the same operational or financial effect.

Work through an illustrative monthly estimate

The following fictional figures show how to calculate the estimate. Assume 200 eligible cases a month, each currently taking 12 minutes. The baseline is 2,400 minutes, or 40 hours.

Suppose the proposed process still requires three minutes of review for every case. That uses 600 minutes. Assume 10% of cases need an additional eight minutes of exception handling, adding 160 minutes. Monitoring and maintenance require another 120 minutes each month.

Total residual work is 880 minutes, or about 14.7 hours. The estimated capacity released is 1,520 minutes, or about 25.3 hours a month. It is not 40 hours.

At an illustrative loaded labour rate of $60 an hour, that capacity has an estimated value of $1,520. If recurring software and support cost $350 a month, the estimated net monthly capacity value is $1,170. With an assumed $4,680 setup cost, the simple capacity-value payback is four months.

This planning estimate excludes financing effects and tax treatment, and assumes steady volume and performance. Capacity value is not necessarily cash saved.

Separate time released from money retained

If the same staff remain on the same hours, releasing time does not directly reduce payroll. It may allow quicker follow-up, more careful work, less overtime or additional capacity without an immediate hire. Name the intended use.

A cash-saving case needs a plausible cash change, such as a reduction in paid overtime or a cost that will genuinely no longer be incurred. Do not label all released staff time as a budget reduction.

Likewise, extra revenue should not be assumed merely because capacity is available. The business still needs demand, the ability to deliver and a credible link between the process change and the result. Keep potential growth benefits in a separate scenario until there is evidence.

In the worked example, the four-month figure is a capacity-value calculation. If none of that capacity reduces expenditure or creates realised additional value, there is no equivalent four-month cash payback to report.

Test the assumptions that could change the decision

Review time, exception rate and eligible volume often matter more than the time taken by the automated step itself. Run a conservative case using slower review, more exceptions and fewer eligible requests. Compare that with your central estimate.

Using the same fictional volume, six minutes of review per case would consume 1,200 minutes. If 25% of cases also needed eight additional minutes, exceptions would add 400 minutes. With the same 120 minutes of maintenance, only 680 minutes, or about 11.3 hours, would be released.

At the same illustrative rate, that is $680 of monthly capacity value before the $350 recurring cost. The net becomes $330, and the simple capacity-value payback on $4,680 becomes about 14.2 months. The workflow has not changed names, but the assumptions have changed the decision substantially.

Use this sensitivity check to define what a pilot must establish. If the proposal only works when review takes under three minutes, measure review time carefully instead of focusing on an impressive demonstration.

Make the pilot answer a business question

Choose a representative sample and keep the completion criteria fixed. Record review time, exceptions, corrections, failed actions and any work shifted to another team. Include the effort spent supervising the pilot itself, while separating one-off learning from likely ongoing maintenance.

For AI-assisted tasks, documented evaluation matters as well as speed. The NIST AI Risk Management Framework provides a useful reference for measuring and managing risk. Your business still needs its own quality thresholds for the particular work.

Agree a decision rule before the results arrive. Continue if quality is acceptable and the conservative economics fit. Narrow the scope if a particular exception dominates the workload. Stop if the remaining supervision removes the benefit or the risk cannot be managed appropriately.

Keep the calculation after launch and replace assumptions with observed figures. A change in volume, software cost or review requirements can alter the result. The useful question is whether the process continues to release valuable capacity at an acceptable cost.

To discuss a bounded process rather than a broad promise of efficiency, explore Grendesign’s AI and automation services. Bring the baseline, the exception types and the work your team hopes to make room for.