Where to Start with Business Process Automation (Hint: Not with AI)

Everyone wants to talk about AI. Meanwhile, most businesses have dozens of processes that could be automated with tools they already own — or simple ones that cost almost nothing. The highest-return automation opportunities are usually the boring ones nobody is excited about.

I built the first automation software robot at an international school in Luanda where I ran the technology department. It wasn’t glamorous — it took a manual administrative workflow and made it run itself. But the hours it saved added up fast, and it taught the organization that automation was practical, not futuristic.

Here’s how to find your own starting point.

Look for high-volume, rule-based tasks

The best automation candidates share a pattern:

• Done frequently — daily or weekly, not occasionally.
• Follows clear rules — if X, then Y. No judgment calls.
• Moves data between systems or people — copying, formatting, routing, notifying.

Walk through your team’s week and ask: what do you do that’s repetitive, predictable, and requires a computer but not a brain? Data entry between systems. Report generation. Approval notifications. Invoice matching. Status updates that copy information from one place to another.

These are the tasks where automation pays for itself in weeks, not quarters.

Approval workflows are the easy first win

Paper or email approvals are slow, untrackable, and full of “I thought you handled that.” Almost every business has approvals that could run through a simple automated workflow:

• Purchase requests above a threshold route automatically to the right approver.
• Leave requests go to the manager, then to HR, with automatic confirmation.
• Expense reports follow a defined path with status tracking.

The technology for this is basic — most office suites include workflow tools. The hard part is deciding who approves what. Once that’s clear, the workflow builds itself.

Automate the reports people actually read

If someone on your team spends hours every week pulling numbers from different places, formatting them, and emailing a summary — automate it first. Scheduled reports that pull from your systems and deliver formatted summaries are one of the highest-ROI automations available.

Start with the weekly numbers your leadership actually uses. If a report takes four hours to produce and gets read for four minutes, that’s a process that should run itself overnight.

Connect systems before building new ones

A lot of manual work exists because System A doesn’t talk to System B, so a person becomes the integration layer — copying, pasting, re-entering. Before buying anything new, check whether your existing systems can exchange data:

• Native integrations built into the tools.
• Middleware platforms that connect common business apps.
• Simple automated exports/imports on a schedule.

The goal isn’t perfect real-time sync. It’s eliminating the human copy-paste step that eats hours and introduces errors.

Measure the hours

Track time before and after automating. Not for a business case — for truth. If a process took three hours a week and now takes twenty minutes, that’s two hours and forty minutes of capacity returned to your team every single week. Multiply by the number of people doing similar work and the total gets real fast.

This measurement also tells you where to automate next. The processes that save the most time are the ones that justify the next round of investment.

AI comes after the basics

Once your rule-based processes are automated — once approvals flow automatically, reports generate themselves, and systems share data without human relay — then AI makes sense as the next layer. Not before.

AI layered on top of manual chaos produces impressive demos and unreliable results. AI layered on top of clean, automated operations produces real value, because it has structured data and defined processes to work with.

One caution before you automate anything

Automate a good process and you get speed. Automate a bad one and you get faster mistakes. Before automating a task, take an hour to confirm the underlying process is actually right — the steps necessary, the rules correct, the output what people need. If it’s not, fix it first; automation will faithfully reproduce whatever you give it, including the parts that shouldn’t exist.

Where to start

1. List the five most repetitive tasks your team does weekly.
2. Score each by frequency, rule clarity, and data movement.
3. Automate the top one — usually an approval or a recurring report.
4. Measure the time saved for four weeks.
5. Use the result to pick the next target.

You don’t need a transformation program or a new platform. You need to identify the work that a machine should be doing, free the humans from it, and repeat. That’s automation — practical, incremental, and measurably better than the alternative.