Which tasks to automate first in your company

When companies think about automation for the first time, they land on the same wish surprisingly often: something with customer communication, something with AI, something you can show off. That is understandable, but it is rarely the task that eats the most time.

The tasks worth doing first are boring. They never come up in a meeting because they don't bother anyone, they just cost twenty minutes every day. Twenty minutes on five days a week is more than 80 hours a year for a single routine step. That is two full working weeks, sliced so thin that nobody ever notices.

An example from a trades business: the owner wrote quotes in the evening, even though the data already existed somewhere, in the measurements, in the customer's email, in the price list. He retyped it. Asked what annoyed him most, he would have said "the bookkeeping". Asked what cost him the most time, the answer was a different one.

Three marks of a good first automation

Before you touch any tool, check your task against three points. All three have to apply, otherwise it costs more than it returns.

The most common mistake: it isn't the task that costs time, it's the handover

When people describe their processes, they describe activities: write the quote, send the invoice, confirm the appointment. But the time rarely disappears inside the activity, it disappears in what sits between them, in the waiting, the chasing and the moving of data from one system into the next.

So add a second column when you write things down: how often does the person or the program change in this process? Every change is a point where something stalls, gets typed twice or gets lost. Processes with three or four handovers are almost always the most rewarding candidates, even when no single activity in them takes particularly long.

The candidates that show up almost everywhere

Trades, retail, law firm or agency: five processes show up in nearly every business, and they are almost always the cheapest place to start.

• Sorting and answering enquiries. A form comes in, someone reads it, someone copies it into another system, someone writes a confirmation. The classic, because every step is rule-based and because a confirmation within seconds instead of the next day decides whether the customer keeps shopping around.
• Producing quotes and invoices. The data already exists somewhere. Typing it into a document by hand is pure transfer work, and that is exactly where the transposed digits happen, the ones that cost you an awkward email later.
• Appointments, reminders and follow-ups. Nobody should have to remember that a quote has been unanswered for ten days. That is exactly where revenue quietly disappears, not because the customer walked away, but because nobody reminded them.
• Filing and preparing receipts. Pulling invoices out of the inbox, renaming them, putting them in the right folder, passing them to the accountant. A task nobody enjoys, which is why it piles up until it hurts all at once at month end.
• Moving data from A to B. Between shop and accounting, between form and CRM, between spreadsheet and tool. Copy-and-paste is the most expensive step there is: it costs time, produces errors and leaves no trace anyone could check afterwards.

A calculation anyone can repeat

Take the first line, handling enquiries. Say twelve a day, each costing four minutes on average from reading to confirmation. That is 48 minutes daily, nearly four hours a week and around 200 hours a year. At an internal hourly rate of €60 we are talking about €12,000 of working time a year, and that for a process nobody would call a problem.

Automation rarely takes all of it. Seventy to eighty percent is realistic, and what's left are the cases that fall outside the pattern. But seventy percent of 200 hours is still 140 hours that go somewhere else, and the setup costs a fraction of that, once.

The hour that reveals your order of play

You don't need a consulting project for this, a sheet of paper and a week of attention will do. Every time a task annoys you, write one line with four entries.

• What was it? Half a sentence is enough: “retyped quote data into the document”.
• How long did it take? A rough estimate, in minutes. Precision doesn't matter here, order of magnitude does.
• How often does it come up? Per day, per week or per month. Pick one unit and stick to it.
• How often did something or someone change? Every jump between program, person or inbox counts.

At the end of the week, multiply minutes by frequency by 50 weeks for each line and sort descending. The top three lines are your candidates, and usually not the task you had in mind at the start.

Almost everyone discovers the same thing: the most expensive line isn't the longest task, it is the shortest one with the highest frequency. And very often it isn't on the list of things that annoy you day to day. It has become so normal that it only stands out once it is written down.

The best first automation is the one nobody misses once it suddenly runs by itself.

How to tell you picked the wrong task

Four warning signs show up early. If one of them applies, it is worth a step back before any money moves.

• You can't write the rules down without constantly saying “usually”. Then the process hasn't been decided yet, it gets renegotiated every time.
• Two people in the company describe the process differently. That isn't a software problem. Automation would only cast one of the two versions in stone.
• The benefit can only be described in adjectives. More professional, more modern, faster: without a number behind it, there is no yardstick to tell you later whether it worked.
• The task only exists because another system isn't set up properly. Then automation is a plaster over a cause that sits elsewhere and would be cheaper to fix.

What you deliberately don't automate first

Two things should wait. First, anything where a mistake reaches the customer directly and gets expensive, so price commitments, contract wording and anything with legal weight. Second, processes that are about to change anyway. Automating a workflow you'll rebuild in three months is double the work.

And a third point that often gets missed: if a process is broken, automation only breaks it faster. Clean it up first, then automate. That costs one conversation and saves a project.

None of this makes those areas off limits forever. They come later, once two things are in place: a process everyone describes the same way, and a defined point where a human looks at it before anything goes out.

What a first automation is allowed to cost

A workable limit is the working time it saves in the first year. With the 200 hours from the example above, that is a generous budget. For a task that costs two hours a month, the budget is very small, and that is the real answer to whether this task is worth it at all.

On top of that comes a line item worth planning for instead of suppressing: running it. An automation doesn't stay unchanged forever. A provider changes an interface, a form field gets added, a rule changes. Budget a few hours a year. Whoever plans for it isn't surprised when it happens.

Starting small isn't a compromise

The first automation isn't meant to impress, it is meant to run. A process that works reliably from week one convinces a team more than any presentation, and after that, ideas for the next steps show up on their own, usually from the team itself.

That is the real reason for the order in this piece. Not because small processes matter more, but because the first working one changes the conversation. After that, nobody argues about whether AI is worth anything, only about what to do next.