Half the headlines say every business already runs on AI and the other half say it is all hype. The payment data says most owners have bought nothing yet. Here is what is actually working in businesses like yours, and how to start without wasting a month.
SAStuart AstaFounder, SevenLens · updated 5 August 2026 · 12 min read
You are earlier than the headlines say
Surveys run hot on this subject. The US Chamber of Commerce found 58% of American small businesses saying they use AI. Bank data tells a quieter story: JPMorganChase Institute examined 4.6 million business accounts and found fewer than one in five had actually paid for an AI service by the end of 2025. In construction it was about one in eleven.
Australia and New Zealand look similar. MYOB's 2026 Business Monitor put Australian small business AI use at 40%, with 46% saying they have no plans to start, and found the adopters growing 2.8 times faster than everyone else. So if you have bought nothing and built nothing, you are not behind. You are the majority, and the advantage is still on the table.
The jobs AI does well in a business like yours
The ranking articles assume everyone works at a desk. Most of our clients do not, so this list starts where the phone rings.
Answering the phone you cannot answer
An AI receptionist takes the 7pm call, captures the job details and books the slot while you are on the tools or at dinner. Set up well, it beats the voicemail it actually competes with. Set up badly, it annoys people, so it needs a proper build and a human path for anything odd.
Drafting the words you write every day
Quotes, follow-ups, review replies, job confirmations. AI drafts them from the details you already hold, and you approve or adjust. The evenings this returns are usually the first benefit owners can feel.
Chasing what needs chasing
Unpaid invoices, unanswered quotes, abandoned bookings. The system watches and nudges politely, so nothing depends on your memory.
Reading and summarising paperwork
Supplier terms, tender documents, long email chains. A usable summary in a minute, with the original still there to check.
Reporting without the spreadsheet evening
Cash, pipeline and job status pulled into one plain view that updates itself.
Where it goes wrong
Xero's Future Focus research across six countries found 69% of owners using generative AI had hit drawbacks, and 16% actually lost time checking its output. The failure patterns repeat, and they are avoidable.
AI slop in front of customers
People can tell. A padded, cheerful, machine-written post or reply erodes trust with the exact people you want to keep. Use AI for drafts and logistics, and keep the public voice yours.
Automating a mess
AI wired into a chaotic process produces chaos faster. Fix the process, then speed it up.
Tool-collecting
A subscription for everything and a result from none of it. One task fixed end to end beats five tools half-adopted.
Trusting it with judgment
It drafts, summarises and schedules well. It should not price unusual jobs, handle upset customers or make hiring calls. Keep the edges human.
Feeding it what it should not hold
Customer records do not belong in free consumer chat tools. Business-grade tools with sensible data settings exist. Use those.
PUT A NUMBER ON IT
Ten minutes with the savings calculator settles most of this.
The un-gated calculator estimates what repetitive admin costs your business a year and what AI would plausibly return. No email required, the assumptions stay visible, and you keep the number either way.
A standard problem, like phone answering or email drafts
A built system
Project fee, light upkeep
Weeks, for one workflow
A defined process is provably costing real money
Do nothing yet
Nothing today
Not applicable
No repeat work at volume, or bigger fires first
The test does not change with the route: the first thing should pay for itself inside a year, on arithmetic someone shows you before you commit.
If you run a trade
The question tradies actually ask is how any of this fits a plumbing, building or electrical business, and the honest answer is: better than it fits most desk businesses. The pain is specific. Calls missed on the tools become jobs lost to whoever answered. Quotes written at 9pm follow the same shape every time. Invoices go unchased because chasing is nobody's favourite part of the evening.
Each of those maps to a fix that exists today: an answering system that captures job details after hours, quote drafting from the details you already hold, invoice chasing that runs itself, and review asks fired on job completion. None of it requires you at a desk, which is the test that matters for a business run from a ute.
Free help before you pay for any
Both governments are subsidising the learning curve. In Australia, TAFE NSW and the National AI Centre run a free introduction course under a scholarship program aimed squarely at workers and small business owners. In New Zealand, the AI Advisory Pilot announced in January co-funds professional AI advice at 50%, up to $15,000, through the Regional Business Partner Network, and business.govt.nz publishes plain-language guidance built around running one pilot task.
The sequence that costs least: use the free material to learn the vocabulary, run one small pilot yourself, and bring in paid help only once you know which workflow deserves building properly. Paying to be told what a free course teaches is the most common first mistake in this market.
How to start without wasting a month
Pick one task, not a strategy
The repetitive job that costs you the most evenings. For most trades and service businesses that is enquiry handling or quoting.
Put a number on it first
Times per week, minutes per time, your rate. If the number is small, stop here and keep your money.
Run it for two weeks with a human checking
Let the tool draft while a person approves. Two weeks tells you whether it is saving time or generating homework.
Then decide like a buyer
Keep it, fix it or drop it, based on the hours that came back. Then, and only then, pick the second task.
When doing nothing is the right call
Xero's research found only 8% of owners rank AI as their top priority, and the other 92% are not wrong. Revenue, customers and costs come first. AI is a way of serving those, not a rival to them.
If the week has no repeated work worth automating, or the business has a bigger fire, doing nothing yet is the right call. Anyone selling you urgency is selling.
COMMON QUESTIONS
Questions people ask next.
01Do I need technical skills to use AI in my business?
No. The mainstream tools work in plain English, and a builder can do the wiring once. The skill that matters is knowing your own process well enough to describe it, which owners already have.
02Can AI replace my staff?
It removes tasks, not judgment. In practice businesses reinvest the hours: the US Chamber found most small businesses adopting AI grew their teams rather than cutting them. The honest promise is fewer evenings on admin, not an empty office.
03Is my business data safe with AI tools?
It depends on the tool and its settings. Business tiers of the major tools do not train on your data by default, while free consumer tools may. The working rule: customer records never go into anything free, and whoever sets up your tools should be able to say where the data goes in one sentence.
04Is an AI receptionist worth it for a small business?
Count your missed calls for a fortnight first. If jobs are going to whoever answered, it usually pays for itself quickly, because it competes with voicemail rather than with a great human receptionist. If you rarely miss calls, spend the money elsewhere.
05How much does AI cost for a small business?
Anywhere from nothing, for drafting with free tools, to monthly subscriptions for specific jobs, to a project fee for a system built around your workflow. The spend is the wrong place to start. Cost the problem first, then let the payback arithmetic pick the route.
06What should I automate with AI first?
The task you repeat most that follows the same steps: usually enquiry follow-up, quoting or invoice chasing. One task, two weeks, a human checking the output, then a decision made on numbers.
IF IT IS USEFUL
Want a straight answer for your business?
SevenLens runs a free 20-minute call. Bring one workflow and the numbers you know, and we will tell you what it is costing and whether fixing it is worth the money. If it is not, we will say so, which is the whole point of asking.