Three tools run most of the automation in small and mid-sized companies. They all connect the same apps. They differ in who maintains them, what they cost at volume, and what happens when a run fails.

Zapier

The widest library of connectors and the fastest to build in. Each run of each step counts as a task, so a workflow that handles 2,000 orders a month with six steps is 12,000 tasks. Right for a team of under ten with a few simple workflows that someone in the business will own.

Make

Visual scenarios with real branching, iterators and error handling, at a lower price per operation. The best choice for most companies between 10 and 100 people. A scenario that does something complex is readable by the next person, which matters more than any feature.

n8n

Open source and self-hostable, with code steps when a connector does not do what you need. Right when you process a lot of data, when data must stay on your own servers, or when the workflows are part of the product rather than the back office.

Our rule of thumb

• Under 5,000 runs a month and no developer: Zapier
• Branching logic, several hundred thousand operations: Make
• Data residency, heavy volume, custom code: n8n

The tool is the smallest decision. The map of what runs, in what order, with what checks, is the work.

What about AI agents?

They sit inside any of the three as a step that reads, classifies or drafts. We give an agent a narrow job with a clear output, then let the workflow decide what happens next.