Twelve quick wins a small business can set up this week: a business brief for your AI tool, reply templates for top enquiries, a review-reply bank, meeting summaries, voice-note tidying, payment reminders, a month of captions, a plain-English policy rewrite, spreadsheet formulas, an enquiry-summary automation, a proposal checker and training switched off. Most take under an hour.
All but one need nothing beyond a $20-a-month AI plan and the software you use now; number 12 needs a paid automation plan. Don't try all twelve. Pick three or four that match your week, set them up, and see which ones you're still using a fortnight from now. That's the real test of a quick win.
Fifteen to thirty minutes each
1. Switch off model training on personal AI accounts
If anyone in the team uses a personal ChatGPT, Claude, Gemini or Copilot account for work, their chats may be used to train the vendor's models unless they switch it off. It's a few minutes' work per account, and it's the one item on this list with no downside. There's a list of where each training switch lives, vendor by vendor.
In practice it's a ten-minute sweep, not a policy exercise. An illustrative four-person photography studio found three accounts on its first pass: the owner's ChatGPT Plus, with training still on because nobody had opened the settings since signing up; a second shooter's free Gemini account on his phone, used for caption ideas; and the editor's free Claude account, used to reword client emails. All three had the model-training switch turned off that afternoon. The editor's chat history also held two client emails pasted in full, names and phone numbers included, which is exactly what the watch-out below is about.
Watch out: the switch covers future chats only. Anything with client details in it should move to a company account anyway.
2. Write a one-page brief and save it in your AI tool
Every other item on this list works better when the AI knows your prices, policies and tone. An illustrative photography studio might write: services and packages with prices, booking and deposit rules, turnaround times for galleries, the tone ("friendly, no jargon, never pushy") and one instruction for gaps: "If something isn't covered here, say I'll confirm it." Save it as custom instructions, in a project, or as a pinned document you paste in.
The difference shows on the first question you test. Asked "How long until I get my wedding gallery?" with no brief, a typical answer is (illustrative):
Most wedding photographers deliver galleries within 4 to 12 weeks, depending on the season and the number of images. It's best to check your contract for the exact timescale.
With the brief saved, the same question gets:
Your gallery will be ready within six weeks of the wedding, and we'll send a sneak peek of 20 images within a week. If you need anything sooner for a particular date, let me know and I'll confirm what's possible.
The first answer is accurate and useless to a client. The second matches the studio's real turnaround and falls back on the "I'll confirm" line for the one thing the brief doesn't cover.
Watch out: the brief goes stale the day your prices change. Put a review date at the top.
3. Turn on meeting summaries in your existing video tool
Zoom dropped the "AI Companion" name in June 2026, so look for the feature by what it does: meeting summary is still included on eligible paid Zoom Workplace plans. Gemini in Meet comes with Google Workspace Business Standard and above. A wedding planner who spends hours on client calls can get a summary and action list after each one instead of typing notes. Using Take Notes for Me in Google Meet walks through the Google version.
After a 40-minute planning call, the summary looks something like this (illustrative, shortened):
Decisions: ceremony moves to 2pm; florist confirmed; favours capped at $400. Actions: planner to send the revised timeline by Friday; couple to confirm final guest numbers by the 12th.
What it left out was the aside at minute 31, when the couple agreed to drop the second photographer to pay for a later bar: a $900 line in the budget, say. The planner caught it because she reads every summary against her own three-line note of anything involving money, which takes two minutes and is worth keeping even when the summaries are good.
Watch out: tell people at the start of the call that notes are being taken by AI, and read the summary before relying on it. Summaries miss quiet agreements ("fine, let's drop the second photographer") more often than headline decisions.
4. Build a bank of review replies
A tattoo studio with a backlog of unanswered reviews can clear it in one sitting. Paste in the text of recent reviews (not the reviewer's other details) and ask for replies:
Here are 10 recent customer reviews of my [business type]. For each,
write a reply of 2-3 sentences in a warm, plain tone. Thank them for
something specific they mentioned. For any complaint, acknowledge it
without arguing, and invite them to contact us directly. Don't use
the same opening twice. Don't offer refunds or discounts.For a mixed review such as "Love my tattoo, but I waited 40 minutes past my appointment with no explanation", the draft that comes back tends to read (illustrative):
Thank you so much for your kind words about your new piece! We're sorry about the wait on the day, and we've spoken to the team to make sure it doesn't happen again. Please get in touch directly if you'd like to talk about it.
Two fixes before posting. The exclamation mark sounds breezy next to a complaint, and "we've spoken to the team" is only acceptable if someone actually has. The reply the studio posted gave the real cause (an earlier session overran) and what had changed: clients now get a text if their artist is running more than 15 minutes late.
Watch out: edit every reply before posting. Identical-sounding replies across a profile look automated, which is worse than no reply.
5. A checklist prompt for outgoing proposals
Before a wedding planner sends a proposal, a second read catches the wrong date, a missing inclusion or a mismatched couple's name copied from an old template. AI makes a decent second reader if you give it a checklist:
Check this proposal against the checklist. For each item, say PASS or
FLAG and quote the line that caused a flag.
1. Client names match throughout.
2. The event date and venue match throughout.
3. Every inclusion in the summary appears in the detailed section.
4. Payment schedule and deposit terms are stated.
5. No leftover text from another client (other names, dates, venues).
Don't check the arithmetic; I'll do that myself.Run on a 9-page draft, the reply comes back in a form you can scan in 20 seconds (illustrative):
1. PASS
2. FLAG - cover page says "Saturday 14 June"; timeline on page 6
says "Saturday 21 June".
3. FLAG - summary lists "welcome drinks reception"; not found in
the detailed section.
4. PASS
5. PASSFlag 2 was the real catch: the date had been changed on the cover when the couple moved the wedding, and nowhere else. Flag 3 turned out to be worded differently in the detail ("arrival cocktails"), which is still worth knowing, because a couple comparing the two pages might wonder the same thing.
Watch out: the last line is deliberate. Language models are unreliable at sums, so check totals yourself; why AI is bad at maths explains why.
About an hour each
6. Reply templates for your ten most common enquiries
A driving school answers the same questions every day: lesson prices, block-booking discounts, availability, what happens on a first lesson, cancellation rules. List your ten, ask your AI tool to draft a reply to each from your brief, edit them, and save them as email templates. In Gmail, templates are switched on under Settings, See all settings, Advanced, Templates; you can store up to 50, and they work in Gmail on the web. In new Outlook, the My Templates option on the message ribbon does the same job.
The "first lesson" template, after the owner's edits, read like this (illustrative):
Your first lesson is two hours and we'll pick you up from home. We start somewhere quiet, go through the controls step by step, and most learners are driving on quiet roads within the first half hour. Please bring the licence you'll be learning on, and your glasses if you need them to drive. Lessons are $45 an hour, or $400 for a block of ten booked together.
The AI's first draft had said the first lesson was free. It wasn't any more: an old promotion was still sitting in the brief. That's the most common way templates go wrong, and it's why the price line goes last, where it's easy to find and update.
Watch out: templates with prices need updating when prices change. Keep a list of which templates mention money.
7. Turn voice notes into tidy records
A personal trainer who records a 60-second voice memo after each session can have AI turn it into a structured note: what was done, how the client found it, what to do next time. Most AI apps accept voice input directly. Use a fixed structure so every note looks the same:
Turn this voice note into a session record with these headings:
Session focus / What we did / How it went / Next session plan.
Keep my wording where possible. Don't add anything I didn't say.A transcribed memo such as "Lower body today, goblet squats three by ten at 16 kilos, then lunges, she found the lunges hard on the left side, next time maybe split squats, keep the weight the same" comes back as (illustrative):
Session focus: lower-body strength. What we did: goblet squats 3 x 10 at 16kg; lunges. How it went: found lunges challenging on the left side; no pain reported. Next session plan: split squats at the same weight.
"No pain reported" was never said. The AI filled a gap under a heading that seemed to want one, despite the instruction not to. On a session record that could matter later, so skim each note for anything you didn't say before saving it.
Watch out: health and injury details are among the most sensitive data a business holds. Keep them out of personal AI accounts, and describe them in general terms if you use AI at all.
8. A three-step payment reminder sequence
Write the reminders once, properly, instead of agonising over each chase. Ask for three versions for an overdue balance, say a yoga studio's unpaid class pack: a friendly nudge at 3 days, a clear reminder at 10 days, and a firm final note at 21 days that says what happens next. Chasing late payments with AI reminders covers automating the send once the wording works.
The 21-day note is the hardest one to write yourself. A good draft, trimmed by the owner, looks like this (illustrative):
Hi [first name], the $150 for your class pack is now three weeks overdue. As mentioned in our earlier messages, we'll pause bookings on your account from [date] until it's settled. You can pay online here: [link]. If something's making it difficult, reply to this email and we'll work something out.
The owner kept the last sentence because most late payers are disorganised rather than unwilling, and "pause bookings" stayed because the booking system can actually do it.
Watch out: the final reminder should only state consequences you will actually carry out.
9. A month of social captions in one sitting
A photography studio can list 12 to 15 images it plans to post (a sentence each on the shoot and the story behind it) and get a month of captions in one go. Ask for three options per image in different lengths, then pick and edit.
For a line such as "Newborn session: baby asleep in her dad's hands, both parents were nervous about holding her for the shot", the short option often comes back as "Pure magic. There's nothing quite like those first precious moments" (illustrative). It could caption any newborn photo ever posted. The option worth keeping used the detail you gave it: "Dad was sure he'd drop her. He didn't. Ten minutes later she was fast asleep." Add "magic" and "precious" to the banned list and the next batch starts from the detail instead.
Watch out: AI captions drift towards the same breathless style. Include two of your own past captions in the prompt as examples of your voice, and ban the phrases you'd never use.
10. Rewrite one confusing policy in plain English
Pick the policy customers misread most often. For a driving school it's usually cancellations: how much notice, what's charged, what happens with a test booking. Ask AI to rewrite it at a reading age of about 12, then to list the five questions a customer would still have. Answer those in the policy itself.
Here's the kind of before and after you get (illustrative). The original: "Lessons cancelled with less than 48 hours' notice shall be chargeable in full save where the instructor, at their sole discretion, deems the circumstances exceptional." The rewrite: "Need to cancel? Tell us at least 48 hours before your lesson and you won't pay anything. With less notice, you'll pay for the lesson in full, unless something serious has happened, such as illness or a family emergency." The AI's list of remaining questions included "Does the 48 hours count weekends?" and "Can I move a lesson instead of cancelling it?", both of which the owner had been answering by phone every week. Note that "such as illness or a family emergency" wasn't in the original. It may be what the owner does in practice, but it's a new promise, so it stays only if they mean it.
Watch out: plain English can change meaning. Compare the new version line by line with the old one, and if the policy has legal weight (refund terms, for example), have whoever wrote the original check it.
Half a day
11. A weekly report from your booking export
Most booking systems export to a spreadsheet. A yoga studio can export a month of attendance and ask AI to write the formulas for a weekly view: attendance by class, by teacher and by time slot, and which classes ran below half full. Describe the column headings rather than uploading member names, paste the formulas into your sheet and check a few rows by hand. Google Sheets AI prompts has more prompts in this style.
Say the export has columns Date, Class, Teacher and Status, where Status is Attended, Cancelled or No-show. Asked for "attendance per class", the first formula the AI suggested for the slow-flow class was =COUNTIF(B:B,"Slow flow"), which counts every booking. It gave 38 for the month. The corrected version, =COUNTIFS(B:B,"Slow flow",D:D,"Attended"), gave 29. Spelling out the Status values in your prompt usually gets the right formula first time.
Watch out: check the first week's figures against the booking system's own reports. Formulas that look right can still count cancelled bookings as attended.
12. Enquiry form to AI summary to notification
This is the only true automation on the list. When a tattoo studio's enquiry form is submitted, an automation reads it, asks AI to pull out style, placement, size, budget and preferred dates, and sends a tidy summary to the studio's inbox or chat. The artist sees a consistent summary instead of a wall of text.
A rambling four-paragraph enquiry turns into something like this (illustrative):
Style: fine-line botanical, black and grey. Placement: inner left forearm. Size: about 15cm long. Budget: "around $300". Dates: weekday afternoons in October. Missing: no reference images attached; doesn't say whether it's a first tattoo.
Ask for the "Missing" line explicitly in the AI step. It's the most useful part, because the artist can reply with one question instead of discovering the gaps three emails in.
In Zapier this is three steps (form trigger, AI step, email), and the Free plan only allows two-step Zaps, so you'll need a paid plan: Professional starts at $29.99 a month billed monthly, or $19.99 billed annually, for 750 tasks. Zapier's help centre says an AI by Zapier step uses one, three or five tasks per run depending on the model tier you pick. Adding AI steps to Zapier shows the setup.
Run the sum before you pick a plan. The form trigger costs nothing, the AI step costs one, three or five tasks, and the email step costs one, so each enquiry uses two to six tasks. At an illustrative 60 enquiries a month that's 120 to 360 tasks, comfortably inside Professional's 750 even on the dearest model tier. At 200 enquiries on the top tier (1,200 tasks) you'd be over, and the cheaper tier becomes the obvious first thing to test.
Watch out: test it with ten real past enquiries before switching it on, including the messy ones. And keep the automation to summarising; replies still go through a person for now.
Picking your three: match them to where your week goes
The right quick wins depend on what fills your week, not on which sound most impressive. Use the chore that eats the most time as your guide:
| If most of your week goes on… | Start with |
|---|---|
| Answering the same enquiries | 2, 6 and, later, 12 |
| Client calls and consultations | 3 and 5 |
| Chasing money and explaining policies | 8 and 10 |
| Marketing and your online reputation | 4 and 9 |
| Writing up sessions, jobs or visits | 7 and 11 |
The driving school shows how the table works in practice. Its owner's week goes on two things: the same pricing and availability questions by email, and learners phoning to ask whether a late cancellation will be charged. That points to 2 and 6 for the first, and 10 for the second, because a clearer cancellation policy stops the calls at source. Captions and meeting summaries sound appealing, but the school posts twice a month and has almost no client calls, so they wait.
Whatever you pick, add number 1. It takes minutes and protects the rest.
A five-day plan if you want to fit several in
| Day | Set up | Time |
|---|---|---|
| Monday | 1 (training off) and 2 (business brief) | About 45 minutes |
| Tuesday | 6 (reply templates) | About 1 hour |
| Wednesday | 3 (meeting summaries) and 4 (review replies) | About 45 minutes |
| Thursday | One of 7 to 10, whichever matches your biggest weekly chore | About 1 hour |
| Friday | 11 or 12, if you have a spare afternoon | 2 to 4 hours |
Monday comes first for a reason: item 1 protects everything you do afterwards, and item 2 improves every draft that follows.
Two weeks on: which quick wins actually stuck
Quick wins fail quietly. Nobody decides to stop using the reply templates; they just drift back to typing from scratch. So put a 15-minute review in the diary for two weeks after you start and ask three questions of each item you set up:
- How many times was it used? Fewer than three times in two weeks means it doesn't fit your week. Drop it without guilt.
- What did it save? A rough figure is fine: "templates cut enquiry replies from about 8 minutes to 3".
- What went wrong? One wrong price in a template, one summary that missed a decision. Fix the cause, not just the instance.
Filled in by the driving school after its fortnight, the review took ten minutes (figures illustrative):
ITEM USED SAVED WENT WRONG
6 Reply templates 31 ~5 min each, about 2.5 hrs one template quoted
last year's block price
10 Cancellation once calls about cancelling none so far
policy rewrite fell from ~6 to 2
4 Review replies 1 ~10 min nothing; just not needed
at 2 reviews a monthTemplates stay, with the price line fixed in every template that mentions money. The policy rewrite was a one-off that keeps paying. Review replies go, because two reviews a month don't justify a bank of replies, and that's a perfectly good result for a quick win: it cost an hour to find out.
Keep the two or three that pass. Those are your evidence for where AI fits in your business, and the best candidates for turning into a proper workflow.
Further reads
- How to Get Started With AI in Your Small Business: First 7 Steps — Turn the quick win you keep into a proper two-week trial.
- What Should a Small Business Automate First With AI? — How to rank bigger tasks once the quick wins are done.
- How to Build a Shared Prompt Library for Your Team — Somewhere to keep the prompts from this list so the team can reuse them.
- AI Workflow Automation Examples: 20 Processes to Automate First — Twenty larger automations for when number 12 works.
- How to Get More Google Reviews With Automated Review Requests — The other half of number 4: getting more reviews in.
- How to Draft Customer Email Replies With AI That Sound Like You — Going further with number 6 so replies sound like you.
- How to Use AI on Your Own Admin First, Then Roll It Out — A six-week owner-first plan: pick three of your own admin jobs, log what AI saves and gets wrong, write a playbook, then hand one task to one person.
- AI Use Cases by Department for Small Businesses (With Examples) — 21 practical AI use cases across seven departments, each shown in a small insurance brokerage, with how to start and what to watch.
- AI Myths That Stop Small Businesses Getting Started — Ten beliefs that keep small firms from trying AI, what is actually true about each, and a small first step that tests it within a week.
- Barriers to AI Adoption in Small Businesses and How to Clear Them — Eight practical barriers that keep small firms from adopting AI, the cheapest way past each, and the few barriers you should respect rather than clear.
- How to Learn AI as a Business Owner: A 30-Day Self-Study Plan — A day-by-day 30-day plan for owners to learn AI on real work, from first prompts to one working automation, with free courses and a day-30 self-test.
- What Can AI Realistically Do for a Small Business? 15 Tasks — Fifteen jobs AI can realistically take on in a small business, rated by how much you can hand over, with tools, costs and the catch for each.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Gmail Help on templates; Microsoft Support on My Templates in Outlook; Zoom Workplace plan and meeting-summary pages; Zapier help centre on the Free plan and AI by Zapier task pricing. Checked September 2026.