Automate the handling, not the thinking. The first thing worth automating in most small businesses is a task that happens often, follows the same steps every time, has an obvious right answer, and currently depends on somebody remembering to do it.
That is a narrow description on purpose. Most disappointing automation projects begin by trying to automate a decision rather than a chore.
Score the candidates
List the repetitive tasks in your week and score each one from one to five on five dimensions.
Frequency
How often does it happen? Daily beats monthly. A task done twice a year rarely repays the effort of automating it, however irritating it is.
Repetition
Does it follow the same steps each time? If every instance is a little different, you are looking at judgement wearing the costume of a routine.
Clarity
Could you write the rule down in a few sentences and hand it to a new starter? If not, the rule is not ready to be encoded.
Error risk
What happens when it is done wrong or skipped? Retyping an address wrongly into an invoice is a different order of problem from forgetting to file a note.
Handoff friction
Does the task sit between two people or two systems, where things stall while someone waits? Handoffs are where work goes quiet, and they are often the highest value place to start.
Tasks that usually score well
- Confirmations and acknowledgements when something is received, booked or completed.
- Copying the same details from one place into another, such as an enquiry into a job record.
- Reminders before an appointment, a renewal, a document expiry or a deadline.
- Scheduled reports that somebody currently assembles by hand each week.
- Routing an incoming request to the right person or queue based on a clear rule.
- Filing documents into a predictable structure with predictable names.
None of these are glamorous. That is the point. They are stable, dull, frequent and easy to verify.
What not to automate first
- Pricing and quoting where discretion is genuinely applied.
- Anything involving a judgement about a person, such as suitability or credit.
- Messages that carry apology, negotiation or bad news.
- A process nobody can currently describe consistently.
- A process that is about to change anyway.
- Anything where a wrong automated action is hard to reverse.
Automating an unstable process locks in whichever version happened to be current on the day it was built.
Start with one narrow workflow
Pick a single workflow with a clear trigger and a clear finish. One trigger, a few steps, one visible result. Run it alongside the manual version for a short period so you can compare outputs before you trust it.
Decide in advance what happens when it fails, because it will occasionally fail. Who is told? Does the work queue up somewhere visible, or does it silently disappear? Silent failure is worse than no automation, because people stop checking.
Measure it honestly
Before you start, note what the task costs today in rough terms: how many times a week, how long each time, how often it goes wrong. Afterwards, compare. Use your own numbers. Industry averages tell you nothing about your business.
We are careful not to promise a saving in advance. What can be said honestly is that removing a repeated manual handoff removes the chance of forgetting it.
When the answer is to do nothing
If a task takes two minutes a week, automating it is a hobby rather than an improvement. If a well supported product already does the job as part of something you pay for, switch that feature on instead of building anything. The best automation decision is often to use the tool you already own properly.
Once one narrow workflow is running quietly and correctly, the next candidate is usually obvious, because the people doing the work will tell you.