Guide
How to implement AI in your business
Start with the work, not the tools. Find out where your paid hours go, put a dollar figure on each repetitive job, build the one that pays back most, keep a person approving anything that leaves the business, and measure the result against your wage bill. Then do the next one.
By Matt Laycock and Zayn Pike. Updated 4 November 2026.
Where Australian businesses are up to
You’re not late. The National AI Centre’s survey of small and medium businesses put adoption at 43% across December 2025 to February 2026, and under 30% in construction and agriculture.
Among businesses with no plans to adopt it, about 65% point to trust or wanting people to stay in control, and 54% say it isn’t relevant to them. Another 19% of SMEs say they don’t know how to use it. Those are fair objections, and the steps below deal with each one.
Why so many AI projects stall
RAND interviewed 65 experienced data scientists and engineers about why AI projects fail. By some estimates more than 80% do, about twice the rate of IT projects without AI. That’s an estimate, not a count, but the causes they found are worth knowing:
- The wrong problem. Nobody agreed what had to change, so it gets built to the wrong measure.
- Not enough good data. The records exist, but they were kept for something else.
- Chasing the new thing. The technology gets picked before the problem does.
- No plumbing. Nothing feeds it clean data or keeps it running once it’s live.
- Too hard for AI. Some jobs can’t be automated yet, and someone should say so early.
Only the last one is about the AI itself. Each step below is there to stop one of these.
How to do it, in eight steps
Find out where the hours go
Walk through each job from first enquiry to final payment. Read your procedures, then ask the people doing the work, because the two rarely match. A fortnight of time logs tells you more than a year of opinions.
Put a dollar figure on it
Hours a week, times 52, times what an hour costs you with on-costs. A project manager’s hour costs far more than an admin’s, so the biggest savings often sit with your dearest people. Here’s the full sum.
Look in this order
First, save the money you’re already spending. Second, scale and make more: answer every enquiry and follow up every quote. Third, automate what you didn’t know you could, like reading the handwriting on a paper docket.
Check the data and the systems
Where does each job’s information live, and can you trust it? You don’t need perfect data or new software. You do need to know where it breaks, because that’s where the build starts.
Pick one build and write down what it must do
One job, one number. “Code 90% of supplier invoices overnight” beats “use AI in finance”. Agree what it should save and how that gets measured before anyone writes code. If it won’t pay, leave it out.
Keep a person in charge
Start read-only. Let it draft, then have someone approve everything that leaves the business until it’s earned their trust, and log every step. That’s the practical answer to the trust problem. About half of the Australian SMEs using AI already check its work before it reaches a customer.
Sort out privacy early
The privacy regulator’s guidance is plain: do your homework on the product, keep personal information out of public AI tools, update your privacy policy and make it clear when a customer is dealing with AI. Put rules for staff in writing, even if all you do today is use a chatbot.
Go live small, measure, then do the next one
Compare the result to what you agreed. If it hit, take the next build off the list. If it missed, find out why before you spend another dollar. RAND’s advice is to pick problems worth sticking with for at least a year, not whatever’s loudest this month.
How long it takes
Longer than the sales pitch, shorter than you’d fear. Our audits take three to four weeks, read-only, and most first builds go live in weeks, not months. The slow part is rarely the AI. It’s getting the data straight and getting a decision. Bigger systems take longer, so we build them in stages.
Do it yourself, hire someone or bring a team in?
| Route | Good for | Watch out for |
|---|---|---|
| Chatbots for staff (ChatGPT, Claude, Copilot) | Drafting, summarising and research. Cheap, and you can start this week. | Saves minutes, not roles. Keep personal information out of public tools. |
| Off-the-shelf automation (Zapier, Make) | A few apps and a simple rule. If that’s your whole problem, use one of these and keep your money. | Gets brittle as it grows, and someone has to own it. |
| Hire an engineer | Big, ongoing software work in a business that can manage it. | Takes months to find, costs $150,000 or more a year and covers one discipline. |
| Bring in an implementation team | Whole functions across several systems, where the saving is large. | Only worth it if the numbers stack up. Ask to see them first. |
When not to bother
If the repetitive work in your business adds up to a few hours a week, a chatbot subscription and a tidy process will get you most of the way. Custom AI pays when there are real wages sitting on repetitive work. If the sum doesn’t stack up, don’t build it. We tell people the same thing.
Where to from here
You can run these steps yourself with a whiteboard and a spreadsheet. Our audit is the same process done for you, with a roadmap in dollars at the end. If you’d rather see one first, ask for demo access and click through a finished audit on a made-up business.
Sources
Next step
The infrastructure gap is costing you more than you think.
The businesses pulling ahead aren’t working harder. They’re running on better infrastructure. We build it, you own it, and it compounds. 15 minutes with a founder, no slides. If there’s no saving worth chasing, we’ll tell you.
Questions, answered first
What’s the first step to implementing AI in a small business?
Write down where the hours go. Pick the three most repetitive jobs, work out what each costs in wages, and start with the one that pays back most. You can do that on a whiteboard before you spend a dollar.
Do I need clean data before I start?
No, but you need to know where it’s dirty. Most businesses run on a mix of software, spreadsheets and paper, and AI can now read a lot of that. Finding where the data breaks is part of any decent audit.
Do I have to change my software?
Usually not. A good build plugs into what you already run, like Xero, OneDrive, your inboxes and your job system.
Will AI replace my staff?
It replaces tasks first, not whole people. The hours that come back can go into skilled work and winning more jobs, or come off the wage bill when someone leaves and you don’t replace them. The audit shows you the number. What you do with it is your call.
How much does it cost to implement AI?
It depends on the work. A chatbot subscription costs little. We price each custom build after the audit, against what it saves you, and we won’t build it unless it saves significant money.