Agency, Freelancer or Build It Yourself: The Decision Guide

We are an agency, and I am still going to tell you when you do not need one. That is pure self-interest: a project that does not fit the client costs both sides more than it brings in.

At nano, mate and MUSTAX we built everything ourselves for years. At Flowhouse we build for clients. So I know both sides of the bill, including the places where it does not add up. Here are the three paths with cost, speed and risk.

The three paths at a glance

The ranges are deliberately wide. A single workflow sits at the low end, a chain across several systems at the high end. The full breakdown including running costs is in What automation really costs.

Path 1: Build it yourself

What it really costs. No money, but time. And more of it than the tutorials suggest. The first workflow takes 15 minutes. A production-ready order flow with error handling takes weeks. For us, around 90 % of the development time went into one question: what happens when something goes wrong?

What you get for it. Understanding. That is the underrated part. Once you have built your own workflow, you can read every quote properly afterwards. You know what two days of work looks like and what two weeks looks like.

When it fits. You or someone on the team enjoys technical work. You have time before peak season. The process is yours and it is not critical if it sits idle for a day.

When it does not fit. When you are the bottleneck. I regularly see founders still tinkering with a workflow in month three while the process keeps running manually. The 3,000 euros for a freelancer would have paid for themselves long ago.

Our own route was in-house, but under a condition that rarely gets mentioned: we had the technical background and a time buffer. Without both, building it yourself is the most expensive path of all. If you want to start anyway, the entry point is lower than its reputation suggests: Your first n8n workflow.

Path 2: Freelancer

What it really costs. Good automation freelancers charge roughly 70 to 120 euros an hour. For a clearly scoped project you land at 1,500 to 6,000 euros. On top comes your own effort for the brief, test data and sign-off. Budget 10 to 20 % of the project scope for that.

What you get for it. The best price per hour of expertise. A freelancer has no overhead, no project management layer, no sales cost.

When it fits. The scope is clear. You can say in one sentence what should be running at the end. You can do the sign-off yourself, meaning you can check whether it actually works.

The real risk. It shows up next year. When an API changes and the freelancer has taken a job somewhere, you are left with a workflow nobody understands any more. Two things protect you: short documentation as part of the sign-off, and access to everything under your own accounts. Nothing may live on their server.

A warning sign. A freelancer who sends a fixed-price quote without asking questions has not understood the process. Good people ask about the edge cases first.

Path 3: Agency

What it really costs. Roughly 5,000 to 15,000 euros one-off depending on scope, plus ongoing support. For context, a real example of ours: n8n setup for fulfilment, tracking and support triage 5,000 euros, AI customer service layer 2,000 euros, dashboard and reporting 1,500 euros. Together 8,500 euros.

What you get for it. Accountability on top of code. The difference becomes visible when something breaks. An agency has a process for that, someone to cover, and an interest in keeping you.

When it fits. Several systems have to work together cleanly. The process is business critical, an outage costs you money immediately. You want to hand the topic over instead of managing it yourself.

When it does not fit. A single workflow. An agency is simply too expensive for that, and any good agency will tell you so.

What to watch for. Who owns the workflows, and where do they run? Correct answer: you own them, on your infrastructure or with full access. Anything else is a new dependency you can see coming. In twelve years of consulting I have seen companies pay millions to get away from a vendor.

The decision table

Look for your situation, not the nicest option.

The second-to-last row is the most important one and gets overlooked most often. Without capacity for maintenance, every solution is a ruin after a year.

The path I recommend most often

This does not have to be a pure decision. The most stable sequence I know mixes all three.

1. Build your first small workflow yourself. Not the most important one, an unimportant one. The goal is a feel for the effort, not the result.

2. Give the first serious process to a freelancer. With clear sign-off and documentation. Now you know what to check.

3. Bring in an agency when several systems have to work together. This is where the price starts to pay off, and where you can brief properly.

We basically did this with our own brands, just without the freelancer and spread over two years. Today we save 33 to 46 hours a week across all processes. The first step was small: at MUSTAX, reporting went from 2 days to 2 hours because we stopped pulling numbers together by hand. After that it was clear where the next lever was.

Five questions for any provider

Freelancer or agency, these questions cost you five minutes and sort candidates more reliably than any reference list.

• Who owns the workflows, and where do they run?
• What does running this cost per month, everything included?
• Who maintains the system, and what does that cost?
• What happens when a connected API changes?
• In which case would you advise me against this?

The last question is the best filter. Anyone without an answer to it will sell you anything, including the wrong thing.

Where we advise against it ourselves

A decision guide includes the cases where we turn enquiries down.

Too little volume. If a task costs you less than 2 hours a week, building something for it rarely pays off. At 5 orders a day, handling them manually is perfectly fine.

The process is still moving. A startup still finding its shape builds something that is obsolete next month, and pays twice.

A failure would threaten the business. Automation has an error rate, and ours is not zero either. Where a mistake makes you liable, a human belongs in the loop instead of more technology.

There is no maintenance capacity. Count on 2 to 4 hours a month per serious setup. If you cannot free those up, buying a finished tool is the better move.

Damian wrote the longer version of this list: When automation makes no sense. If you recognise yourself there, save yourself the intro call.

The downside of every option

So the table does not look too clean, here are the uncomfortable points.

Building it yourself costs you focus. Every hour on the workflow is an hour away from the product. That is the line item founders see least clearly.

Freelancers bring a knowledge risk. The build is cheap, the continuation is uncertain. Without documentation you are buying a black box.

Agencies bring a price risk. You pay for accountability rather than hours. If you do not need that accountability, you are paying too much. That applies to us as well.

There is no path without a downside. There is only the downside you choose on purpose.

Conclusion

What decides this is what is scarce for you right now: money, time or knowledge.

Money tight and time available: build it yourself. Time tight and the scope clear: freelancer. Several systems involved and the process critical: agency. Process still changing: none of the three, tidy up first.

And if the task costs you less than 2 hours a week, leave it alone. That is the answer you will hear least often, and it is frequently the right one.

If you want to talk your situation through once, without a pitch: short message to Flowhouse. You get an assessment and a number. Sometimes the assessment is that you do not need us.