How I Calculate Automation ROI

When a client asks me whether a makecom automation is worth building, I do not start with the tool. I start with economics, risk, and repeatability. As Theo K, founder of TK-Agency, my automation consultancy in Munich, I have seen small Business owners get excited about Make.com, zapier, n8n, ChatGPT, Claude, and every new AI-Agent idea before checking whether the workflow actually deserves to exist.

The uncomfortable truth is simple: not every task should be automated. Some processes are too rare, too messy, too political, or too dependent on human judgment. A good Automation project should save time, reduce errors, improve visibility, or unlock revenue. Ideally, it does several of these at once. My job is to separate useful Process Optimisation from technical decoration.

My makecom ROI rule: start with time, not tools

The first number I calculate is the monthly time currently spent on the process. I write down how often the task happens, how long it takes each time, and who does it. This sounds basic, but it often changes the entire conversation. A founder may say that invoicing follow-ups are a huge problem, but after calculation the task takes only 90 minutes per month. Another process, such as copying lead data from email to a CRM, may quietly consume 18 hours per month across the team.

My basic time formula is:

Monthly time cost = frequency per month x minutes per task x hourly cost divided by 60

If a manager in Germany spends 12 hours per month on manual reporting and their true hourly cost is 70 euros, the monthly labor cost is 840 euros. If an automation costs 2,500 euros to build and 150 euros per month to maintain, the project can pay back quickly. If the same automation only saves 100 euros per month, I usually recommend not building it yet.

This is where many founders make a mistake. They compare automation cost against salary only. I prefer to include opportunity cost. If the same person could spend those 12 hours on sales, client onboarding, or product improvement, the value of automation is higher than the wage saving alone.

The five questions I ask before building

Before I open Make.com, Zapier, n8n, or an AI workflow builder, I ask five questions. These questions prevent unnecessary complexity and help me estimate whether an automation has a real business case.

1. Is the process repeated often enough? Daily or weekly tasks are usually better candidates than quarterly tasks.
2. Are the inputs structured? A form submission is easier to automate than a vague email thread with missing information.
3. Is the desired output clear? If the human team cannot describe the correct result, an automation will not magically clarify it.
4. What happens when it fails? A broken Slack notification is annoying. A broken payment workflow can be expensive.
5. Can I measure the improvement? If there is no visible metric, the ROI discussion becomes emotional instead of practical.

In my experience, the best automation candidates are boring. Lead routing, invoice reminders, ticket triage, quote generation, status updates, CRM hygiene, customer onboarding, analytics snapshots, file naming, and internal approval flows are not glamorous. But they are perfect because they happen frequently and follow predictable rules.

How I compare Make.com, Zapier, n8n, and custom AI

I do not believe in tool loyalty. I believe in fit. Make.com is often excellent for visual scenario building, multi-step operations, and business users who want some visibility into the flow. Zapier is fast and convenient when a client needs a simple connection between common SaaS tools. n8n is powerful when data control, self-hosting, custom logic, or developer flexibility matters. In Germany, n8n can be especially interesting for companies that care about data residency and strict compliance requirements.

When AI enters the discussion, I slow down even more. ChatGPT and Claude can summarize, classify, draft, rewrite, extract, and reason over messy text. An AI-Agent can also decide between steps, call tools, and complete more complex tasks. But AI adds variability. That means I calculate not only time saved but also review effort, error risk, and governance cost.

For example, using Claude to summarize long support tickets before they enter Jira may save 5 minutes per ticket. If there are 400 tickets per month, the theoretical saving is more than 33 hours. But if every summary requires human correction, the saving may drop to 10 hours. If incorrect summaries cause poor prioritization, the automation can create hidden costs. I always estimate the realistic saving, not the marketing version.

My simple ROI framework

I usually calculate automation value in four layers. This keeps the decision clear and avoids overengineering.

1. Direct time saving

This is the easiest layer. How many hours disappear from the manual process? I multiply those hours by the hourly cost of the people doing the work. If an assistant saves 20 hours per month at 35 euros per hour, the direct saving is 700 euros per month.

2. Error reduction

Manual work creates errors: wrong invoice numbers, forgotten follow-ups, mistyped CRM fields, duplicate tickets, outdated spreadsheets. I estimate how often these errors happen and what they cost. Sometimes the error reduction is worth more than the time saving.

3. Speed and responsiveness

An automation that sends a lead to the right salesperson within 30 seconds can increase conversion. An onboarding flow that triggers instantly can improve customer experience. Speed has business value, especially in sales, support, and operations.

4. Management visibility

Some automations are worth building because they create reliable data. If a founder finally gets a weekly dashboard without chasing five people, decision quality improves. This is harder to quantify, but it matters.

After these four layers, I compare value against total cost. Total cost includes discovery, build time, testing, documentation, user training, tool subscription, AI API usage, monitoring, and maintenance. Many automation projects fail because people calculate only the initial build and forget the ongoing care.

The payback period I like to see

For small businesses, I like automation projects that pay back within three to six months. A one-month payback is excellent. Six to twelve months can still be sensible if the workflow is strategic, reduces serious risk, or supports growth. Anything longer than twelve months needs a strong reason.

Here is a simple example. A sales operations workflow costs 3,000 euros to build. It saves 900 euros per month in time, reduces missed follow-ups by an estimated 300 euros per month, and costs 100 euros per month to maintain. The monthly net benefit is 1,100 euros. The payback period is roughly 2.7 months. I would usually approve this project.

Now compare that with a sophisticated AI-Agent that costs 8,000 euros to build, saves 300 euros per month, and needs 200 euros per month in monitoring and API usage. The payback period is unrealistic. Even if the technology is interesting, I would recommend simplifying the workflow or postponing it.

When I recommend not automating

I say no to automation more often than some people expect. I avoid building when the process is unstable, when the team keeps changing the rules, when the data source is unreliable, or when the task requires sensitive human judgment. I also avoid automating a bad process. Automation makes a good process faster, but it can make a broken process fail at scale.

Before building, I often suggest a lightweight Process Optimisation exercise. Remove unnecessary steps. Standardize naming. Create a clean intake form. Define ownership. Sometimes these changes reduce the workload enough that no automation is needed. Other times they make the future automation much easier and cheaper.

My practical scoring model

For fast decisions, I score each opportunity from 1 to 5 across five criteria:

• Frequency: how often the process happens.
• Time saving: how much manual effort can be removed.
• Reliability: how predictable the inputs and rules are.
• Risk: how damaging a failure would be.
• Strategic value: how much the workflow supports growth, service quality, or management clarity.

A high-frequency, high-saving, low-risk workflow is an easy yes. A high-risk workflow can still be a yes, but only with stronger testing, alerts, fallbacks, and human approval steps. I like human-in-the-loop automation for sensitive workflows, especially when AI is generating text, classifying requests, or making recommendations.

Conclusion: makecom is a means, not the business case

My conclusion is straightforward: makecom, Zapier, n8n, ChatGPT, Claude, and any AI-Agent platform are only useful when the numbers and the process support the decision. I calculate time saved, errors reduced, speed gained, visibility improved, and ongoing cost. Then I look for a realistic payback period and a workflow that can be maintained without drama.

If the automation saves a few minutes once a month, I leave it alone. If it removes repetitive work, improves quality, and gives a small business team in Germany more time for customers and growth, I build it. That is how I decide whether an automation is worth building.