AI automation for Australian small business, without the hype.
Fewer than half of Australian SMEs use AI at all. In construction it is under 30%. Not because it is expensive or hard — because nobody has shown them a specific, boring example that works. This page is that example, with the prices and the sources attached.
AI automation for an Australian small business typically costs $2,500 to $20,000 to build and $200 to $800 a month to run, and the right first project is whichever repeatable task is currently costing you the most money every week. For most service businesses that is answering the phone.
Only 43% of Australian SMEs report any AI adoption, according to the National AI Centre's monthly survey of 400+ SME decision-makers (December 2025 – February 2026). In Construction and Agriculture it is under 30%. Nineteen percent say plainly that they do not know how to start.
You are not behind. Most of the country hasn't started either.
There is a particular kind of anxiety that comes with reading about AI in 2026. Everyone is apparently doing it. You are apparently late. The data says otherwise, and it is worth sitting with the actual numbers for a second.
The National AI Centre runs a monthly tracking survey called the SME AI Pulse, conducted by Fifth Quadrant, sampling at least 400 Australian SME owners and decision-makers every month and weighting by industry, state and size. Across the December 2025 to February 2026 quarter, 43% of Australian SMEs reported some level of AI adoption. That was down from 45% the quarter before. The line is not climbing neatly upward — it is wobbling around somewhere below half.
Break it down by industry and it gets more useful. Health and Education sit above 50%. Construction and Agriculture are below 30% — the lowest of any sector tracked. So if you run a trade business and have not touched AI beyond asking ChatGPT to write a text message once, you are not the straggler. You are the majority.
The reasons non-adopters give are the interesting part. Around 65% cite distrust of AI decision-making or a preference to keep humans in control of their processes — which is a completely reasonable position, not a failure of education. And 19% said they simply do not know how to use AI in their business. That group is not opposed to any of it. They have just never been shown a specific example that looks like their week.
This is the gap. Not capability, not price. Nobody has walked a plumber through a boring, specific, working thing that saves him four hours. So here is the attempt.
Three tiers, and most people should stay in tier one
The most common expensive mistake is skipping to tier three because it sounds impressive. Work up, not down.
Tier 1 · Tools — about $0–$60/mo
ChatGPT or Claude at roughly $20–$30 a month. If you already pay for Google Workspace or Microsoft 365, some AI features are already sitting in there unused. Use it for quotes, emails, scope descriptions, awkward customer replies.
Spend a month here before you spend a dollar anywhere else. It costs almost nothing and it teaches you what these things are actually good at, which makes you a much harder person to sell to later.
Tier 2 · One built workflow — $2,500–$20,000
This is where the money actually is for a service business. One repeatable task, automated end to end and wired into the software you already run. AI receptionist, missed-call SMS, quote follow-up, invoice chasing.
Across the Australian market these builds commonly run $2,500–$20,000 depending on integration depth, with ongoing costs around $200–$800 a month. Square AI starts at $2,500 + $349/mo.
Tier 3 · Custom development — $35,000+
Bespoke models, deep legacy integration, genuinely complex multi-system work. Real, and occasionally the right answer — for businesses with high-volume processes and a proper data foundation.
For a business under about 20 staff, this is almost never the right first move. If someone leads with this, ask them what tier two would look like and watch their face.
Which workflow first?
Pick the row that sounds most like your week. This is the same table we work through on the free audit call.
| If this is your problem | Start with | Rough cost | What good looks like |
|---|---|---|---|
| "I miss calls when I'm on the tools" | AI receptionist | $2,500 + $349/mo | Every call answered and booked, 24/7. Details |
| "I send quotes and never chase them" | Quote follow-up | From ~$1,500 | Automatic nudges on day 3, 7 and 14. Stops the second they reply. |
| "Customers pay whenever they feel like it" | Invoice chasing | From ~$1,500 | Xero or MYOB reminders on a schedule, with a payment link. |
| "I only get a few calls but they're big jobs" | Missed-call text-back | From ~$800 | 60-second SMS. Cheapest useful thing on this list. |
| "I have no Google reviews" | Review automation | From ~$800 | Request fires when the job is marked complete. |
| "I don't know where my time goes" | AI Business Assessment | $2,500 (credited) | Written roadmap, ranked by return. 100% credited against a build within 90 days. |
| "I'm just curious, not ready to spend" | The free quiz | $0 | 12 questions, a score, and a roadmap PDF. Take it |
Costs shown are Square AI's own pricing. Quote follow-up, invoice chasing, missed-call and review builds are usually scoped together rather than sold individually, because the integration work overlaps — see Tradie Admin Autopilot. Every price is fixed and agreed in writing before work starts.
Most AI projects fail. These are the reasons
RAND Corporation interviewed 65 data scientists and engineers with five or more years of experience for its 2024 report on AI project failure. It found that by some estimates more than 80% of AI projects fail, about twice the rate of IT projects without AI. Notably, the leading causes were not technical. This is a qualitative study, so read it as "the large majority fail" rather than a precise figure — but the failure patterns are worth knowing before you spend anything.
Nobody agreed what success meant
RAND found leadership-driven failure the most common category — and the most common version of it is starting without a shared, specific definition of what the project is supposed to achieve. Your protection: name the number before anyone builds anything. Calls answered. Quotes converted. Hours back. If nobody can name it, do not start.
Chasing the technology, not the problem
RAND describes teams pursuing the latest shiny thing rather than a business outcome. It shows up in small business as buying an AI tool because a competitor mentioned it. Your protection: start from the task that annoys you most this week, not from a product someone demoed.
The data underneath is a mess
Data quality was the second most common failure cause in RAND's interviews. In a small business this is usually the client list living in three places and a pricing sheet that lives in your head. Your protection: we tell you during scoping if your data needs work first, even when that delays the sale.
It got built and then abandoned
Sponsorship fades, nobody owns it, it quietly rots. Automation is not a set-and-forget appliance — an agent that nobody reviews drifts out of date the moment your prices change. Your protection: this is exactly what the monthly retainer buys, and why we will not sell a setup-only build.
How to not get sold to
Ask where the number came from
If a vendor tells you missed calls cost Australian business "$8 billion a year" or that the average small business loses "$126,000 a year", ask for the study. We went looking. Those figures circulate between vendor blogs citing each other and we could not trace either to a primary source. A provider who cannot source their own headline stat is telling you something.
Get the price before the demo
Demos are designed to make you want it before you know what it costs. Ask for the number first. Anyone who will not give you a range before a 45-minute call is managing you. Ask specifically about setup fees, per-minute overage and what happens to the price when you get busy.
Ask what you own at the end
If you cancel, do you keep the prompts, the workflows, the integration configuration? Some providers hand it all over. Others hand you nothing, because the lock-in is the business model. Neither is dishonest — but you should know which one you are buying before you sign.
Ask where the data goes
Where is it processed? Is it used to train a model? Who can read the transcripts? These are fair questions and a real provider will answer them plainly. Worth knowing: from 10 December 2026, Privacy Act transparency obligations (APP 1.7–1.9) require covered entities to disclose automated decision-making that significantly affects people.
Pick your next step
AI receptionist
The highest-return first project for most service businesses. What it does, what the market charges, and a calculator that uses your numbers.
What it costs
Our prices and the market's prices in one table, with sources. Including the hidden costs vendors leave off the brochure.
Free readiness quiz
Twelve questions, two minutes, a score and a personalised roadmap PDF. Useful whether or not you ever hire us.
AI for plumbers
Emergency triage, after-hours calls, and the arithmetic on plumbing job values.
AI for electricians
Quote-heavy workflows, switchboard and EV enquiries, Level 2 routing.
Tradie Admin Autopilot
All five workflows as one package for trade businesses. From $2,500 + $349/mo.
Small business AI questions
It is software that completes a repeatable task from start to finish without you touching it — answering a call and booking the job, chasing a quote on day three, raising and following up an invoice. The distinction that matters is between a tool and a workflow. ChatGPT is a tool: it helps you write the follow-up email faster. Automation is a workflow: the follow-up email goes out on day three whether or not you remembered, and stops when they reply.
Fewer than half. The National AI Centre's SME AI Pulse — a monthly survey of at least 400 Australian SME decision-makers run by Fifth Quadrant — found 43% reported some level of AI adoption across December 2025 to February 2026, down slightly from 45% the previous quarter. Construction and Agriculture sit below 30%. If you have not started, you are not behind; you are roughly average.
With the task that is costing you the most money right now and happens every single week. For most service businesses that is the phone. For quote-heavy trades it is quote follow-up. For anyone with receivables it is invoice chasing. Do not start with a strategy, a chatbot on your website, or a platform migration — start with one workflow that has a number attached to it.
Two separate budgets. Tools like ChatGPT or Claude are around $20–$30 a month per person and you should just try them. Built automation is the other budget: across the Australian market, scoped single-workflow builds commonly run $2,500 to $20,000 with ongoing costs of roughly $200 to $800 a month. Square AI builds start at $2,500 setup plus $349/month. Anyone who quotes you without asking what systems you run is guessing.
Mostly for non-technical reasons. RAND Corporation's 2024 study, based on interviews with 65 data scientists and engineers, reports that by some estimates more than 80% of AI projects fail — roughly twice the rate of IT projects without AI. The leading causes RAND identified were leadership and scoping failures: no shared definition of success, chasing the technology rather than the problem, and sponsorship fading before delivery. Small businesses actually have an advantage here, because the person who decides is the person who does the work.
No, and be sceptical of anyone who says otherwise. Automation should connect the tools you already pay for — ServiceM8, Tradify, Xero, MYOB, Google Calendar. If a provider's first recommendation is migrating your business onto their platform, they are selling a platform, not solving your problem.
It depends entirely on the build, and you are entitled to ask. Ask where the data is processed, whether it is used to train anyone's model, and who can see it. Also note a real date: from 10 December 2026, transparency obligations under the Privacy Act (APP 1.7–1.9) require covered entities to disclose in their privacy policy where a computer program makes decisions that could significantly affect someone's rights or interests. It is a disclosure duty, not a ban.
Not in a business your size, and that is not the pitch. In a one-to-ten person business the work is not being done badly by a person — it is not being done at all. Nobody chases the day-seven quote. Nobody answers at 8pm. Automation picks up the work that currently falls on the floor, which is a different thing from replacing someone.
Some of it, yes, genuinely. Zapier, Make and the AI features already inside Xero and ServiceM8 will get a capable owner a long way, and a $99/mo self-serve receptionist platform is a real option if you enjoy configuring things. You are paying us for scoping, integration and the tuning that stops it decaying. If you have the time and the inclination, do it yourself — we will still tell you which workflow to start with on a free call.
Work out your first move
Fifteen minutes. We look at where your time and money actually go, and you get a one-page recommendation — including whether you should just use a $20 tool instead of hiring anyone.
Phone
0466 887 485 — Joel answers his own phone.
Response time
Within one business day. Usually within four hours.
Location
Penrith, NSW — working with businesses across Australia.
Start with one thing that works.
Not a strategy. Not a platform. One workflow, built properly, with a number attached to it.
Or call Joel: 0466 887 485
Sources
- National AI Centre, AI adoption insights: December 2025 to February 2026 — SME AI Pulse, conducted by Fifth Quadrant, minimum 400 Australian SME owners and decision-makers surveyed monthly, weighted by industry, state and employee size. ai.gov.au
- Ryseff, J., De Bruhl, B. & Newberry, S., The Root Causes of Failure for Artificial Intelligence Projects and How They Can Succeed: Avoiding the Anti-Patterns of AI, RAND Corporation, 2024 (RR-A2680-1). Qualitative study based on interviews with 65 data scientists and engineers. rand.org
- Australian Bureau of Statistics, Counts of Australian Businesses, July 2021 – June 2025, released 26 August 2025. abs.gov.au
- Office of the Australian Information Commissioner, APP 1 — Open and transparent management of personal information. APP 1.7–1.9 commence 10 December 2026 under the Privacy and Other Legislation Amendment Act 2024 (Cth). oaic.gov.au
- Market build and subscription ranges compiled from Australian AI consultancy and automation provider published pricing, 2026. Vendor-published, not independently audited.