How I work on n8n & Make
Most automation stacks I audit grew one urgent workflow at a time. Someone connected the form tool to the CRM, someone else added a Slack alert, an agency built a reporting scenario, and eventually nobody knows which workflows still matter or what happens when one fails. I treat automation as production software: every workflow has an owner, a documented purpose, an error path and a cost you can see.
Lead enrichment and routing
A new lead should land in the CRM complete, deduplicated and assigned within minutes. I build workflows that catch form and ad leads through webhooks, search for an existing record before creating anything, enrich company data from the sources you already pay for, and route by fit, segment or territory. Enrichment writes to dedicated fields or fills only empty ones, so it never overwrites what a rep typed by hand.
Reporting pipelines and alerts
Weekly reports assembled by copying numbers between tabs are slow and error-prone. I build pipelines that pull ad spend, sessions, leads, pipeline and revenue from their APIs into a sheet or warehouse on a schedule, then post a short summary where your team already works. Pacing and anomaly alerts flag overspend, broken tracking or a sudden drop in leads the same day, not at month end. I cover the approach in how to automate marketing reporting.
Content operations with LLM steps
LLM steps are good at research summaries, classification, first drafts and turning call notes into structured fields. They are not good at being left alone. I ask the model for structured output, validate it before anything downstream uses it, and put a human approval step in front of anything customer-facing or hard to undo. In n8n the run can wait for a reviewer’s response; in Make, the approval usually triggers a second scenario. Prompts, inputs and outputs get logged, so you can see what the model did and improve it.
Error handling, logging and cost control
This is where most automations fall short. In n8n, I set an error workflow that fires on failure, retry on fail for unreliable APIs, and error outputs on steps where one bad record should not stop the whole run. In Make, I choose error handler routes such as Resume, Ignore and Break on purpose and store incomplete executions, so failed runs can be retried instead of lost. Every failure posts to a monitored channel with the workflow, step and record.
On cost, I replace polling triggers with webhooks where the source app supports them, filter early so records you don’t need never reach expensive steps, and use smaller models for simple classification. Usage gets reviewed per workflow, so a runaway loop shows up before the invoice does.
Self-hosting n8n
Self-hosting the n8n community edition removes per-execution pricing, not the work. A production instance needs Postgres rather than the default SQLite, backups, a safely stored encryption key, restricted editor access, execution data pruning and a tested upgrade routine. At higher volume, queue mode with separate workers keeps long runs from blocking everything else. If nobody on your side wants to own a server, n8n Cloud is usually the better call.
When you need an n8n & Make specialist (and when you don’t)
You likely need one when:
- Leads go missing or get duplicated and nobody can say which workflow caused it
- Workflows were built under someone’s personal account, and that person has left
- Your Make credit usage or LLM bill keeps climbing without more output
- You want AI in your marketing workflows, but not unchecked AI writing to customers
- You self-host n8n and nobody is sure when it was last updated or backed up
You probably don’t need one when you run a handful of simple automations at low volume with no AI steps and they work. Whoever uses them can maintain them. And if the process you want to automate is not stable yet, fix the process first: automation makes a broken process run faster, not better. If you are still choosing a platform, start with my comparison of n8n, Zapier and Make.
How it fits the rest of your growth system
Automation is never the goal on its own. It sits inside a growth engagement, next to acquisition, CRM and measurement, because workflows are the plumbing those systems depend on. Lead routing decides how fast sales follows up on paid leads. Reporting pipelines decide which campaigns get budget. Enrichment decides whether lead scoring has anything useful to score. I design workflows around those decisions, not around which apps happen to have a connector.
n8n and Make work is part of my AI automation service, delivered as an add-on from $2,500 a month alongside Growth Foundation (from $3,500 a month) or Fractional Growth OS (from $6,000 a month).
What the first 30 days look like
| Week | Focus | What you get |
|---|---|---|
| 1 | Audit | Workflow inventory, owners, failure points, cost review and a prioritized fix list |
| 2 | Stabilize | Error handling, failure alerts, credential cleanup and fixes to the workflows that matter most |
| 3 | Build | First new workflow in production, such as lead routing or a reporting pipeline |
| 4 | Harden and hand over | Logging, cost limits, documentation and a backlog for the next month |
Week one is the Growth Audit: a fixed $1,500, credited if you continue. I work in your n8n instance or Make organization, use shared service accounts rather than personal logins, and document every workflow so your team can maintain it.
Get it built
I do not advise on automation. I build it. If your workflows fail quietly, cost more every month or still leave someone copying data between tabs, I’ll audit what you have, fix what matters and build what’s missing. Start with a Growth Audit and we’ll begin with the workflow that costs you the most time or pipeline.