Stop Automating the Wrong Thing First: A Simple Filter for Small Businesses
July 1, 2026
The mistake almost everyone makes
When people get excited about automation, they reach for the flashy thing. An AI that writes their marketing. A clever chatbot with a personality. Then a month later nothing about the business actually feels lighter, because the flashy thing was never what was eating their week.
The task you should automate first is rarely the interesting one. It is the boring, repetitive, slightly annoying one that quietly costs you hours and occasionally loses you money. I decide what to build first with a five-point filter. You can use the same one.
Score every candidate task out of 15
Take any task you are thinking of automating and rate it 1 to 3 on each of these. Add it up.
| Factor | Score 1 | Score 3 |
|---|---|---|
| Frequency | Rarely | Many times a day |
| Time per instance | Seconds | Many minutes |
| Rules vs judgment | Needs human judgment | Clear, repeatable rules |
| Cost of delay or error | Nobody notices | Lost sale or angry customer |
| Drain on the team | Fine to do | Everyone hates it |
Anything scoring 11 or above is a strong first project. Anything under 7, leave it for now, whatever your gut says. The filter beats the gut here almost every time.
The categories that pay back first
Across the businesses I have worked with, the same handful of jobs keep scoring high. If you are not sure where to look, start here.
- Lead capture and first response. The gap between an enquiry arriving and someone replying is where money leaks. A website or WhatsApp agent that answers instantly and collects contact details usually earns its keep in the first week.
- Moving data between systems. Anything copied by hand from a form to a sheet, a sheet to a CRM, an email to a spreadsheet. High frequency, pure rules, easy to get wrong. Perfect first target.
- Support and enquiry triage. Sorting incoming messages and answering the repetitive ones. The genuinely hard questions still reach a human, faster, because the noise is filtered out.
- Booking and scheduling. Back-and-forth to agree a time, or building a fair staff rota by hand, both eat hours and both follow clear rules.
- Quotes, invoices and onboarding. The same document, the same welcome sequence, the same steps, every single time.
- Reporting. Turning raw numbers into a dashboard someone actually reads, instead of a spreadsheet nobody updates.
How this played out in real projects
I did not start clients on ambitious AI. I started on the expensive-but-boring tasks. A WhatsApp to Streak CRM link so records got updated without anyone opening a laptop. An email workflow that classified every message and auto-handled the routine enquiries. A staff shift scheduler that produced fair rotas automatically and lifted team satisfaction. A pipeline that turned raw operational data into a clean dashboard for decisions. None of these are glamorous. All of them scored high on the filter, which is exactly why they worked.
What not to automate first
Three things belong later, or never. Tasks that need real human judgment, where a wrong call is costly and hard to define in rules. Tasks that happen rarely, because you will spend more time building than you ever save. And any process that is not yet stable, because automating chaos just produces faster chaos. Get the process clean, then automate it.
Where to start this week
Write down the five tasks that annoy your team most. Score each one with the table above. Automate the highest score first, learn from it, then move to the next. That order, boring and high-scoring before exciting and low-scoring, is the whole secret.
If you would rather have someone map your top three candidates with you, that is exactly the kind of review I run. I will also tell you which ones are not worth it, which saves you more than the pitch that says everything is.