What owners get wrong about automation before they start

Before a business automates its first task, it usually already has firm views on what automation is. Those views come from adverts, from acquaintances’ stories and from offers that land in the inbox. Some of it is true. Much of it isn’t, and those wrong assumptions often decide whether the first project turns out cheap and useful or expensive and disappointing.

These are the ones we come across most.

“We’re too small for that”

Company size isn’t the yardstick. Repetition is.

A firm of a dozen people where one person spends half the day retyping delivery notes has a better candidate for automation than a large firm where every job is done differently. The real question is: which task is done every day, the same way, with information that already exists in some system or on paper that can be photographed?

Small firms even have an advantage here. Fewer people need to agree, so change goes in faster.

“First we need to buy a big system”

Many people believe they need an ERP or some large platform before automating anything. That is rarely the case. A single step can often be automated around what you already have: reading supplier invoices straight into the software you use today, connecting the online store to the stock system, a reminder ahead of a deadline.

Sometimes a big system really is needed. But it is not a precondition for the first step. It’s worse when one gets bought “so we’re ready”, and automation is then put off until it has been rolled out.

“If it isn’t right every time, it’s worthless”

People retyping data make mistakes too. The difference is that nobody measures them.

The real question isn’t whether the system ever gets it wrong, but how often, and what it does when it isn’t sure. A good system says so: a doubtful item goes to a person to check rather than being guessed. If it reads most documents on its own and sets the doubtful ones aside, people check the exceptions, not every line.

That is why accuracy should be measured on your own documents before go-live, with an actual figure. Not on the example from the demo.

“Automation means artificial intelligence”

A large share of useful automation has nothing to do with AI. A rule such as “if the amount is above the limit, ask for approval”. A link between two programs that don’t talk to each other today. A reminder three days before a deadline. A document generated from data that already exists.

AI is needed where the input is untidy: a photo, a scanned document, an email written in free text. Paying for AI where a rule would do means paying more for something less predictable.

“The biggest cost is the build”

The build price is in the quote, so it is visible. What is less visible:

• preparing data — product lists, duplicates, old records
• agreeing how the work is done before it gets automated
• exceptions, of which real work always has more than the meeting room suggested
• training and support in the first weeks
• changes after the first months, once it is clear what doesn’t fit in practice
• monitoring and maintenance while the system runs

Ask for the whole picture: what is included, what isn’t, and what the first year after go-live costs, not just the build.

“It’s quickest to do everything at once”

It looks efficient: one project, one rollout, one training session. In practice people get a pile of new things while their old work is still sitting on the desk. Errors arrive from five places at once and nobody can tell which came from where. And the first measurable benefit arrives late, so patience runs out before it does.

Step by step is only slower on paper. Each step pays for itself and shows what the next one should be.

What usually holds instead

• repetitive work decides, not company size
• the first step is built around the software you already have
• accuracy is measured, and there is a rule for uncertain cases
• AI goes only where a rule can’t
• you cost the first year, not just the build
• you go in order, not all at once

Checking whether a task is worth automating at all, before you pay for the build, is the first step of our AI automation service. Sometimes it ends with the conclusion that a simple rule is enough, with no model involved. If you have a task you’re unsure about, describe it to us.