What to automate first in a small business

Most automation projects fail because they start with the wrong question. The wrong question is “what could we automate.” Almost anything could be. The right question is “what is the repetitive thing that costs us the most and changes the least.”

Find it by asking two questions

Ask everyone on the team:

1. What did you do this week that a computer should have done?
2. What do you do every Monday morning?

The second question is the useful one. Weekly recurring work is where automation pays, because the saving repeats forever. One-off work almost never justifies the build.

You will get answers like: copying orders from email into the spreadsheet, re-typing form submissions into the CRM, chasing people for timesheets, generating the same report, reconciling two systems that do not talk.

Score them honestly

For each candidate, write down four numbers:

• Frequency. Times per week.
• Duration. Minutes per occurrence.
• Error cost. What a mistake costs when it happens.
• Stability. Has this process changed in the last year?

Annual hours saved is frequency times duration times 50, divided by 60. A task done daily that takes 20 minutes is about 83 hours a year. At a loaded cost of $40 an hour, that is $3,300 a year, every year. That is the number to compare against a build cost, and it is usually the number nobody has calculated.

Stability is the one people skip. Automating a process that changes every quarter means rebuilding it every quarter. If the process is still in flux, stabilise it on paper first. This is genuinely the most common reason automation projects become maintenance burdens.

Fix the process before automating it

If a process is inconsistent, automating it locks in the inconsistency and makes it harder to see. Write the process down as steps first. Frequently this alone removes half the work, because someone notices two steps that cancel each other out, or an approval nobody actually reads.

Automating a bad process gives you a faster bad process.

The cheap wins, in order

Most small businesses have the same first three, and none require a big project:

1. Stop re-typing data between two systems. If information is entered into one system and manually copied into another, that is an integration. Most business software has an API or at minimum a webhook.
2. Replace the shared spreadsheet that everyone edits. Not because spreadsheets are bad, but because concurrent editing with no validation is where the errors come from.
3. Schedule the recurring report. If someone assembles the same numbers every Monday, that is a scheduled job with an email attached.

Know when not to build

Do not automate when the task happens rarely, when the process is still changing, when the tool already exists and costs less than the build, or when the real problem is that nobody agreed who owns the task.

That last one is worth repeating. A surprising number of automation requests are actually unresolved ownership questions wearing a technical costume. Software will not settle who is responsible for approving something.

Start small enough to throw away

The best first automation is one that takes days rather than months. It proves the value, it teaches you where the edge cases hide, and if it turns out to be wrong you have lost days.

If you want help working out which of your manual processes is worth attacking first, describe what your team does every Monday.