n8n workflows built to run, and fixed when they don't.
New automations, the connections between tools you already pay for, and rescue work on n8n builds that quietly stopped firing. Scoped to the actual job, built by the person you talk to.
Automation rarely fails loudly. It just stops.
It's Monday. You open the sheet where new leads are supposed to land and the last row is from Thursday afternoon. Nothing crashed. Nothing sent an alert. A stored credential expired, one node started failing, and the workflow has been reporting to nobody for four days.
That is the ordinary way an automation dies, not the exotic way. n8n has real defenses against it and they are worth knowing about before you hire anyone. You can nominate an error workflow in Workflow Settings and start that workflow with an Error Trigger node, so a failed run pages you instead of disappearing. At the node level you can switch on Retry On Fail, which reruns a node until it succeeds, and set On Error to continue down a separate branch rather than halt everything. None of that is on by default. Somebody has to decide to turn it on, then test that it actually fires.
The second call is the opposite problem. Nothing is broken because nothing ever went live. Six nodes wired together on a Sunday, working perfectly against the Test URL, which n8n's docs say stays active for 120 seconds after you hit Listen for Test Event, and then never published against the Production URL. The build was fine. The last mile was not.
Both of those are fixable in an afternoon by somebody who has seen them before. What is worth paying for is the part after the fix: making the next failure announce itself.
What a workflow actually looks like.
Pick a starting point, then walk the steps. Select any step to read what it does in plain language, run the whole thing to watch it move, or break a step to see what happens when one fails. Simplified, and illustrative rather than any one business’s system.
The shape most businesses start with: answer everything, sort it, and put it where somebody will see it.
- New callerAlready a customer
- When that step fails
Where speed decides the outcome. The difference between a two-minute reply and a next-morning reply is usually the whole job.
- EmergencyRoutine
- When that step fails
The one almost nobody runs, and the one that quietly compounds. Reviews are the cheapest marketing a local business has.
- They left oneThey did not
- When that step fails
Illustrative. A real build is scoped to the systems you already run, and usually has fewer steps than people expect.
Three kinds of n8n work.
New workflows
- Intake and routing: what happens between a form submission and a booked job
- Follow-up sequences that fire off real events, not off somebody remembering
- Scheduled work: the weekly report that writes itself, the reminder that sends itself
- Error workflow and retry behavior configured during the build, not after the first outage
Systems connected
- n8n lists 1,900+ integrations, and the HTTP Request node reaches almost anything that isn't on the list
- Phone, CRM, calendar, inbox, spreadsheets, billing, connected in the order that touches revenue first
- Credentials created inside your accounts, so access does not run through somebody else's login
- AI steps only where they earn the slot: classification, extraction, drafting. Not on every node
Rescue and repair
- The execution log gets read before anything gets changed
- The failing node isolated and reproduced with the real input data
- Auth, pagination, rate limits, malformed input: the usual four, checked in order
- The fix, then the alerting that would have caught it, so the silence does not repeat
What a rescue pass actually covers.
Diagnosis before repair, in this order
- The execution history. Retention depends on the plan. n8n's published cloud limits put the Starter plan at 7 days of execution log retention, so a failure older than that leaves no log to read and has to be reproduced instead.
- Credentials. Expired OAuth tokens and rotated API keys are the most boring cause of a dead workflow and get ruled out before anything clever gets attempted.
- The contract with the other system. APIs change shape. A field that used to be a string arrives as an object, and every run since has been failing on the node downstream.
- Volume. Something built for twenty records a day falls over at two thousand. Usually that is pagination that was never implemented, or a rate limit that only appears under load.
- The workflow's own logic. Branches nobody ever exercised, loops that assume the list is never empty, a field quietly dropped three nodes upstream of where the error surfaces.
- Then the alerting. An error workflow, so the next failure tells you about itself the same hour instead of the following Monday.
Some of these are a two-hour fix and some mean the workflow should be rebuilt. Which one it is becomes clear after the first step, not before, and you hear the answer either way. There is no version of this where a small problem gets quoted as a large one.
A related piece of honesty: if your automation currently runs on Make or Zapier and it works, there is no reason to move it. Migration is worth doing when you can name the wall you are hitting. Wanting a more capable tool is not a wall.
Put a number on the leak first.
If the broken workflow is the one that chases inbound leads, the cost is not abstract. Run your own volume through the missed-call cost calculator, then decide whether this is worth a call.
Book a 30-minute callBuilt for you, owned by you.
Where it runs. Your account, not a shared one. n8n is source-available under its Sustainable Use License, which permits self-hosting and internal business use, so the build can live on n8n Cloud or on infrastructure you control. That is a scoping decision, made with you, not a default.
What you own. The workflows, the credentials, and the documentation. Workflows export as files you can take anywhere. Acuity does not bill you per run, per execution, or per workflow, and does not take a cut of what the platform costs. n8n's own plan is a different matter: executions there are metered and capped by tier, and a busy month can push you into a higher one. That is billed to you by n8n directly, which is why real volume gets estimated during scoping rather than discovered on an invoice.
What it costs. Builds start at $1,500 one-time and are quoted after scoping. A care plan is optional and starts at $75 a month, covering monitoring, tuning, and small changes. A more active management tier costs more, in the range of $300 a month, and is a choice rather than the default. Neither is required for anything to keep working. The full breakdown lives on the pricing page.
Platform costs are separate. You pay n8n directly, at their published rates, and Acuity neither marks that up nor resells it. Their published Starter plan is 20 euros a month billed annually and includes 2,500 workflow executions, where an execution is one run of a whole workflow regardless of how many nodes it has. That allowance is a real cap, and what you pay n8n scales with how much you run, so those numbers get walked through while the work is being scoped.
Who does the work. One senior operator, start to finish. The person on the discovery call is the person in the workflow editor. Nothing gets handed to a bench.
Best fit: you already run n8n, or you are about toIf the thing you actually need is the phone answered rather than the back office wired together, start with AI voice agents instead. If you want the sequence of a build laid out step by step, that is on the process page, and the wider service tiers are on the services page. There is also a longer read on AI automation for service businesses if you would rather do homework before talking to anyone.
What people ask before they hand over a workflow.
What is n8n, in plain terms?
n8n is a workflow automation platform. You build automations by connecting nodes on a canvas instead of building an application from scratch. n8n's own integrations directory lists 1,900+ integrations, and a generic HTTP Request node covers services that do not have a dedicated one. It is source-available under n8n's Sustainable Use License, which the company is careful to say is not the same as open source, and it can run on n8n's cloud or on infrastructure you control.
Can you fix an n8n workflow somebody else built?
Yes. That is one of the two reasons people land on this page. The first pass is diagnostic: read the execution history, isolate the node that is actually failing, reproduce the failure with the real input rather than guessing at it. You get told what broke and what it takes to fix before any build work starts. If it turns out to be a twenty-minute fix, you get told that too.
Do I need my own n8n account, or do you host it for me?
You keep your own account. Whether that is n8n Cloud or a self-hosted instance on infrastructure you control is a scoping decision, and n8n's license permits self-hosting for internal business use. The workflows, the credentials, and the exported files belong to you either way. Nothing is parked inside a vendor account you cannot reach.
What does an n8n build cost?
Builds start at $1,500 as a one-time cost, scoped to what the work actually is rather than to a package. After that, a care plan is optional and starts at $75 a month for monitoring, tuning, and small changes. A more active management tier costs more, in the range of $300 a month, but it is a choice rather than the starting point, and nothing stops running if you decline both. n8n's own platform cost is separate: you pay n8n directly at their published rates, Acuity does not mark it up, and Acuity does not bill you per run or per execution. Their published Starter plan is 20 euros a month billed annually, which includes 2,500 workflow executions, and executions are metered and capped by plan tier.
Should I move off Zapier or Make?
Only if you can name the wall you are hitting. Per-task cost at volume, branching logic a linear builder cannot express, or a requirement to keep data on your own infrastructure are real reasons to move. Wanting a more capable tool is not a reason on its own. If what you have works and the bill is fine, keep it and spend the money somewhere it changes an outcome.
How do I know it will not break the same way again?
You do not get a guarantee, and anyone offering one on an integration that depends on somebody else's API is selling you something. What you get is the boring version: an error workflow that fires when a run fails, retry behavior switched on for the nodes that talk to flaky services, and a written description of what each workflow does so the next person can read it. Failures still happen. They stop being silent.
More general questions are answered on the FAQ page.
Bring the workflow. We'll read it.
Thirty minutes. If it's broken, you'll leave knowing what broke. If it isn't built yet, you'll leave knowing what it would take and what it would cost.
Book a 30-minute call