n8n, Make or custom code? Choosing how to build an automation

01The short answer

Use the simplest tool your team can own. For most single workflows that is a visual automation platform. Move to custom code when the workflow needs heavy logic, high volume, strict security or an interface your customers use.

02Visual platforms: n8n, Make and Zapier

• Zapier: the easiest to start and the largest app catalogue. Costs grow with every task run, so busy workflows get expensive.
• Make: more control over branching and data shaping, usually cheaper per operation than Zapier.
• n8n: can be self-hosted, so data stays in your infrastructure and running costs stay flat. Handles code steps and AI agents well. Needs someone to keep the server healthy.

03When custom code wins

Custom code makes sense when the automation is part of your product, processes high volumes, needs careful error handling, or must meet security requirements a shared platform cannot. It costs more to build and less to run at scale.

export async function POST(req: Request) {
  const lead = await req.json().catch(() => null);
  if (!lead?.email || !/^[^\s@]+@[^\s@]+\.[^\s@]+$/.test(lead.email)) {
    return Response.json({ ok: false, error: "Valid email required" }, { status: 422 });
  }
  // Hand the clean record to your workflow tool or CRM.
  const res = await fetch(process.env.CRM_WEBHOOK_URL!, {
    method: "POST",
    headers: { "Content-Type": "application/json" },
    body: JSON.stringify({ ...lead, receivedAt: new Date().toISOString() }),
  });
  return Response.json({ ok: res.ok }, { status: res.ok ? 200 : 502 });
}

04The hybrid most teams end up with

In practice the best setup is often both: custom code for the parts that need to be fast, safe and tested, such as the AI agent and the customer-facing chat, and a visual platform for the glue your team wants to change without a developer.

05A quick way to decide

1. Who will change this workflow next month? If it is not a developer, start visual.
2. How many runs per month? Price the platform at ten times today's volume.
3. Where must the data live? If it has to stay in your cloud, self-hosted n8n or custom code.
4. What happens when it fails? If a failure costs money or trust, invest in custom error handling.