Automation · 4 min
The five tasks I'd automate first in any small business
Every small business I talk to has the same five jobs being done by hand. Not clever jobs. Not jobs anyone enjoys. Just the work that fills the gap between what one system knows and what the next one needs.
None of these need artificial intelligence, a platform migration or a consultant with a deck. They need a machine doing the part a person should never have been doing.
Here they are, roughly in the order I would fix them.
1. Moving the same information between two screens
Someone reads a number off an email, a PDF or one system, and types it into another. Orders into the accounting package. Invoices into the ledger. Form submissions into a spreadsheet.
This is the most common automatable task in small business, and the easiest to justify, because the work is completely defined: read this, check it, put it there. There is no judgment in it. That is exactly why it is miserable to do and straightforward to automate.
What fixing it looks like: the document arrives, gets read, gets checked against what you already have, and lands where it belongs. A person sees the exceptions only. Typically $2,500 to $5,000 depending on how many systems have to agree.
2. Following up on quotes and enquiries
You send a quote. Then you either chase it or you forget it. Most small businesses forget it, not from carelessness but because chasing is nobody's actual job.
This is the one that usually pays for itself fastest, because it does not save time, it makes money. A quote followed up twice converts noticeably better than one sent into silence. You already did the expensive part, which was pricing the work.
What fixing it looks like: every enquiry gets chased on a schedule, in your words, until someone says yes or no. You get told when someone replies. Around $2,500 to $4,000.
3. The report someone rebuilds every Monday
Numbers pulled from two or three places into a spreadsheet, formatted, and sent round. Same shape every week. An hour or two of somebody's Monday, every Monday, forever.
Worth automating not only for the hour but because a report a person assembles by hand is a report that quietly stops happening when they are on holiday.
What fixing it looks like: it builds itself overnight and is waiting when you open your email. Same shape, same time, nobody's morning. Usually the cheapest of the five, often $2,500 or less.
4. Answering the same handful of customer questions
Opening hours. Delivery areas. Whether you cover a postcode. What a service includes. Whether you can come out on a Saturday.
Most businesses answer these dozens of times a week, and lose the enquiries that arrive at eleven at night when nobody is there to answer.
This is the one people reach for an AI assistant to solve, and sometimes that is right. Often it is not: if it is genuinely five questions, a good FAQ page and a form does the job for the price of an afternoon. An assistant earns its keep when the answers depend on your own material, your prices, your policies, your stock.
What fixing it looks like: either better self-serve pages, or an assistant that answers from your own documents and hands over when it does not know. $5,000 to $10,000 for the assistant version, plus running costs of tens of dollars a month.
5. Appointment reminders and scheduling
Confirmations, reminders, rescheduling, the phone call to the person who did not show. Every no-show is an empty slot you already paid for.
What fixing it looks like: booking, confirmation and reminders happen without anyone remembering to send them. Often this is configuration of tools you already pay for rather than something built, which makes it the one I most often tell people they do not need me for.
The honest arithmetic
Here is the part most articles like this leave out.
Say one of these takes three hours a week. Loaded cost of the person doing it, call it $30 an hour. That is $90 a week, about $390 a month, roughly $4,700 a year.
Against a $2,500 build, that is a payback of six to seven months, then it is free forever. Good, but it is not "pays for itself inside a month" and anyone who tells you it is has something to sell you.
Two things change the maths in your favor. The first is that the recovered hours usually go somewhere better than nowhere. The second is number two on this list, where the return is not saved time but closed work, and one extra job won can cover the whole build.
Where to start
Pick the one that annoys you most, not the one that looks biggest. The first automation in a business is as much about finding out how this works as it is about the hours saved, and an annoying job you understand deeply is the safest place to learn that.
If none of these describe your week, you probably do not need me yet. That is a perfectly good outcome for a free half hour.