n8n consulting
n8n consultant: what to ask before you hire one
The results for this search are mostly staffing marketplaces renting you a seat by the month. A seat is not the thing that fails. What fails is an estate nobody is watching, so hire against that instead.
Search this term and you get rate cards. Hourly bands, monthly retainers for a dedicated developer, forty-eight hour onboarding promises. What none of the listings do is tell you which of the three jobs above they are covering, which is the only question that changes the outcome.
The seven questions that sort operators from suppliers
Ask these before the rate conversation. The answers are diagnostic on their own.
- Show me an error workflow you have shipped. Anyone who has run n8n in production has one and can describe what it routes to. Anyone who has only built demos will describe the concept.
- How do you find out a workflow stopped? The correct answer names a channel and a person. "The execution log" means nobody finds out.
- What is your pruning policy on execution history? Every execution writes to the database repeatedly. A team that has not thought about this has not run an instance long enough for it to matter.
- Which of my integrations has no existing node? A consultant who answers this before quoting has read your list. One who quotes first has not.
- Who owns the credentials after handover? If the OAuth grants sit in an agency account, you have bought a dependency rather than a system.
- What would you tell me not to automate? Anything worth hiring says no to something. A yes to everything is a sales process.
- What happens to this if you disappear? The answer should involve documentation you can read and an instance you control.
What the four engagement shapes actually cover
| Shape | What you get | What it does not cover | Fits when |
|---|---|---|---|
| Audit | A read of the current estate and a prioritised list of what to fix | Any of the fixing | You have workflows and no confidence in them |
| Fixed-scope build | A named set of workflows, built, tested and handed over with documentation | What happens on week nine | The scope is genuinely knowable up front |
| Rented seat | A developer's time by the month | Judgement about what to build, and accountability when it breaks | You already have someone senior directing the work |
| Operated delivery | Build plus monitoring, error handling and incident response | Nothing, which is why it costs the most | The automation touches money or customers |
The mistake we see most often is buying a rented seat when the real gap was judgement. You end up with more workflows and the same problem, plus a monthly invoice.
The four things that break on every self-hosted estate
These are not exotic. They are on every instance we have inherited, and a consultant worth hiring will name them before you do.
- Silent failure. A workflow with no error workflow attached fails into a log nobody reads. The estate goes quiet for weeks before somebody spots it. We have had this happen on a weekly reporting workflow we run for a US email-marketing agency client, and the fix was moving it somewhere that shouts.
- The database, not the CPU. Every execution writes to Postgres repeatedly, so adding workers past a point makes the instance slower rather than faster and exhausts the connection pool. The full mechanism is in why self-hosted n8n slows down as you add workers.
- Credentials in the wrong account. OAuth grants held by whoever built it, which turns a handover into a re-authentication project across every service.
- Webhook URLs with no inventory. Every inbound webhook has an address something else calls. Nobody has the list until the day it matters.
What we charge against, and why it is not a seat
Digiton is an AI consulting firm in Lisbon. We price against the operation being handed over rather than against hours, because hours reward the slow version of the work. The practical effect is that the scoping conversation is about your workflow inventory and your integration list, not about a rate.
Two things we will say early. If your automation spend is small and nobody wants to own a server, we will tell you to stay on the hosted platform. And if the estate is one expensive workflow rather than forty cheap ones, the answer is to rewrite that workflow, not to migrate. The automation cost calculator gives you a build band and a payback range before you talk to anyone, ours included.
Proof you can check
The work we can point at publicly is on this site rather than in a case-study deck. A WhatsApp assistant for a Portuguese municipal water utility that had to reach a billing system built in 1996, written up with what we would flag to anyone attempting it. The hour-by-hour breakdown of a Zapier to n8n migration by workflow type. And an operator post-mortem on n8n in production. We build and operate automation for clients across 8 countries, from Lisbon.
If you are still choosing the tool rather than the partner, start with the self-hosted Zapier alternatives compared on running cost.
Frequently asked questions
What does an n8n consultant do?
Three separable jobs. Design, meaning deciding what to automate and what to leave alone. Build, meaning workflows, credentials and error paths. Operation, meaning somebody watches executions and responds when they stop. Staffing marketplaces sell the build by the month, which is the cheapest of the three to replace and rarely the one that is hurting.
How much does an n8n consultant cost?
It depends on which of the three jobs you are buying, which is why published hourly bands tell you little. An audit prices against the size of the estate, a fixed-scope build prices against the workflow inventory and the integration list, and operated delivery costs the most because it carries the incident response. Digiton prices against the operation being handed over rather than against hours.
Should I hire a freelancer, an agency or a marketplace?
Hire against the gap. If you have someone senior directing the work, a rented seat from a marketplace is fine. If the gap is judgement about what to automate, a seat makes it worse by producing more workflows and the same problem. If the automation touches money or customers, buy operated delivery, because the failure mode is silence rather than downtime.
What should I ask an n8n consultant before hiring?
Ask them to show an error workflow they have shipped, describe how they find out a workflow stopped, state their pruning policy on execution history, name which of your integrations has no existing node, say who owns the credentials after handover, name something they would tell you not to automate, and explain what happens to the system if they disappear.
What is an n8n implementation partner?
Usually a firm that builds and hands over, sometimes with a listing in the n8n partner directory. The label matters less than the scope. Confirm whether monitoring, error handling and incident response are inside the engagement or outside it, because that boundary is where most disappointment lives.
Can you take over an n8n instance somebody else built?
Yes, and the first pass is always the same four items: whether error workflows exist and where they route, whether execution history is being pruned, whose account holds the OAuth credentials, and whether anyone has an inventory of the inbound webhook URLs other systems are calling.
Do we need n8n at all?
Not always. A visual canvas earns its place when a non-engineer has to read the logic, and stops earning it when the logic is a script wearing thirty nodes. Digiton moved a large share of its own recurring jobs to plain scheduled scripts for that reason, so the question we ask first is who needs to read the workflow.
Related
Want the number before the call?
The cost calculator gives you a build band, a monthly running cost and a payback range from our published 2026 price list. No email needed to see it.
Open the AI automation cost calculatorTell us the operation, get a scoped answer
One line about the work you would hand over. We reply within one working day with the band it falls in and whether it is worth building at your volume.