By Rakibul Hasan, Moonlab
Business process automation works in five steps: map how the work happens today, pick the steps worth automating, build a small version, test it on real cases, and measure the result. Most projects that fail skipped the first or the fourth step.
This guide walks through each step in plain terms, with a hypothetical example of a small removals company handling quote requests. You do not need special software to follow it. A pen and a spreadsheet are enough for the first three steps.
Step 1: map the process as it really runs
Write down what actually happens from the first trigger to the final result, not what should happen. Ask the person who does the work to walk you through it. Note each step, who does it, which tool they use, how long it takes and what goes wrong.
For the removals company: a quote request arrives by email or form, someone reads it, checks the date in a diary, works out a price from a rate card, types a reply, sends it, and follows up if there is no answer. Seven steps, three tools, and some steps depend on one person's memory.
Step 2: pick what to automate
Not every step should be automated. Score each one on a few questions.
- How often does it happen? More often means more time saved.
- How similar is each case? Repetitive steps are easier to automate.
- Are the rules clear? You should be able to explain them to a new starter.
- What happens if it goes wrong? Low-risk, reversible steps first.
- Does it touch customers or money? Keep a person in the loop there.
Step 3: design the future process on paper
Redraw the process with the automation in it. Mark which steps are automatic, which need approval and what happens to exceptions. Often you find a step you can remove altogether, which is better than automating it.
In the example: the form captures all details upfront, the system checks the diary and applies the rate card to draft a price, the draft is sent to the owner for a one-click approval, and a follow-up goes out after three days if there is no reply. Anything with unusual items goes to the owner directly.
Step 4: build a small version
Build the simplest working version in the tools you already have, using a platform like Zapier, Make or n8n (compared in moonlab.space/guides/zapier-vs-make-vs-n8n). Use your own accounts, name the workflow clearly and set up an error alert from day one.
Resist adding extras. A first version that handles the common case and passes the rest to a person is better than a complex one that tries everything.
Step 5: test on real cases
Run the workflow on at least twenty real past examples, including strange ones. Check every output. Then run it in parallel with the manual process for a week or two, so a person still does the work and you compare the results.
- Missing information in the input.
- Duplicates, such as the same form sent twice.
- Unusual values, such as a very large order.
- Another language or odd formatting.
- A tool being down or a login expired.
Step 6: measure and improve
Decide your measures before launch: time per task, response time, errors caught, number needing a person. Compare with your notes from step 1. If you saved three hours a week and it costs a small monthly fee, you have a clear answer. Our payback method is in moonlab.space/guides/business-automation-cost.
Review the logs every week for the first month, then monthly. Fix the rules, update documents and note changes. Automations drift as your business changes, so assign an owner.
A worked example in numbers, and when not to automate
Continue the removals example. Say the firm receives about 25 quote requests a week and each takes roughly 12 minutes to read, check the diary, work out a price and reply. That is 25 x 12 = 300 minutes, or 5 hours a week. After mapping, they decide to automate three steps: capturing details in a structured form, checking the diary and drafting the reply. The owner still approves each price with a one minute check.
If the new process takes about 3 minutes per request, that is 25 x 3 = 75 minutes, so around 3.75 hours a week is saved. If the owner's time is worth a certain amount per hour, multiply and compare with the build and running costs. Use your own numbers, and be honest about the time spent on checking and fixing exceptions, because that is where hopeful estimates fall apart.
Some processes should stay manual for now. Rare tasks, processes that change every week, work that depends on personal judgment or relationships, and anything where the rules are not agreed yet are poor candidates.
- Process not agreed: settle the rules first, then automate.
- Fewer than a handful of runs a month: the build rarely pays back.
- High stakes and no way to undo: keep a person doing it, with software helping.
- Data is poor or scattered: clean it up first, or the automation spreads the mess.
- Nobody will own it afterwards: it will break and be ignored.
Common mistakes
Automating a broken process just makes the mistakes faster, so fix the process first. Skipping exception handling leaves cases stuck without anyone knowing. Building in one person's personal account means the workflow disappears when they leave. And launching without a person checking early output is the quickest route to a customer-facing error.
If you want help mapping and picking, the free automation audit at moonlab.space/contact covers exactly these steps, and moonlab.space/services/operations describes our process work.
