A logistics owner once told us that her most expensive employee was a person she had never hired. Every evening, for two hours, someone in her family retyped the day's orders from Viber messages into a spreadsheet, and every few weeks a typo sent a truck to the wrong barangay. The cost was not the two hours. It was the quiet assumption that this was simply how the business ran.
Automation has a branding problem. To many owners it sounds like robots, large budgets and consultants who speak in acronyms. In practice, it is closer to a very patient assistant who does the dull, repeatable parts of your work the same way every time. The question is never whether to automate. It is what to automate first, because the order you choose decides whether it pays for itself in a month or never.
A story about the wrong first project
A restaurant owner we met had spent a large sum on a custom mobile app for loyalty points. It was handsome. Almost nobody used it, because the cashier had no reliable way to look up a customer in the middle of a rush. Meanwhile the kitchen still counted stock on paper, and the waste was costing more per month than the app had cost in total.
The app was not a bad idea. It was simply the wrong first idea. It touched few customers, saved no time and fixed no errors. The stock count touched every single day.
A simple way to rank your options
Take every recurring task in your business and score it on three questions, one to five each.
- How often does it happen? A daily task beats a quarterly one.
- How rule based is it? If a clear set of steps always applies, a machine can follow it. If every case needs judgment, leave it with a person.
- What does a mistake cost? Errors that lose money or customers deserve attention first.
Multiply the three scores. The tasks at the top of the list are your starting point, and you will often find that the top candidates are boring ones. That is a good sign.
Good first candidates
Order and booking intake. If customers message you on three platforms and someone copies the details into a sheet, a simple form that feeds directly into that sheet removes the retyping and most of the errors.
Payment reconciliation. Matching bank, e wallet and cash records against sales is dull, exact and exactly what software is for.
Reminders and follow ups. Overdue invoices, appointment confirmations and reorder alerts. These are small messages that humans forget and systems never do.
Recurring reports. If you assemble the same weekly summary by hand, automate it. The hours return every single week.
What to leave alone, for now
Do not automate a process you do not yet understand. If the steps change every week, or two staff members do it in two different ways, you will only automate the confusion. Write down the current process first. The writing alone often reveals steps that can simply be deleted, and a deleted step is better than an automated one.
Also hold back on anything that depends on a relationship. A thoughtful message to a key customer is not a task to hand to a script.
Automate the work you would never want to be proud of, and keep the work you would.
Plan for the failure case
Every automated process eventually breaks, so decide in advance what happens when it does. Who is notified? Is there a manual backup? Can a person see what the system did and correct it? Businesses that skip these questions end up more fragile than before, because the one person who understood the old manual process has moved on and the new one is a black box.
A good rule is that automation should reduce your dependence on any single person, including the person who built it. Document it in plain language and train two people to handle it.
Where AI genuinely fits
Artificial intelligence earns its place where the work is repetitive but not perfectly rule based. Reading free text and sorting customer messages into categories is one. Predicting how many units of a product you will sell next week, so you buy the right amount of stock, is another. These are valuable once the basics are in order, and clean data is the entry ticket. A forecast built on messy records is only a more confident guess.
The sensible order is the one most businesses skip: fix the process, clean the data, automate the steady parts, and only then add intelligence to the parts that need it.
The best automation is rarely impressive to look at. It is the evening that used to take two hours and now takes ten minutes, the error that stopped happening, the owner who goes home on time. Start there, and let the clever things wait their turn.
The examples are composites and the figures are illustrative. If you want help choosing where to begin, get in touch.