How much can a business save with automation? How to calculate it using real data

Quick answer: a business should not estimate savings using a generic percentage

It is common to see claims such as “automation reduces time by X%” or “AI cuts costs by Y%”. The problem is that those percentages do not describe your process.

The same automation can have a huge impact in a business that performs a task 3,000 times a month and be almost irrelevant in another that does it ten times.

To calculate savings, we need data from the current process:

• number of executions per week, month or year;
• manual time per execution;
• time that will still be required after automation;
• approximate cost of the work involved;
• initial implementation cost;
• licences, APIs and recurring maintenance;
• errors, delays or other costs that can actually be measured.

With those variables, we can move from “automation saves time” to an estimate that can be compared with the investment.

Money savings, time savings and recovered capacity are not the same

This is one of the most important distinctions when calculating return.

Direct cash savings

This exists when automation reduces an expense the company genuinely stops paying: overtime, outsourced work, usage of a tool that is removed or other costs that are directly avoided.

Recovered time

This is the number of hours the team no longer spends on the process. It can be measured even when there is no immediate cash saving.

Recovered productive capacity

If someone recovers ten hours a month but is still paid the same, the company has not reduced payroll by ten hours. It has recovered ten hours that can be used for customers, sales, operations or work it previously had no capacity to handle.

Costs avoided through fewer errors or delays

They can form part of the benefit when enough data exists: repeated work, penalties, incidents, lost opportunities or administrative errors with an identifiable cost.

I would not add these benefits if they are only an assumption. A conservative calculation is usually more useful than an impressive ROI built on assumptions that cannot be verified.

Basic formula: how much time can automation save?

We can start with a simple formula:

Hours saved = number of executions × (manual time − time after automation)

If time is measured in minutes, divide by 60 afterwards.

Quick example:

• 200 executions per month;
• 6 manual minutes per execution;
• 2 minutes of review after automation.

The automation saves 4 minutes per execution:

200 × 4 minutes = 800 minutes = 13.33 hours per month.

Those 13.33 hours are the starting point. We should not yet call them “money saved” until we know how the company uses that time.

How to calculate the value of an hour to estimate the benefit

For an initial estimate, we can use the approximate business cost of the person or resource performing the task.

The important thing is not to confuse net salary with the real cost of an hour to the company.

Depending on the level of precision available, the calculation can include:

• employer wage cost;
• associated employment costs;
• external services linked to the task;
• specific costs that disappear after automation.

If the objective is only to compare candidate processes, even a reasonable internal hourly rate can be useful as long as we use the same criterion for all of them.

Then:

Monthly value of recovered time = hours saved × hourly cost.

Which costs should be deducted before claiming an automation is profitable

Gross savings are not the real return. Automation also costs money and time.

I would separate at least two groups:

Initial costs

• workflow analysis and design;
• development or configuration;
• integrations;
• testing;
• data migration or preparation when required;
• training or launch.

Recurring costs

• tool licences;
• API usage;
• use of AI models where applicable;
• hosting or infrastructure where applicable;
• maintenance;
• human supervision time;
• adjustments caused by changes in external systems.

An automation that saves €300 a month but costs €280 a month in tools and maintenance is very different from one that achieves the same saving with €30 of recurring cost.

How to calculate ROI and when the investment pays back

Once benefits and costs are quantified, we can calculate return on investment.

ROI (%) = ((total benefit − total cost) / total cost) × 100

For an annual calculation, total cost should include both implementation and the recurring costs for those twelve months.

It is also useful to calculate the payback period:

Months to recover the initial investment = initial cost / monthly net operating benefit

These formulas are useful as long as the input data is reasonable. If we estimate unrealistic times or count money as savings when the company does not actually stop spending it, the result may look mathematically precise while still being wrong for decision-making.

Example: calculating savings and ROI for a simple automation

This example is fictional and is only intended to explain the method.

A business receives 120 requests per month. Recording the data, creating the task and sending the confirmation takes 8 minutes per request.

After automation, a person spends an average of 2 minutes reviewing each case.

The approximate payback period for the initial investment would be:

€850 / €219 = 3.88 months.

Over twelve months:

• value of recovered capacity: €264 × 12 = €3,168;
• initial cost: €850;
• recurring cost: €45 × 12 = €540;
• total first-year cost: €1,390;
• estimated net benefit: €3,168 − €1,390 = €1,778;
• estimated first-year ROI: approximately 128%.

Again, this example does not mean your business will achieve that return. Its purpose is to show which data we need to calculate it using your real process.

7 mistakes that can artificially inflate automation savings

1. Counting all current time as time saved

Many automations still need review, exception handling or human intervention. That time must be deducted.

2. Automatically converting recovered hours into cost reduction

If payroll expenditure remains the same, we are talking about recovered capacity, not necessarily less money leaving the bank account.

3. Ignoring licences and API usage

A workflow that is cheap to develop can still have significant operating costs if it runs frequently.

4. Forgetting maintenance and external changes

APIs, tools and processes can change. Total cost should account for the reasonable ongoing operation of the solution.

5. Using the best-case scenario as the normal forecast

If the possible range is between four and eight minutes saved, building the entire ROI on eight minutes can create an overly optimistic expectation.

6. Monetising benefits that cannot be demonstrated

“Better experience”, “less stress” or “more control” can be real benefits, but they should not be converted arbitrarily into euros.

7. Measuring only the first month

At the beginning there may be learning, corrections and exceptions. Measure enough executions before assuming performance will remain stable.

How to measure savings before and after automation

The calculation improves significantly when a baseline exists beforehand.

Before implementation, I would record over a reasonable period:

• number of executions;
• average time per execution;
• errors or repeated work;
• waiting time;
• people involved;
• associated direct costs.

After automation, I would measure exactly the same variables, adding:

• successful automated executions;
• cases that required intervention;
• review time;
• workflow errors;
• actual recurring costs.

This approach matches the way enterprise automation platforms are incorporating savings measurement: comparing a manual baseline duration with workflow executions and then applying a financial value.

The process with the highest potential saving should not always be first

An automation can have high potential savings and, at the same time, be too complex, risky or unstable to become the first project.

That is why this page answers how much can be saved, but it does not replace prioritisation analysis.

If you already have several candidates and need to decide which one to tackle first, read the guide on which processes to automate first .