AI for small business: what to automate first, and what to keep human

AI for small business is mostly about time. Not robots or science projects, just taking the repeat jobs off the people who do them, so they can spend that time on customers. The trick is choosing the right first job, because automating the wrong thing creates a new mess you then have to manage.

I’ve built apps, and now I build AI agents and automations. The pattern I keep coming back to is simple: automate work that is frequent, repetitive and low-risk, and keep a person on anything that needs judgment, empathy or a decision about money.

How many Canadian businesses use AI

Statistics Canada’s Canadian Survey on Business Conditions for the second quarter of 2025 (fielded April to May 2025) gives a useful reality check (Statistics Canada, 2025; figures as reported by BetaKit). It’s a point-in-time figure, not today’s level.

• 12.2% of businesses said they had used AI to produce goods or deliver services in the previous 12 months, up from 6.1% in the 2024 survey.
• Use was highest, at roughly 30% to 35%, in information and cultural industries, professional, scientific and technical services, and finance and insurance.
• It was lowest, under 2%, in accommodation and food services, agriculture, and transportation and warehousing.

So most small businesses are still at the start. The ones I’ve seen get value didn’t just switch on a tool. They changed how the work flowed and showed people how to use it. For the marketing side specifically, see AI for small business marketing.

The test: what to automate first

Run each repeat job in your business through four questions.

1. Does it happen weekly or more?If not, it's rarely worth it
2. Same steps each time?Rules beat AI for fixed steps
3. Cheap to get slightly wrong?If not, keep a person on it
4. Can you measure the time saved?If yes, it's a good first job

Good first jobs for most small businesses

Job	Automation or AI?	Why it’s a good start
Sending every enquiry to one list	Automation	Same steps every time, stops leads getting lost
After-hours “we got your message” reply	Automation	Instant, low risk, helps speed to lead
Appointment reminders	Automation	Cuts no-shows, needs no judgment
Sorting and summarising free-text enquiries	AI	Messy input, a person still reviews
First drafts of routine replies	AI	Saves typing, a person edits and sends
Copying data between tools	Automation	Removes double entry and typos
Weekly numbers summary	Automation, sometimes AI	Same report, saves an hour on Monday

What to keep human

• Pricing and quotes. A wrong number is expensive and hard to walk back.
• Complaints and refunds. People want to be heard by a person.
• Anything sensitive. Health, money, legal or personal situations.
• The first real reply to a new lead, in most service businesses. Software can draft it. A person should read the enquiry and send it. More on that in AI lead follow-up.
• Final decisions. Who to hire, what to stop, where to spend.

An AI agent can do more than a simple rule, such as reading an email, checking your calendar and drafting a reply. That’s exactly why it needs limits. I wrote about how a small team can start with agents safely in AI agents for a small team.

How to start without a big budget

1. List your repeat jobs for one week. Write down anything you or your team do more than twice.
2. Pick one that passes the four questions above.
3. Use what you already have. Your email, calendar, form builder and CRM probably have automation built in.
4. Measure before and after. Minutes per week, errors, missed leads.
5. Add AI only where the input is messy. Free text, notes, long emails.
6. Keep a person on the approval until you trust it, then decide whether they still need to be.

How to tell if it worked

Give each automation a month, then check three things. Did it save the time you expected, counted in minutes per week rather than a feeling? Did anything go wrong that a person had to fix, and how often? And did customers notice, for better or worse? If the time saved is small and the fixes are frequent, switch it off. That’s not a failure. It’s the cheapest way to learn which jobs in your business are worth handing over, and which ones are better left with the person who knows them best.

An example, so you can picture it

This is an illustration, not a client story.

A three-person physiotherapy clinic spends about four hours a week on admin around new patients: copying form details into the booking system, sending intake forms and chasing missing ones. The first automation sends the intake form as soon as a booking is made and reminds the patient two days before if it isn’t done. No AI involved. Once that runs smoothly, the clinic adds an AI step that summarises long intake answers for the physio to read before the appointment. A person still reads every summary.

Common mistakes

• Starting with the most impressive idea. Start with the most boring repeat job. It pays back fastest.
• Using AI where a rule would do. Rules are cheaper and more predictable.
• No one owns it. Every automation needs a person who notices when it breaks.
• Automating a bad process. If the steps are messy, fix them first, or you’ll just make mistakes faster.
• Forgetting the rules. Automated marketing messages in Canada still need consent and an unsubscribe. See the CASL compliance checklist.

Where this fits

Automation is one tool for one kind of problem. If your growth is stuck because the offer isn’t clear or the ads aren’t working, software won’t fix it, which is why I start with where growth gets stuck. If you’d like help choosing what to automate, the Growth Diagnostic includes it, and my automation company is listed on the companies page. Or book a call.

Sources

• Statistics Canada. “Analysis on artificial intelligence use by businesses in Canada, second quarter of 2025.” Analysis in Brief, Catalogue 11-621-M, 2025008. statcan.gc.ca
• BetaKit. “StatsCan finds AI adoption has doubled in businesses, but hasn’t yet affected headcounts.” 2025. betakit.com

Questions people ask

What should a small business automate first?

Start with work that happens often, follows the same steps every time and is cheap to get slightly wrong: routing enquiries, reminders, data entry between tools, first drafts of routine emails and simple reports. Leave pricing, complaints and anything a customer would want a person for.

How many Canadian businesses use AI?

Statistics Canada found that 12.2% of businesses reported using AI in the 12 months before the second quarter of 2025. That was up from 6.1% a year earlier. Use was far higher in information and cultural industries, professional services and finance than in accommodation and food services.

Do I need AI, or just automation?

Often just automation. If a job is the same every time, a simple rule in your existing tools does it more reliably than AI. AI earns its place when the input is messy, such as reading a free-text enquiry, summarising notes or drafting a reply.

How much does it cost to start automating?

Less than most people expect. Many first automations use tools you already pay for, such as your email, calendar, CRM or form builder. Start with one job, measure the time it saves, then decide on the next.