I build the workflows that remove repetitive manual work from a business, using n8n, Zapier and Make — and write application code when those tools run out.
These three tools are not interchangeable, and picking the wrong one is the most common reason an automation project becomes expensive to maintain.
Zapier is the fastest to build in and the easiest to hand over. It suits linear workflows across popular apps where the value is in shipping this week. It becomes costly at volume, because pricing follows task count.
Make handles branching and iteration far better. When a workflow needs routers, error handling, or has to process a list of items and behave differently per item, Make expresses that clearly where Zapier would need several chained Zaps.
n8n is the one I reach for when a workflow needs real logic, custom code, or self-hosting. It runs on your own infrastructure, which matters when data cannot leave your environment or when per-task pricing would make the workflow uneconomical.
Most business automation does not need a model, and adding one where deterministic logic would do makes a workflow slower, more expensive, and harder to debug. A model earns its place when the input is unstructured — free text, a document, a transcript, an email — and the output needs judgement rather than a lookup.
That is where I use the Anthropic Claude API: classifying inbound messages, extracting structured fields from documents, summarising call transcripts into CRM notes, drafting replies a human then approves. The workflow around it stays deterministic, so when something breaks it is clear whether the failure was the logic or the model.
Some processes should not be automated in a no-code tool. If a workflow needs state across many steps, complex retries, or has to serve a user interface, stitching it together in Zapier produces something nobody can maintain. At that point the honest answer is to build the application — Next.js on Supabase and Vercel — and keep the automation platform for what it is genuinely good at.
Knowing where that line sits is most of the value. Recommending a build when a Zap would do wastes money; recommending a Zap when it needs a build wastes more.
Zapier for speed and simple linear flows across common apps. Make when the workflow branches or iterates. n8n when you need custom code, self-hosting, or when per-task pricing makes the others uneconomical at volume.
Yes. All three platforms can call the Anthropic Claude API and similar services. The useful pattern is a deterministic workflow with a model handling the one step that needs judgement, not a model orchestrating the whole process.
A server you control, a database, and someone responsible for updates and backups. It removes per-task pricing and keeps data in your environment, at the cost of owning the infrastructure.
Built by Eliser Eguia Tual, AI Automation & Operations Specialist, working with businesses in the US and Australia from Cagayan de Oro, Philippines (UTC+8).