What Is Workflow Automation?

What does a workflow look like?

A workflow is a trigger followed by steps. Here is one for a small contractor, an invented example for illustration. When a quote is marked accepted in the CRM, the workflow:

1. Creates the job in the team calendar.
2. Creates a draft invoice in the accounting tool.
3. Emails the customer a confirmation with the start date.
4. Posts a message in the team chat.
5. Adds a row to the sheet used for weekly reporting.

Nobody retypes the job into four places, and nobody forgets the confirmation. That is workflow automation: the same steps, in the same order, started by an event and finished without a person.

How the same workflow is priced three ways

The platforms count different things, so one workflow produces three different bills. This is where reading a pricing page pays off. The counting rules below come from each vendor's own pricing page, checked in November 2026.

• Zapier counts a task each time a unit of work completes successfully, such as a successful step. Triggers, polling and built-in tools like Formatter and Paths do not count, and failed actions do not count.
• Make counts credits. Most actions use 1 credit, and reading, searching, creating, updating and transforming data all count. Router modules and error handlers do not.
• n8n counts one execution for each run of the entire workflow, however many steps it has.

Take the contractor's workflow and assume 300 accepted quotes a month, each with an instant trigger and five actions. This is illustrative arithmetic.

Now scale to 3,000 quotes a month, ten times as many. Zapier would need 15,000 tasks, which is above the entry tiers we saw, and the page lists higher tiers. Make would need 18,000 credits, which is above the 10,000-credit tier. n8n would count 3,000 executions, above Starter's 2,500, so you would move to Pro, with 10,000 executions at EUR 50 a month billed annually.

No tool is cheaper in general, because the shape of the workflow decides. A workflow with many steps and high volume is cheap on a per-run model, and a workflow with few steps and low volume is cheap on a per-task model. Work out your steps times your runs before you choose. Our n8n, Zapier and Make comparison puts three scenarios side by side.

What happens when you hit your limit?

Limits behave differently, and the difference matters for a workflow that runs unattended. According to each vendor's pricing page:

A workflow that silently stops on the 28th of the month is a real way for follow-ups to vanish. Send yourself a weekly count of runs, and set an alert when you reach 80% of your allowance.

What kinds of workflow automation are there?

Most workflows fall into four types, and most small businesses need the first two.

• Point to point. One event creates one result, such as a form submission creating a CRM contact.
• Multi-step sequences. A chain like the contractor's example, with a few conditions.
• Approval workflows. A person must approve at a step, such as a manager signing off a quote over a set value. These keep humans in the loop on purpose.
• Orchestration across systems. Many steps, branches and retries across several tools. This is where cost and complexity grow, and where a specialist helps.

Rule-based steps and AI steps mix freely. A common pattern is a plain workflow with one AI step in the middle that reads a message and labels it. Our guide to what AI automation is explains where that fits, and business process automation covers the wider discipline.

Which tool is an example?

Zapier, Make and n8n are the usual starting points for small businesses. Microsoft Power Automate is the usual choice where a firm already runs on Microsoft 365. Its pricing page lists a Premium plan at USD 15 per user a month paid yearly, which covers cloud flows and attended desktop automation, and Process and Hosted Process plans at USD 150 and USD 215 per bot a month for unattended desktop automation.

How do you design a workflow before building it?

Draw it on paper first. Five questions cover it:

1. What is the trigger? One event, clearly defined.
2. What data is needed, and where does it live? If it is in someone's head, the workflow cannot use it.
3. What are the rules? Write every branch, including the odd cases.
4. What happens when a step fails? Retry, alert a person or stop.
5. How will you know it is working? A log, a weekly count, a sample check.

If you cannot answer one of these, the process is not ready to automate. Fix the process first. If a workflow touches customer data, you also need a data processing agreement with the platform, as GDPR Article 28 requires of processors. Our practical guide to AI automation for business shows how to choose which job to start with, and our AI automation service can help you design one.