Where to start when automating a small business
The most painful task is often the worst candidate. Four tests for choosing, a four-week run, and the only indicator that tells you anything at the start.
The questions always arrive in the same order: first "does it work?", then, once convinced, "where do I start?". The second is the harder one, and it is usually where these projects stall.
The common mistake is to start with the most painful task. It is often the worst candidate: if a task is painful, that is frequently because it is irregular, poorly defined, or needs information scattered across several places.
Four tests for a good first task
A task suits delegation when it meets four conditions. The first three are obvious once stated; the fourth is the one people skip.
- It recurs often. At least weekly. A quarterly task will teach you nothing for six months.
- It starts from written material. An email, a document, a spreadsheet. If the information lives in your head or in a phone call, it has to be written down first, and that is a separate project.
- Its output can be checked in under a minute. You must be able to say "fine" or "wrong" at a glance. Without that, you will never know whether the delegation is working.
- Failure is harmless. For the first task only, pick something where a bad result costs nothing but your time. You are learning to tune the tool; better not to learn on a customer.
A task meeting all four is rarely the impressive one. That is deliberate: the goal of the first month is not to save time, it is to find out where the limit sits.
Frequent candidates
| Task | Why it fits | The trap |
|---|---|---|
| Sorting incoming requests by type | Daily, written, result visible at once | Categories must be defined beforehand, not discovered along the way |
| Pulling data out of supplier invoices | Repetitive, stable format, checkable against the document | Check the amounts, not just the names |
| Drafting replies to recurring questions | The substance repeats; only the wording changes | Never automatic sending at the start |
| Turning meeting notes into minutes | Written material supplied, nothing to invent | Review what was left out, more than what was written |
| Updating a tracking sheet from emails | Tedious, mechanical, checkable row by row | Errors accumulate silently: inspect weekly |
A four-week run
Pace matters as much as the choice of task. This progression avoids both usual failure modes: handing over everything at once, or never daring to.
Week 1 — run it in parallel. The task goes to the tool, and you also do it the usual way. You compare. It is extra work for five days, and it is the only honest way to measure the gap.
Week 2 — fix the instruction. The gaps from week 1 rarely come from the tool alone; more often the instruction was implicit. What you knew without saying it has to be written down. This is the most useful week of the process.
Week 3 — move to review only. You no longer redo the task, you review the output. If you correct fewer than one time in five, the instruction is good. If not, back to week 2: the problem is the instruction, not the tool.
Week 4 — decide. Three outcomes, all acceptable. The task becomes routine with a quick review. Or it stays under full approval because the exposure warrants it. Or you drop it, because checking takes longer than the task saves. That last case is a result, not a failure: you now know something precise about your own operation.
What not to do at the start
- Starting several tasks at once. You will not know which one is causing trouble.
- Starting with customer-facing work. It is the area where errors show most and recover least. It comes second, not first.
- Automating sending on day one. Preparing and sending are two steps; combine them only once the first is under control.
- Judging on a single attempt. One good result proves nothing, and neither does one bad one. Consistency over two weeks is what tells you.
The indicator actually worth watching
Time saved is the metric everyone looks at, and the least reliable at the start: it is distorted by the learning curve. There is a better signal, and it is easy to observe.
Count how often you correct. If that frequency falls week over week, the delegation is taking hold, even if the time saved stays modest. If it plateaus after three weeks of adjustment, the task was the wrong one. Change tasks rather than persist: there is always a better-suited one in a small business.
To place your own tasks, the diagnostic starts from what your business does and proposes a starting point, and the employee catalog shows how the roles are divided.