Where should I start with AI in my Australian business?

Start by putting a dollar figure on where your business already loses time and money each week, before you buy more AI tools. A licence in the building is not the same as a working system. At Versantly, that first figure comes from the 45-minute diagnostic. Then you can decide what, if anything, to fix first.

Why start with a number, rather than another tool?

The first useful step is a cost, not a shortlist of products. Until you know what a weekly process is already costing you, you cannot tell whether a new tool would help. The 45-minute diagnostic is how we get that figure at Versantly. You leave with that number, and then you can decide what, if anything, to buy.

This is not because Australian leaders are unaware of AI. KPMG Australia's Keeping us up at Night survey, published on 5 January 2026, asked 274 C-suite executives and board members between 23 October and 21 November 2025. 63% named "new technologies, including AI, and the use cases and ethics that arise when implementing" as the top business challenge for 2026, up from fourth place in the 2025 survey. KPMG reported that this top ranking was consistent for both larger corporates and mid-market respondents. The survey is not limited to firms of 50 to 200 staff. The useful point is simpler: many leaders already know this matters. The open question is where to begin.

The National AI Centre's "Get ready for AI" guidance makes a similar first move, in plainer official language. It tells organisations to be clear about the problem they want to address, the outcomes they are aiming for, and how AI could help, and to define what success will look like. It also warns that spreading effort across too many initiatives is a way to lose that focus. A dollar figure on the leak is how I would make that problem concrete.

We already have ChatGPT or Copilot. Is that a start?

It can be a start if you can already say what the tool is fixing, and what that is worth each week. If you cannot, you have a licence, not a starting plan. Plenty of teams already use tools and still cannot tell you what the unfixed work is costing.

I use these tools in my own brokerage. The mistake is treating a login as proof that the business has started. Official planning guidance from the National AI Centre, in "Identify opportunities," puts it neatly: not everything needs AI, and you should match the tool to the task after you understand the process. Their key rule is that if a simpler fix can deliver most of the benefit at much less cost or complexity, you should start there. Templates, a cleaner handover, or cutting a step can be the right first move. AI comes in when the number says it is worth it.

How do I know which process to fix first?

Fix the process that is already costing you the most in time and money, provided the work is clear enough to measure. That is usually a weekly process nobody has priced, not the newest product a vendor wants to show you. You leave the 45-minute diagnostic with that figure.

The National AI Centre's opportunity guidance is useful here. It asks you to look at how often the work happens, how repetitive it is, whether you have the data, how bad an error would be, and what the current cost is. It also asks you to think about impact: would a wrong result harm someone, or be hard to undo? High-impact work, such as hiring decisions, needs more care, and may need the process itself to be fixed before anyone brings in a model.

For a CEO of a firm with 50 to 200 staff, I would still begin with the dollar figure. Score the work after you know what it is worth, not before. Otherwise you can spend a week ranking tasks that do not move the numbers.

Should I write an AI policy before I do anything else?

Write the rules you need, but do not wait for a perfect policy before you put a number on the leak. The National AI Centre's "Get ready for AI" guidance includes oversight and accountability: know who is responsible for decisions about AI use, and which uses need greater scrutiny. That can sit alongside the first measurement. It should not replace it.

If your organisation is covered by the Privacy Act, you already have obligations when staff use AI with personal information. I have not tried to settle those questions in this article. The first practical step remains the same. Know what the unfixed work is costing you. Then design the work, and the rules around it, to match that problem.

What does the 45-minute diagnostic cover, and what can follow?

The 45-minute diagnostic is the first meeting. We work through your numbers with you and, if it helps, whoever runs operations. You leave with a dollar figure on what the leak is costing you. If it's small, we'll say so. We don't talk price until you have that figure.

If you want help after that, the work that can follow is a Fractional Chief AI Officer advisory, which I have already defined in that earlier article. The advisor builds the AI roadmap with you, and every project is tied back to the figure from the diagnostic. Whether that help stays part-time, or whether a full-time hire later makes more sense, is a decision you can make once you have the number. You can read both of those pieces rather than have me repeat them here.

The Fractional Chief AI Officer advisory is the page for that later work.

Put a dollar figure on the problem

Book the 45-minute diagnostic and leave with a dollar figure on the leak, before you spend more on AI.

Author

I'm Johnny Sukkar, founder of Versantly, which is operated by Lycaon Media Pty Limited (ABN 22 671 975 983). I've been in lending since 1990 and in broking since 2008, and I've been building and running AI workflows inside my own brokerage since 2023.