AI marketing examples for small business: ten, one per marketing job.

01What makes a good AI marketing example for a small business?

A good AI marketing example is one marketing job that an AI system runs every week, with a named person approving anything that reaches a customer and one number that shows whether it is working. It fits a business of 5 to 50 staff and runs on a few hours of review a week rather than a department.

Owners want examples closer to home than the global brands on most lists: Xero's May 2026 research on New Zealand small businesses found 43% want real-world case studies of AI and 47% want practical training¹.

So here are ten worked examples, one per marketing job, each with the situation, what the AI system does, where the human approves and what to measure. The businesses are illustrations, invented to make the mechanics concrete. They are not Flow AI clients, and no example claims a result.

An AI marketing system, in this guide, is a set of written instructions, data connections and review steps that lets AI run one marketing job on a schedule, with a human in the loop before anything ships. The 90-day AI marketing transformation roadmap sets the order of installs.

02Which examples lay the foundation: positioning and search?

Positioning and search come first because every other example reads from them. Positioning gives the AI system a written claim to repeat, and search puts that claim where buyers and AI answer engines look for a recommendation. With these two in place, the other eight examples produce work that sounds like the business instead of like everyone else.

1. Positioning: a one-page claim the system can read

Take a 12-person plumbing firm in Hamilton. Customers choose them for punctuality and fixed-price cylinder replacements, but nothing is written down, so every quote and web page says something different.

• What the system does: reads the last fifty quotes, the website, the Google reviews and three competitors' sites, then drafts a one-page file of who the firm serves, what it claims and the proof for each claim.
• Where the owner approves: every line. Claims and prices are the owner's decisions, and the approved file becomes the source the other nine examples read.
• What to measure: whether the same claim now appears on the home page, the quote template and the Google Business Profile. A monthly yes or no.

2. Search and AI search: answer pages for the questions buyers ask

Take an eight-person accounting practice in Tauranga. ChatGPT never names it when asked for a Bay of Plenty accountant, which matters more each year: BrightLocal's 2026 Local Consumer Review Survey of 1,002 US consumers found 45% had used AI tools such as ChatGPT for local recommendations, up from 6%².

• What the system does: monthly, it asks ChatGPT, Claude, Perplexity and Google the twenty questions buyers actually ask (taken from enquiry emails), records who gets named, and drafts one answer page a fortnight where the practice is missing, structured for answer engine optimisation.
• Where the partner approves: every fee, service and qualification before a page goes live.
• What to measure: how many of the twenty questions name the practice, plus enquiries that say they found it online.

03Which examples keep the business visible every week: content and social?

Content and social are where most small businesses already point AI, and they work best as one loop. One researched piece a week gives social something true to say, and social gives the piece an audience. The AI system drafts and repurposes, and the owner reads for accuracy and adds one real detail.

That is where businesses start: the National AI Centre's tracker for December 2025 to February 2026 found content generation and data analytics were the leading uses, each by 54% of Australian SME adopters³.

3. Content: one guide a week from questions customers already asked

Take a six-person landscape design studio in Christchurch. The designer answers the same questions on every site visit, and none of the answers are on the website.

• What the system does: transcribes the voice notes the designer records after each visit, turns the week's most common question into a guide written against the positioning file, and queues it for review.
• Where the designer approves: prices, timeframes and plant choices, plus one detail from a real job, used with the customer's permission.
• What to measure: search visits to the guides and enquiries whose first page was a guide. See the guide to content marketing for small business in NZ.

4. Social: one channel, two posts a week

Take a 15-person physiotherapy clinic in Wellington posting now and then across four platforms. The audience exists: DataReportal's Digital 2026 New Zealand report puts Facebook's ad reach at 3.45 million people, 65.6% of the population, and Instagram's at 2.65 million⁴.

• What the system does: reviews a year of past posts against booking sources, recommends one channel, then turns each week's guide into two posts scheduled in the clinic's own tool.
• Where the practice manager approves: every post, with extra care for anything that reads as treatment advice or shows a patient, who must consent in writing.
• What to measure: bookings where the intake form names social as the source.

04Which examples bring in new customers: outbound and conversion?

Outbound and conversion turn attention into enquiries. Outbound reaches people who have never heard of the business, and conversion work makes sure the people who already arrive go on to ask for a quote. Both need the closest human review of the ten, because both put words in front of a named person who can say yes or no.

5. Outbound: short, researched emails to a defined list

Take a 30-person commercial cleaning company in Auckland that wants five more office contracts. Small firms answer more often: Belkins' 2026 study of 7.5 million cold emails found reply rates of 0.72% at companies of up to 10 staff and 0.49% at 11 to 50, against 0.45% overall, measured per email sent⁵.

• What the system does: builds a list of office managers in the service area, researches each for a reason to write (a new lease, a fit-out), and drafts a short three-step sequence.
• Where the owner approves: the list and every first email. Consumer Protection's summary of the Unsolicited Electronic Messages Act says commercial messages need consent, accurate contact details and a free unsubscribe that works within five business days, with fines up to $500,000⁶, so someone owns the consent basis for each name.
• What to measure: meetings booked per hundred contacts, with unsubscribes as the early warning. See the guide to outbound sales automation.

6. Growth and conversion: quote follow-up that happens every time

Take a ten-person kitchen and joinery company in Nelson that sends forty quotes a month and follows up perhaps a third of them.

• What the system does: watches the quote folder or CRM, drafts follow-ups at day three and day ten that answer the likely objection for that job type, and flags large quotes for a phone call.
• Where the owner approves: the templates once, then any message that changes price, scope or timing. The calls stay human.
• What to measure: the share of quotes followed up, and the quote-to-job rate against the previous three months.

05Which examples handle the busy moments: launch and community?

Launch and community are the two jobs that arrive in bursts. A launch packs a month of work into two weeks, and community work (reviews, events, local groups) depends on replying quickly and personally. AI absorbs the volume in both, while the owner keeps the voice and the relationships that make either one work.

7. Launch: a kit for a new service or location

Take a 20-person café and catering business in Melbourne opening a second site, with six weeks to go and no launch plan.

• What the system does: from one approved brief, produces the location page, the new Google Business Profile description, an email to catering clients, a two-week social calendar, a note for local media and a dated checklist.
• Where the owners approve: the brief, then the whole kit in one sitting, checking dates, hours and any opening offer, since those are what customers act on.
• What to measure: catering enquiries and bookings for the new site in its first thirty days, against the target written in the brief.

8. Community: a reply to every review within a day

Take a 14-person veterinary clinic in Dunedin that replies to one review in ten. BrightLocal's 2026 survey found 89% of US consumers expect business owners to respond to reviews, and 50% are put off by generic or templated replies².

• What the system does: drafts a reply to each new review using the reviewer's specifics and the clinic's tone file, flags complaints for the practice manager, and drafts posts for the clinic's puppy classes.
• Where the practice manager approves: every reply before it posts. A public reply should never confirm details of an animal's care.
• What to measure: the share of reviews answered within a day, plus monthly review count and rating.

06Which examples show what is working: paid, data and measurement?

The last two examples close the loop. Paid and data work checks where the money went each week, and measurement turns everything above into one report the owner reads on the same day every week. The first eight examples can run without them, but nobody would know which of them to keep funding.

9. Paid and data: a weekly ad and spend check

Take a five-person mobile car detailing business in Adelaide running Google Ads and Meta on a modest budget, checked when the owner has a spare evening.

• What the system does: every Monday, pulls ad data, search terms and enquiry records, flags wasted terms, ads showing in suburbs the business does not serve and ad enquiries that never got a quote, then drafts a change list with reasons.
• Where the owner approves: every change that moves money, and any new ad copy.
• What to measure: cost per enquiry and cost per booked job, by campaign.

10. Measurement: one screen, read every Monday

Take a 25-person IT support provider in Palmerston North whose analytics, CRM and phone system have never been connected.

• What the system does: assembles a one-screen Monday report of enquiries by source, cost per enquiry, what shipped and one recommendation, plus a monthly memo on which job earned more time.
• Where the owner approves: the decision. The report recommends; the owner chooses what to fund.
• What to measure: one written decision a month.

07The ten examples side by side

The table puts all ten examples on one page with the marketing job, the illustrative business, what the AI system does, what a person approves and the number to watch. Most small businesses run two or three at once. Start with positioning unless a written claim already exists, then add the job your business is missing most.

08What do all ten examples have in common?

Every example follows the same pattern: the AI system drafts, a named person approves anything that reaches a customer or moves money, and one number decides whether the job continues. The pattern matters more than the tool. Businesses that skip the approval step tend to get caught on a claim about their own prices or services.

Approval is where many businesses are still thin. The National AI Centre found about half of current Australian SME AI users check outputs before they affect customers, and around 65% of non-adopters cited distrust of AI decision-making or a preference for human control³.

The legal exposure stays with the business whoever drafted the words: the Commerce Commission lists Fair Trading Act maximum fines of $200,000 for an individual and $600,000 for a business, per offence⁷. AI will happily write a guarantee the business never offered, so claims and prices sit on every approval list above. The guide to building an AI marketing strategy for a small business turns these rules into a one-page plan.

09Where Flow AI fits

Flow AI is an AI marketing transformation practice: the strategy, the systems and the agents that run a marketing function, engineered end to end. Each example above is a marketing job Flow AI installs as a system on the business's own accounts, with a person approving before anything ships.

It starts with a written scope, a written plan naming the first system to install and why. The methods are public: the skill library and the free agent templates can be read or run by anyone.

10The short version

Useful AI marketing examples for a small business are small and repeatable: one job, run weekly, drafted by an AI system and approved by a named person, with one number that says whether it is working. Positioning comes first because the other nine read from it.

Pick the two jobs your business is missing most, run them for a month, and let the Monday report choose the third. For the wider picture, start with the guide to AI marketing for small business in NZ.

To find out which of the ten examples your business should run first, talk to Amber.