Most small businesses don't need an "AI strategy". They need a few hours back each week. The fastest way to get them is to automate three small, boring tasks well, not ten big ones badly.
Here is how we help teams pick those first three. You can do all of it with a notebook and a spreadsheet, with or without AI.
Step 1: Log your time for one week
Before you choose a tool, find out where the hours actually go. Guessing doesn't work. The tasks that feel heavy are often not the ones that eat the most time.
For one normal week, ask everyone on the team to keep a simple log. Each time they switch tasks, they write down:
- What they did ("answered price questions on WhatsApp")
- Roughly how long it took
- How often it happens (daily, weekly, per customer)
- Where the information came from and where it went
Keep it rough. Fifteen-minute blocks are fine. The point is to see patterns, not to watch people.
At the end of the week, group similar entries together. You now have a list of recurring tasks. That's your raw material.
Step 2: Score each task
Not every repetitive task is a good first automation. Give each one a score from 1 to 3 on four questions:
- Frequent. Does it happen often? Daily beats monthly.
- Repetitive. Is it done the same way each time?
- Rule-based. Could you write the steps as "if this, then that"? If it needs judgement every time, it's harder to automate.
- Low-risk. If the automation gets it wrong, is the damage small and easy to fix?
Add up the scores. The tasks at the top are your candidates. A task that scores high on the first three but carries real risk isn't off the table. It just needs a person to check the result before it goes anywhere.
Say you run a small bakery that takes custom cake orders. Copying each order from Instagram messages into a notebook happens daily, the same way, with clear rules, and a mistake is easy to spot. That's a strong candidate. Deciding which customer gets a discount is not.
Step 3: Check the usual suspects
Your log is specific to your business. Still, the same kinds of tasks show up in many small businesses. Look for these on your list.
Answering the same customer questions
Opening hours, prices, delivery areas, "do you have this in stock?". If your team types the same answers on WhatsApp or email every day, start with saved replies. Then move to an assistant that drafts replies for a person to approve.
Moving leads into one list
Leads arrive from a website form, Instagram, WhatsApp and phone calls. They end up in five places, and some get lost. Sending every form entry into one sheet or CRM automatically, with the date and the source, is often the quickest win.
Quotes and invoice reminders
Building a quote from a template, and sending a polite reminder when an invoice is overdue. The template part is easy to automate. Reminders should follow rules you set, and the amounts should come from your accounting system, not from AI.
Social post drafts
Turning a product photo and a few notes into a first draft of a caption, in Arabic and English. A person still edits it and posts it.
Weekly reports
Pulling numbers from sales, ads or bookings into one short summary every Monday. Nobody enjoys copying numbers between tabs.
Where AI helps, and where it shouldn't act alone
Plenty of good automations don't need AI at all. A form that writes to a spreadsheet is plain automation, and it's more reliable for it. Use AI where the task involves language or messy input.
AI is useful for:
- Drafting: replies, captions, quote descriptions, first versions of emails
- Summarising: long email threads, meeting notes, customer feedback
- Sorting: tagging incoming messages as "price question", "complaint" or "new order"
AI shouldn't act alone on:
- Anything sent to a customer without review. A wrong answer in your name costs more than the minutes you saved. Let AI draft. Let a person send.
- Money. Prices, discounts, refunds and payment amounts should come from your systems and rules, not from a model's guess.
- Personal data. Be careful what customer details you paste into AI tools. Check where the data goes and who can see it. Keep sensitive information out unless you're sure.
A simple rule: AI can prepare the work. A person approves anything that leaves the business.
Start with one, measure, then the next
You have your top three. Don't build them all at once.
Pick the one with the best score and the lowest risk. Before you switch it on, note how long the task takes today, using your time log. Run the automation for two to four weeks. Then check:
- How long does the task take now?
- How often did the automation get it wrong?
- Does the team actually use it, or work around it?
If it saved time and people trust it, move to the second. If not, fix it or drop it. A dropped automation is a lesson, not a failure.
Once all three are running, log your time for another week. Your list will look different, because the work has changed.
Give every automation an owner
This is the step most teams skip, and it's the one that decides whether an automation lasts.
For each automation, name one person on the team who owns it. They should be able to explain what it does, where the data goes and what to do when it breaks. Write it down in a short note: what starts it, the steps, the tools it uses and who to call.
Automations break quietly. A form field gets renamed, a password expires, a tool changes its plans. If nobody understands the setup, the business depends on something no one can fix. It's worse if the person who built it was an outside freelancer who has moved on.
That's why, when we build automations for clients, we train the team to own them. The goal is simple: your people can read, adjust and switch off every automation without us.
Want to work through this with your team?
If you'd rather do this together as a team, our AI and Automation for Business workshop is a one-day session on exactly this topic. You can see it with the rest of our workshops.
And if you'd like a second pair of eyes on your list, book a free 30-minute call. We'll tell you honestly which tasks are worth automating first, and which aren't.