What should a small business automate first?

Automate the job you do most often, that eats the most time each go, and that follows the same rules every time. For most small businesses that turns out to be handling new enquiries, or the handover between winning a job and starting it. It is almost never the job that annoys you most. Those two are usually different, and the gap between them is where automation money gets wasted.

Here is the method I use on every audit. You can run it yourself in a week, for free, before you talk to anyone.

Step one: find out where the week really goes

Not where you think it goes. Where it actually goes.

For one week, ask everyone who does operational work to keep a rough log. Three columns: what the task was, roughly how long it took, roughly how often it comes up. Round to the nearest fifteen minutes. You are not trying to be precise. You are trying to spot the outliers.

Two things fall out of this every time:

• A task nobody ever mentions turns out to eat several hours a week. It is so routine that it stopped feeling like work.
• A task everyone complains about turns out to take forty minutes a month. It is genuinely irritating. It is still not worth automating.

That gap — between what feels expensive and what is expensive — is the main reason automation budgets get spent in the wrong place. You cannot think your way past it. You have to write it down.

Step two: check each job against four questions

Pick your top candidates from the log. For each one, all four answers must be yes. Not three.

1. Does it happen often enough? As a rough floor: if it does not take up at least four hours a month, the cost of working it out, building it and handing it over will be more than you get back. How often matters more than how long. Five minutes twenty times a week beats two hours once a month.

2. Can you write the rules down? Not "we just know". If you cannot list the decisions as a set of rules, it is not ready. Writing them down is the right next step anyway — and it often improves the manual version enough that the automation stops being urgent.

3. Does it stay the same? A process that changes shape every few months needs rebuilding every few months. Automate the parts that have held steady for a year. Leave the rest alone until they settle.

4. Would you notice if it broke? Something has to catch it when it goes wrong. If a silent failure could run for weeks without anyone spotting it, the monitoring has to be designed before the automation. That is real work, not an afterthought.

Step three: work out the numbers

For each job that passed all four:

what it costs you now = times per month × minutes each time ÷ 60 × hourly cost
how long to pay back  = build price ÷ what it costs you now

Use the real hourly cost of whoever does the job — their pay plus tax, software and overhead, not just their salary divided by hours.

Under six months to pay back is an easy yes. Six to twelve months is a judgement call, and it comes down to how sure you are the job will still exist next year. Over twelve months is usually a no, unless there is another reason — a rule you have to comply with, or a bottleneck that is genuinely holding up growth.

Two honest adjustments. Add the cost of mistakes if the manual version loses you money or customers; that is often bigger than the time cost. Then take off the human time the automated version will still need — checking it, handling the odd case, occasional fixes. Automation almost never takes a job to zero. Any estimate that assumes it does is overselling.

Where the answer usually lands

The same few jobs come out on top in most small businesses. Not because they are interesting, but because they score well on all three counts at once.

Handling new enquiries. Frequent, quick each time, entirely rule-based — and the delay costs you money directly. An enquiry that waits a day for a reply is measurably less likely to turn into work. This is usually the most valuable automation available to a small business, and it is rarely the one people ask about first.

The handover from "won" to "started". Setting up the project, the folder, the invoice schedule, the intake form, the kickoff email. Very repeatable, spread across several tools, and almost invisible in most people's sense of where their time goes.

Regular reporting. Easy to justify because you can measure it exactly, and easy to check because you can hold the automated version up against last month's manual one. For agencies, where the work multiplies by the number of clients, this is nearly always first.

Chasing invoices and payments. Unglamorous, and reliably the fastest to show up as money. It turns "I keep meaning to send that reminder" into cash actually arriving.

Pulling data out of documents. Reading invoices, receipts and forms and getting the numbers into a system. Frequent, clearly defined, and a job people are genuinely worse at than machines by the twentieth one in a row.

What to leave alone

Just as important, and said out loud far less often.

Anything that needs real judgement each time. Pricing a one-off project. Handling a genuinely upset customer. Deciding whether to take on a difficult client. You can automate the gathering of information around these. Do not automate the decision.

Jobs that happen a few times a month. The numbers do not work, however irritating they are.

Anything you are about to change. If you are switching CRM next quarter, do not automate around the old one. Wait.

Your hardest process, as a first project. There is a strong pull towards starting with the biggest problem. Resist it. The first automation should be one that works, because that is what earns you permission to do the second. A team that watched the first attempt fail will push back on the next one no matter how good it is.

Anything where the manual process is broken. Automating a bad process gives you a bad process running faster, with fewer people watching to catch it. Fix the process first. Quite often that removes the need to automate at all.

If you only do one thing

Keep the log for a week. Most people find their answer in it without any help, and the ones who do not at least arrive at a conversation with real numbers instead of a hunch.

If you would rather have that done on your actual processes, that is what the free audit is — a written map of your workflows with hours and money attached, and an honest note on which ones to leave alone.