Business Process Automation: Where to Start

TL;DR: Process automation doesn't start with picking a tool. It starts with identifying the right process — the one that repeats most often, has a measurable output, and doesn't require human judgment at every step. Realistic starting cost: under $40/month.

Quick Answer — for people who need it now

Start automating with 3 steps: (1) pick one process that repeats more than 5 times a week and has a clear output, (2) build the simplest workflow that can actually run — no code, no IT required, (3) measure the results after 30 days before expanding. Make.com's free plan + Claude Pro (~$22/month) is enough for your first 3–5 basic workflows.

Most founders hear "process automation" and picture a complex IT project — a dev team, a big budget, 3–6 months to deploy.

That's not how it has to go.

I automated our internal weekly reporting process — from 2 hours of manual work every Friday down to a 15-minute review — in a single afternoon, without writing a single line of code.

This post is how I did it, and how you can apply the same approach to your own business.

What process automation is — and isn't

Before diving in, one common misunderstanding needs clearing up.

Process automation means redesigning a chain of work so a system handles the repetitive steps instead of a person — while people stay involved only where real judgment is required.

It is not: replacing people. It is not a bot that answers everything automatically. It is not "install some software and you're done."

The key distinction: good automation doesn't remove people — it puts them exactly where they belong, at the decisions that matter, instead of wasting them on repetitive steps that require no thinking at all.

Which process to automate first

Not the most important one. Not the biggest one. The one that satisfies all 3 conditions:

Condition 1 — Repetition: Happens at least 3–5 times a week, following the same pattern.

Condition 2 — Clear output: You can describe the correct output in two sentences. If you can't describe it, AI can't do it.

Condition 3 — No complex judgment at each step: The logic is clear, not "it depends on the situation."

A 3-step framework to get started — no IT required

Step 1: Map the process in 30 minutes

Draw out the current process, step by step:

[Trigger] → [Step 1] → [Step 2] → ... → [Output]

For each step, note: who does it, how long it takes, whether it's error-prone, and what the output is.

Once you've mapped it out, highlight the steps that repeat, require no creativity, and produce a measurable output. That's the automatable zone.

Real example from our own operations: our weekly reporting process used to have 7 steps — after mapping, only 2 required a human (reviewing unusual numbers + writing strategic commentary), the other 5 were fully automated.

Step 2: Pick the right tool — not the hottest one

Principle: Choose the minimum — just enough to run that one process. Add tools later, once the first process is stable.

Step 3: Build the simplest workflow that can actually run

The MVP principle: automate one step, not the entire process.

Example: instead of automating the entire post-sale customer care process, start with: automatically send a thank-you email + onboarding checklist right after a deal closes (triggered from a Google Sheet).

That single step saves 15–20 minutes per deal. At 10 deals a month, that's 2.5–3 hours a month. Not huge, but enough to build the habit of measuring results and learning how to operate an automated workflow.

3 common mistakes — and how to avoid them

Mistake 1: Automating before the manual process is stable

If the manual process still changes every week, automation will just produce errors faster. Rule of thumb: the manual process should run stably for at least 4 weeks before you automate it.

Mistake 2: No human-in-the-loop for important output

Any workflow that automates important output (external reports, emails to customers) needs at least one human review point before the next step triggers. Not because AI isn't good enough — but because you're accountable for that output, not the AI.

Mistake 3: Choosing the most complex tool because it's "the most feature-rich"

Self-hosted n8n can do almost anything. But if it takes 2 weeks to set up, your first process still isn't running, and nobody on the team knows how to maintain it — you've just spent resources building something no one uses.

For SMEs under 20 people: Make.com Free + Claude Pro already covers 80% of real automation needs.

A real internal case: a weekly AI report in 15 minutes

Before: Every Friday, 2 hours spent pulling numbers from 4 channels, compiling them into a Google Doc, writing commentary, and sending it to the team.

After automating (built in one afternoon):

• Make.com pulls data from Google Sheets every Friday morning
• Claude takes the raw data and drafts a report using a fixed template
• I receive the draft on Slack at 8am
• 15-minute review, edit the strategic commentary, send

Result: 2 hours → 15 minutes. What I actually do is the part that genuinely needs me.

Build cost: $0 (Make.com Free + Claude Pro I already had). Build time: one afternoon (~3 hours).

Checklist before you start

• I've identified one specific process (not "automation in general")
• That process repeats at least 3–5 times a week
• The output can be described in two sentences
• The manual process has been stable for at least 4 weeks
• I know who will review the output before it's used for real
• I have a plan to measure results after 30 days

Go deeper — based on where you are

If you want to understand more complex AI workflow architecture:

• AI Workflow for Vietnamese Businesses — A Real-World Architecture
• What Is Agentic Workflow — and Why It's Completely Different From Traditional Automation

If you want the full picture of applying AI to your business:

• AI for Vietnamese SMEs — Where to Start, What to Do, How Much It Costs

You already know which process you want to automate — but you're not sure where to start?

That's exactly what the Audit session is for. I listen to your process for 30 minutes, point out which step to automate first, and which tool actually fits your scale and real budget.

Not generic advice — a specific analysis for your specific process.

Book a Process Audit →

FAQ

Do I need a developer to automate a process?

No — not if you start with a simple workflow using a no-code tool like Make.com or Notion automation. You need a developer once the workflow gets more complex, connects to internal systems, or needs high-throughput processing. For an SME just starting out: you don't.

How long until I see results?

If your first workflow is designed correctly, results are visible within the first week. Judging quality and reliability takes about 30 days of real operation.

Is company data safe when using Claude/Make.com?

Claude offers an option to opt out of training on conversation data. Make.com processes data through its own servers — for sensitive data, review their terms or consider an on-premise solution. This is a question to audit before deployment, not after.

Should I start with a big process or a small one?

Small — always start small. Bigger processes are more complex, error-prone, and harder to debug. Start with the smallest process that still has measurable value — build the habit and operating experience before you scale.