Nobody needs a developer to build an n8n workflow. That is the entire point of n8n, and anyone quoting you by the workflow is charging you for the easy part.
Hire an n8n developer when your workflows write, spend money, or contact a human, not when you have a lot of them. Building in n8n is deliberately easy, so the node count and the workflow count tell you almost nothing about what the work costs. The expensive part is what happens on the run that fails at 2am, on the retry that fires the same action twice, and on the execution that finished green having done the wrong thing. Count those, and the hiring decision makes itself.
I build and run these systems for a living, 500+ projects and 1,500+ clients, and I have never once been called in because somebody could not drag the nodes together.
A workflow that only reads is free to get wrong. A workflow that writes, spends or contacts a human is cheap right up until the first retry.
The mistake: scoping the job by workflow count
The brief nearly always arrives as a number. We need about six automations, what would you charge? It is a reasonable question and it is the one number that does not predict the work. Six workflows that pull data into a dashboard are a weekend. One workflow that issues refunds is a quarter of careful work, and it stays careful work forever rather than once.
The other half of the mistake is pricing the project against the subscription. n8n Starter is 20 euros a month billed annually. That figure is real, it is on n8n's own pricing page today, and it is quietly misleading, because it prices the editor. It does not price the morning you need to know what a workflow did last Tuesday and find out you cannot.
"We need about six automations. What do you charge per workflow?"
The count is the one number that does not predict the work. Five read-only reports and one workflow that touches billing are not the same job, and only one of them can lose you a customer.
"One of these writes to our billing system. What happens on a retry, and who finds out when it fails?"
That is a scoping question and a vetting question at once. The answer prices the build and tells you whether the person answering has ever run one in production.
Blast radius is the unit, not the node count
Sort every workflow you own into one of four classes by what it can do to the world outside n8n.
A workflow that only reads can be wrong all day and the worst outcome is a stale number on a dashboard. A workflow that writes can create a duplicate record that somebody has to clean up. A workflow that spends can charge a card twice. A workflow that contacts a human can send the same invoice reminder four times to your largest account, and that is the one you hear about by phone.
These classes do not fail the same way either. Reads fail loudly. An API returns a 500, the execution goes red, you see it in the list. The dangerous class is the run that completes successfully having done the wrong thing. A field arrived empty, the mapping passed the empty value along without complaint, the update wrote, and the execution is green. Nothing alerts, because nothing errored.
That is what you are actually hiring against. Not whether somebody can build the workflow, but whether they write it so a wrong-but-successful run is either impossible or immediately visible. In practice that means asserting on what the workflow changed rather than on whether it finished, putting an idempotency key on anything that writes, and setting a hard cap on anything that spends or contacts a person.
What your plan retains decides how much you can debug
Here is the part almost nobody checks before they commit, read from n8n's own pricing page today. The plan you are on does not just cap how many times your workflows run. It caps how much of their past you can see.
Starter, at 20 euros a month billed annually, gives you 2,500 executions, 5 concurrent executions, one day of workflow history and 7 days of execution log retention. Pro, at 50 euros a month billed annually, gives you 10,000 executions, 20 concurrent, five days of workflow history and 30 days of log retention. Business, at 667 euros a month billed annually, adds separate environments, scaling options and version control using Git. Enterprise goes to 200+ concurrent executions and unlimited log retention.
Read that Business line again, because it is the one that changes how you plan. Version control using Git is a Business plan feature. Below it, your automations have one day of history on Starter and five on Pro. Revert this workflow to what it was before Tuesday is not a support ticket on those plans. It is a rebuild from memory.
Concurrency deserves the same attention. Five concurrent executions on Starter is not a monthly budget, it is a width. A webhook that arrives in bursts does not care that you have 2,500 executions left in the month; it cares how many can run at the same moment, which is a different failure and one that shows up as delay rather than as an error. If you are already hitting that ceiling, the fix is architectural rather than commercial, and queue mode and response offloading are where that conversation starts.
One more piece of context that cuts both ways: n8n counts a single run of an entire workflow as one execution no matter how many steps it contains. That is genuinely generous compared with per-step billing, and it is exactly why teams underestimate what they are building. Cheap to run and cheap to build does not make it cheap to reconstruct what happened.
Four moves before you hire anyone
The bottom line
n8n removed the reason to hire someone to build a workflow. It did not remove the reason to hire someone to run one. The skill you are paying for is not node placement, it is the discipline of assuming every run will eventually go wrong in a way that does not announce itself, and building so that when it does, the damage is bounded and somebody finds out.
If you want a second pair of eyes on what you have already built, I will do it on a call. Bring the workflow list and I will tell you which ones I would not let run unattended, what I would add before I did, and whether you need a dedicated n8n developer or two afternoons of cleanup. Our published rate for a dedicated developer is from $2,497 a month, and plenty of the audits I run end with me telling somebody they do not need one yet. See the rest of what we build in AI automation and back office automation, or who I am.
Book a 30 minute call: cal.com/zeeshanwaheed/30min or email [email protected].