What Automation Really Costs: The Numbers Agencies Keep to Themselves

Ask three agencies what automation costs and you get three pitch decks and no number. Here are the numbers, including the uncomfortable ones.

Why I can give them to you: we run three of our own brands, nano, mate and MUSTAX, and we built their automations ourselves. We also build for clients as an agency. So I know both sides of the invoice. Search for "automation costs" and you mostly find either "from 99 euros a month" (tool marketing) or "it depends" (agency dodge). Neither helps you. Here are the numbers: three routes with price ranges, the running costs, the hidden costs, and the cases where you should keep your money.

The three routes and what they cost

For the same automation, say a full order flow with a fulfilment provider connected, you have three routes:

The ranges are deliberately wide, because scope decides everything. A single workflow sits at the bottom end, a chain across several systems at the top. Our own experience as a reference point: for one client the n8n setup for fulfilment, tracking and support triage came to 5,000 euros, the AI customer service layer to 2,000 euros, dashboard and reporting to 1,500 euros. That is 8,500 euros together. The same client later paid 12,000 euros under time pressure, because he came to us too late. The full story is in The 10,000 Euro Mistake.

Building it yourself is the cheapest route in cash and the most expensive in time. It pays off if you or someone on your team is technical and has hours to spare. Getting started is easier than its reputation suggests: Your first n8n workflow takes 15 minutes. A production-ready order flow with proper error handling takes weeks. For us, roughly 90 percent of development time went into the question "what happens when this breaks?".

Freelancers are the middle route. Good automation freelancers charge, roughly, 70 to 120 euros an hour. The risk is availability a year from now, when an API changes and nobody knows the workflow any more.

Agencies cost the most and earn their keep when several systems have to work together cleanly and you want to hand over responsibility. Check who owns the workflows in the end and where they run. If both sit with the agency, you have bought yourself a fresh vendor lock-in.

The running costs: AI automation per month

This is where people fudge the numbers, in both directions. The realistic ranges:

• Server for self-hosted n8n: around 15 euros a month. That is our real number, and it stays the same whether a hundred or ten thousand workflows run on it.
• AI usage (API costs): around 50 euros a month for us, covering customer service across all three brands. At very high volume, more accordingly.
• The SaaS alternative: if you go with off-the-shelf tools instead of building, you pay more fast. Zapier lands at 200 to 350 euros a month under serious use, ready-made AI support tools at 300 to 500 euros per brand. The tool comparison is in n8n vs Make vs Zapier.

The difference sounds small and adds up anyway. Between self-hosted n8n and a Zapier setup, we are talking about more than 2,000 euros a year for us.

The hidden costs nobody puts in the pitch

This is the most important section of this article. The one-off costs are only the beginning.

Maintenance. APIs change, Shopify versions move on, fulfilment providers switch formats. Budget 2 to 4 hours a month per serious setup. An automation without maintenance rots, and a rotten automation is worse than none, because it quietly does the wrong things.

Monitoring. You need alerts when something jams. Building that belongs in every budget and is rarely offered. With us every error goes out as a Slack message. Without it you are flying blind.

Cleaning up the process first. Automation makes a bad process faster, and it stays bad. Documenting and straightening out the process often costs more time than building. That time appears in no quote, and you pay it anyway.

Your own effort. Even with a freelancer or an agency: briefing, test data, sign-off, follow-up questions. Plan 10 to 20 percent of the project scope as your own time.

The exit. What does it cost to switch providers? With self-hosted workflows: little. With logic locked inside a SaaS tool or at an agency: a lot. In my consulting years I saw companies pay millions to get away from one provider.

How to read a quote: five questions for every agency

When you collect quotes, ask these questions. They tell you whether someone is doing the maths properly:

• "Who owns the workflows, and where do they run?" Correct answer: you do, on your infrastructure or with full access. Anything else is lock-in announced in advance.
• "What does running this cost per month, everything included?" Anyone who only names the setup price is leaving out half the invoice.
• "Who maintains the system, and what does that cost?" Without a clear answer you pay for maintenance later, in emergency mode and at a higher rate.
• "What happens when a connected API changes?" Good providers have a process for that. Bad ones have a shrug.
• "In which case would you advise me against this?" Anyone without an answer here will sell you everything, including the wrong thing.

These questions cost you five minutes and sort providers more reliably than any reference list.

When it is NOT worth it

Yes, we make money with automation. We still advise against it regularly. In these cases, keep your money:

• Too little volume. If a task costs you less than 2 hours a week, building for it rarely pays. At 5 orders a day, handling them manually is completely fine.
• The process keeps changing. Startup still figuring things out? Then you build something that is obsolete next month and pay twice.
• Nobody can look after the system. Without capacity for those 2 to 4 maintenance hours a month, any investment turns into a ruin.
• A failure would threaten your business. Automation has an error rate, and ours is not zero either. Where a mistake makes you liable, put a human in the loop instead of more automation.

Damian wrote the longer version of this list: When automation makes no sense.

The other side: what not automating costs

Fairness means looking at the other side too. Manual processes cost you founder hours, and those are your most expensive asset. 20 hours of operations a week, at a conservative founder hourly rate of 100 euros, is around 8,000 euros of opportunity cost a month. Add the cost of errors: wrongly shipped orders barely happen any more since we automated.

For us every automation paid for itself after 2 to 3 months, some after 2 weeks. Treat that as a realistic order of magnitude, without any guarantee for your case. The full calculation is in The true cost of manual processes.

Conclusion: the key takeaways

• Realistic one-off costs: building it yourself costs time instead of money, a freelancer roughly 1,500 to 6,000 euros, an agency roughly 5,000 to 15,000 euros depending on scope.
• Running costs are small when you self-host: for us around 15 euros of server plus about 50 euros of AI per month.
• The hidden costs are maintenance, monitoring and cleaning up your processes. Skip them in the budget and you pay them later with interest.
• Under 2 hours saved per week, with unstable processes, or without maintenance capacity: leave it.
• Always do the maths on both sides: the cost of automating against the cost of muddling through.

If you want to know what your specific case would cost, come talk to Flowhouse. You get a number, no pitch deck. And if it is not worth it, we will tell you that too.