How to Calculate Automation ROI: Inputs, Costs and a Worked Example

To calculate automation ROI, you need two honest numbers. One is what the automation gives you back each year, the other is what it costs to build and keep running. ROI is the benefit minus the cost, divided by the cost. Payback is the upfront cost divided by the net amount you get back each month. The formulas are simple. The hard part is the inputs, because ROI figures get inflated by counting time that is never really freed and leaving out the costs that arrive after launch. This guide shows how to calculate automation ROI from your own data, works through one example, and lists the mistakes to avoid.

Measure the task as it runs today

Pick one task, not a department. Then write down, from your own records, how long one run of the task takes, how many times it happens in a normal week, how many people do it and how often it goes wrong. Time it on a few real days rather than estimating from memory. If the task is invoice entry, count invoices and minutes per invoice. If it is lead follow-up, count leads and minutes per lead.

Also note what happens when the task goes wrong today. A wrong price on an order, a missed follow-up or a duplicate payment has a cost you can usually trace in your own books. Those costs belong in the benefit side only if you can point to them.

Put a fully loaded cost on the hours

An hour of staff time costs more than the hourly wage. The US Bureau of Labor Statistics reports that compensation costs for private industry workers averaged $46.89 per hour worked in June 2026, of which wages and salaries were $32.82 and benefits $14.07, so benefits made up about 30 percent of the total. If you value saved hours at the wage alone, you undercount them.

Use your own payroll figures where you have them. The BLS average is a national figure across all private jobs, so a bookkeeper in your city may cost more or less. What matters is that you include benefits and payroll costs, not only the wage.

Count every cost, including the ones after launch

The cost side is where many ROI estimates are too thin. A guide on process automation ROI from BOC Group lists licensing, implementation, system integration, training and ongoing maintenance, and warns that many initiatives underestimate costs or overlook parts of the value. For a typical small business workflow, the list looks like this.

• The build, whether you pay a developer or count your own team’s hours.
• The platform subscription for the automation tool.
• Usage charges that grow with volume, including any AI model calls made on each run.
• Maintenance, because connected apps change their APIs, forms change and edge cases appear.
• Someone’s time to check failed runs and handle the cases the automation passes back.

Usage is easy to underestimate because tools count it differently. Zapier counts tasks, Make counts credits and n8n counts executions, and the same workflow can produce very different numbers on each. Our comparison of how n8n, Zapier and Make count usage walks through one workflow on all three, which is the number you need for the running cost line.

Decide what the saved time really becomes

This step is easy to skip. Saved hours only turn into money if they become something. They can avoid a planned hire, cut overtime or temp costs, or move a person to work that brings in revenue you can name. If none of that happens, the hours are spread across the week and the cash cost of the team stays the same.

So split the saved hours into two lines. Count as hard savings only the hours tied to a cost that actually goes away. Show the rest as freed capacity, and say what it will be used for. A plan that rests on capacity can still be a good decision, but it should not be presented as cash.

The BOC guide also points out that automation runs existing logic, so automating an unclear or unstable process can make its problems bigger. If the task changes every week or nobody agrees how it should work, fix the process first and calculate ROI afterwards.

A worked example

The figures below are invented round numbers to show the arithmetic. They are not prices, quotes or results from a real project.

Say two people spend a combined 10 hours a week entering supplier invoices, for 50 working weeks a year. That is 500 hours. At the BLS average of $46.89 per hour, the task costs about $23,445 a year. You expect automation to remove 70 percent of that time, because some invoices will still need a person, so the yearly benefit is about $16,412, and you have confirmed that the hours replace overtime you currently pay.

On the cost side, assume a $6,000 build, a $150 monthly platform and usage cost, which is $1,800 a year, and two hours of maintenance a month at the same hourly cost, which is about $1,125 a year. First-year cost is about $8,925.

First-year ROI is ($16,412 minus $8,925) divided by $8,925, which is about 84 percent. For payback, take the yearly benefit minus the running and maintenance costs, about $13,486, and divide by 12 to get about $1,124 a month. The $6,000 build pays back in a little over five months. From the second year, without the build cost, the return is higher, as long as the volume and the process stay the same.

Now change one input. If the invoices drop to half the volume, the benefit halves but most costs do not, and the first-year ROI falls below zero. Run your estimate at a lower and a higher volume before you decide, because a result that only works at one volume is fragile.

Mistakes that make automation ROI look better than it is

The common ones are easy to spot once you know them. Counting 100 percent of the task time as saved, when exceptions still need a person. Valuing freed hours as cash when no cost goes away. Leaving out maintenance and failed-run handling. Adding revenue gains with no evidence behind them. And comparing the automation with a full salary when nobody’s job is actually changing.

AI agents need extra care here. An agent that answers customers still needs a person for complaints, unusual questions and anything it is not allowed to answer, and that time is a running cost. Our guide to setting up an AI customer service agent covers planning that handoff, which you should cost before you count any savings. If you are still deciding whether a task needs AI at all, the section on when plain automation is enough in our guide to intelligent automation is a useful check, since a rule-based workflow can be cheaper to run.

Run your own numbers

Start with one task, your own timings and your own payroll costs, and calculate ROI and payback at two or three volumes. If your team is in the UAE, our free AI ROI calculator takes salary, visa, insurance and workspace costs in AED. If you want a second opinion on which task to automate first, or on the cost side of a quote you already have, book a free strategy call and bring your numbers.