What an automation really costs: build, run and maintain (with a calculator)

“Should we automate this?” is a money question, but it’s usually answered by enthusiasm. Here’s a simple way to answer it with numbers, including the costs that usually get left out.

The three costs

1. Build cost (one-off)

• Planning: mapping the process, edge cases and exceptions
• Building and connecting the steps
• Testing with real data, including failure cases
• Documentation and handover (see how to document automations)

People often budget only for the building. Planning and testing are where reliability comes from.

2. Running cost (monthly)

• Platform fees, which scale with volume, measured in Zapier tasks, Make credits or n8n executions. See n8n vs Make vs Zapier for how the same workflow is counted on each.
• Paid connectors or apps some integrations need
• Per-message fees, for example WhatsApp or SMS
• AI usage, if steps call an AI model (see LLM API costs explained)
• Hosting, if self-hosted (see self-hosting n8n: the real cost)

3. Maintenance cost (ongoing)

The cost most often forgotten:

• Fixing it when a connected app changes a field, an API version or a login method
• Renewing expired credentials
• Responding to failure alerts
• Small changes as the business process changes

A planning assumption we use: one to two hours a month per business-critical automation, more for complex or high-volume ones. Track your own actuals and replace it.

The value side

• Time saved: times per month × minutes each ÷ 60 × the hourly cost of the person who did it. Use their actual cost to the business, not only salary.
• Errors avoided: what the mistakes cost today, such as lost leads, wrong invoices, late payments, re-work. Estimate conservatively.
• Speed: replying to leads in minutes instead of hours, or collecting payment sooner. Real, but hard to quantify. Count it in your judgement, not the maths.

The calculation

Monthly value     = hours saved × hourly cost + error cost avoided
Monthly cost      = platform fees + maintenance hours × hourly rate
Monthly net       = monthly value − monthly cost
Payback (months)  = build cost ÷ monthly net
3-year TCO        = build cost + 36 × monthly cost

A hypothetical example: syncing store orders to accounting

• 400 orders a month, each taking about 3 minutes to enter by hand → 20 hours a month
• Staff cost assumed at AED 60 an hour → AED 1,200 a month of time
• Errors (wrong amounts, missed orders) assumed to cost AED 300 a month to find and fix
• Build: 12 hours at an assumed AED 200/hour → AED 2,400
• Running: platform fees AED 100 a month
• Maintenance: 1.5 hours a month at AED 200 → AED 300

This one is an easy yes. Change the inputs to a task done 20 times a month, and the same build cost takes years to pay back.

Rules of thumb

• Payback under 6 months: usually worth doing, if the process is stable.
• 6–12 months: worth it if it also removes risk (lost leads, compliance) or will grow with volume.
• Over 12 months: simplify (use a native integration, cut edge cases) or leave it manual.
• Volume growth matters both ways. It increases value, but also platform fees on per-task pricing.

Ways to lower the cost

• Use native integrations first. They’re often free and maintained by the vendor.
• Cut rare edge cases. Let the automation handle the 90% and send exceptions to a person.
• Choose the platform for your volume. Per-step pricing gets expensive for long workflows at high volume.
• Build in failure alerts from day one. Silent failures are the most expensive kind of maintenance.

Frequently asked questions

What hourly cost should we use for staff time?

Their full cost to the business divided by productive hours, which is higher than salary alone. If you’re unsure, use a conservative number. A good automation still pays back.

Should we count the cost of our own time to build it?

Yes, even if no invoice is involved. Hours spent building automations are hours not spent selling or serving customers.

What about automations that save customers’ time rather than ours?

Faster responses and fewer errors improve conversion and retention. Hard to put in the formula, but worth noting next to the numbers when you decide.

If you’d like help sizing up and building your first automations, that’s part of our tech stack service.