The short version: automate first the steps that happen often, follow clear rules, carry little risk and mostly move information around: acknowledging inquiries, reminders, copying form details into your CRM and weekly summaries. Never fully automate decisions about money, people or safety, or anything a customer needs a person for. Fix unclear steps before you automate them.
There's no shortage of tools promising to automate your business. The hard part isn't finding a tool. It's knowing which of your tasks are worth automating at all, and which ones will cause more trouble than they save.
Here's a simple way to decide, a list of good first candidates, and the things we'd never hand over to a machine completely.
How do you decide what to automate?
List the repetitive tasks your team does in a typical week. Then score each one from 0 to 2 on five questions:
| Question | 0 points | 2 points |
|---|---|---|
| How often does it happen? | A few times a month | Many times a week |
| Are the rules clear? | "It depends" every time | You can explain it in a sentence |
| What if it goes wrong? | Costly or embarrassing | Easy to spot and fix |
| Is it mostly moving information? | Needs judgment or creativity | Copying, sorting or reminding |
| Does it annoy your team? | Not really | Everyone hates it |
Score 8 or more? It's a strong candidate. Between 5 and 7, it might be worth it after you tidy the process. Below 5, leave it alone for now.
What are good first automations?
These tend to score well in most small service businesses:
- Inquiry acknowledgements. An instant "we got your message, here's what happens next" reply.
- Routing new inquiries. Sending each one to the right person based on service, location or type.
- Form to CRM. Website form details landing in your CRM or spreadsheet automatically instead of being re-typed.
- Follow-up reminders. A nudge to the owner when a quote or proposal hasn't had a reply in a few days.
- Appointment reminders. A text or email the day before a booked visit or call.
- Invoice reminders. Polite automatic reminders when an invoice passes its due date.
- Missed-call text back. A text to anyone whose call you couldn't answer.
- A weekly summary. New inquiries, open quotes and overdue jobs, pulled together for Monday morning.
Notice what these have in common: none of them make an important decision. They make sure the right person knows the right thing at the right time.
What should you never fully automate?
- Pricing and discounts. A tool can draft a quote. A person should approve it.
- Hiring and people decisions. Screening, rejecting or evaluating people needs human judgment and care.
- Anything with legal, financial or safety impact. Contracts, refunds, safety instructions, professional advice.
- Complaints and upset customers. Automation can flag them fast. A person should answer.
- Messages you'd be embarrassed to see go wrong. If a mistake would damage a relationship, keep a person in the loop.
Where does AI fit?
Most of the automations above don't need AI at all. They're simple "when this happens, do that" rules that many tools you already pay for can handle.
AI earns its place where the input is messy: reading a free-text inquiry and tagging it, summarizing a long email thread, or drafting a first reply for a person to check. The rule we use is simple: AI can draft, sort and summarize. A person decides and sends anything important.
How do you roll out your first automation?
- Pick one task with the highest score.
- Write down the steps as they happen today, including who does what. If people disagree, settle that first. (Our guide to mapping a process in one hour walks through this.)
- Build the smallest version that works. One trigger, one action.
- Run it alongside the manual way for a week. Check every result.
- Measure the time it saves, roughly. If it saves little, stop there.
- Write down how it works and how to switch it off, so the business doesn't depend on one person remembering.
How much time will it actually save?
A rough estimate is enough to decide. Multiply how often the task happens each week by how many minutes it takes each time. A two-minute task done 40 times a week is over an hour of someone's week. A ten-minute task done twice a month barely registers. Then add a little for the mistakes it prevents, like the re-typed phone number with one digit wrong. If the total is small, a better checklist may be all you need.
What does this look like in practice?
For example, a property management company might score its tasks and find that logging maintenance requests from email and texts scores 9, while choosing which vendor to send scores 4, because it depends on the building, the urgency and the owner's preferences. The sensible first step is to automate the logging and the "we've received your request" reply, and leave vendor choice with a person, supported by a clear checklist.
What are the common mistakes?
- Automating a process nobody agrees on, so the automation bakes in the confusion.
- Buying a new platform before checking what your current tools can already do.
- Building five automations at once, so nobody knows which one broke.
- Not telling customers when they're talking to an automated system.
- No owner. Every automation needs a named person who checks it still works.
Not sure where automation would pay off? That's what our automation and AI work starts with, for example on request handling in property and facility services. A Find the Opportunity review maps one process with your team and marks the steps worth automating, before you buy anything. $500, delivered in 3–5 business days.