How to Calculate the ROI of AI Automation

“This automation saves ten minutes” sounds valuable. But if it runs four times per month, costs $99, takes an hour to maintain, and still requires five minutes of human review each time, the business may be paying more than it saves.

That is why the useful question is not “How much time can AI save?” It is “What is the net monthly value after the workflow's real costs?”

1. Start with monthly task volume

Count how often the task actually happens. Do not use the number of employees who could theoretically use the automation or the number of features the tool could eventually support.

If a proposal is created 25 times per month, use 25. If customer-call summaries happen 80 times per month, use 80.

Low volume can kill an otherwise impressive business case. Automation has fixed costs, so the task needs to occur often enough for the savings to accumulate.

2. Measure minutes saved after review—not before it

Time the current process first. Then time the automated process from beginning to end, including review, corrections, copying, approvals, and exception handling.

If a report takes 18 minutes manually and the AI draft plus review takes 8 minutes, the net saving is 10 minutes. It does not matter that the model generated the draft in six seconds.

3. Put a realistic value on the time

Use the approximate hourly cost of the person doing the work. That might be wages plus payroll burden, a blended internal rate, or another consistent figure your business already uses.

The goal is not accounting precision. It is avoiding the mistake of treating every saved hour as if it were worth an arbitrary premium consulting rate.

Worked example #1: customer follow-up drafts

Imagine a small service business sends 60 follow-up emails per month.

• Current time per email: 9 minutes
• AI-assisted time including review: 4 minutes
• Net saving: 5 minutes
• Monthly volume: 60
• Hourly labor value: $32

The monthly time saved is 60 × 5 ÷ 60 = 5 hours. At $32 per hour, the gross monthly time value is $160.

That is the value before technology and maintenance costs—not the ROI.

4. Add every recurring workflow cost

The monthly subscription is only one cost. Include anything required to keep the workflow functioning.

Do not allocate the entire price of a tool to one workflow if you are genuinely using it across several workflows. But do not pretend the cost is zero just because another department also uses it.

5. Convert setup time into a payback calculation

Setup is usually a one-time cost: mapping the process, configuring software, building an integration, preparing examples, training employees, and fixing the first round of errors.

Keep setup separate from monthly cost so you can calculate how long the project takes to pay back.

The core ROI formula

For a simple operating workflow, use this:

Then:

If recurring cost is very small, ROI percentage can become huge and not particularly useful. In that case, focus on net monthly value and payback period.

Worked example #2: the full calculation

Take the follow-up workflow above. It produces $160 per month of gross time value.

Now assume:

• Software allocation: $39/month
• Automation-platform allocation: $15/month
• Maintenance: 30 minutes/month at $32/hour = $16

Total recurring cost is $70/month.

Monthly net value is $160 − $70 = $90/month.

Monthly ROI is $90 ÷ $70 × 100 = about 129%.

If setup cost $270 in internal time and implementation effort, the payback period is $270 ÷ $90 = 3 months.

That is a much more useful statement than “AI saves five hours per month.”

Break-even volume: how often must the task happen?

Break-even volume tells you how many successful workflow runs are needed each month to cover recurring cost.

Suppose each completed use creates $3 of net time value before software costs and the workflow costs $90 per month. You need roughly 30 successful uses per month just to break even.

If the task only occurs 12 times per month, the automation may still have non-financial value, but the labor-savings argument does not work.

Do not automatically count all saved time as cash

Saving ten hours does not necessarily mean the business earns ten hours of wages back. If the employee is salaried, the financial benefit depends on what happens to the recovered capacity.

That does not make the savings fake. It means you should describe the value accurately.

If you cannot explain what happens to the saved capacity, keep the business case conservative.

Error reduction can matter more than time

Some workflows are valuable because they reduce missed follow-ups, incomplete records, duplicate entry, or avoidable rework. You can include that value if you can measure it without making heroic assumptions.

For example, if a missing-field problem causes ten hours of monthly rework and the pilot consistently cuts that to four hours, the six-hour difference can be valued using the same labor method.

Avoid assigning imaginary dollar values to “better decisions,” “innovation,” or “employee happiness” just to make the spreadsheet turn green. Those may be real benefits, but they should remain qualitative unless you have evidence.

Use three scenarios instead of one forecast

AI workflow performance is rarely perfectly predictable. Build three versions of the business case.

If the project only makes sense in the strong scenario, the business case is fragile. A first automation should ideally remain reasonable under conservative assumptions.

A five-minute ROI worksheet

Then calculate:

1. Net minutes saved: current minutes − automated minutes.
2. Monthly hours saved: volume × net minutes ÷ 60.
3. Gross monthly time value: monthly hours saved × hourly value.
4. Monthly net value: gross value − recurring monthly cost.
5. Payback months: setup cost ÷ monthly net value.

If monthly net value is zero or negative, there is no payback period. That is not a failed analysis—it may have saved you from buying something.

When ROI should not be the only decision

Some workflows involve security, legal obligations, customer trust, employee decisions, safety, or other risks that cannot be justified by a positive spreadsheet alone. A process can have attractive ROI and still be a bad automation candidate.

Use the 20-point AI readiness checklist alongside the financial model. If the project has unresolved red flags, fix those before optimizing the return.

The sequence we recommend

First, choose a task with the Automation Fit Test. Next, decide whether it needs AI or simpler automation. Then score readiness, run a small pilot, and calculate ROI from the pilot's actual review time and failure rate.

Only after that should you use the AI tool buying framework to decide whether the software is worth keeping.