Where To Start With AI In Your Business: The 20-Minute Audit

The fastest way to find where to start with AI is to spend twenty minutes on one honest exercise. List the recurring tasks that eat your week, note how often each runs and roughly how long, mark which ones are rule-driven instead of judgement calls, then circle the work that’s frequent and time-heavy and low-judgement. The one at the top, highest volume, lowest judgement, most hours, is where you start. Not the whole business. One task. You can do this today, on a single sheet of paper, without buying anything or booking a call. This is the quick self-audit a busy owner can actually run between meetings, and below is exactly how to run it, minute by minute.

The Bottom Line

• You’re not stuck because AI is hard. You’re stuck because nobody gave you a simple way to pick a starting point.
• In twenty minutes you can list your recurring work, score it on frequency and judgement, and find your first candidate.
• The best first build is the most frequent, lowest-judgement, most time-heavy task you do. Rule-driven work that piles up.
• Run the quick audit today, then use the deeper scoring model when you’re ready to build. Start with one, prove it, expand.

Why You’re Stuck On Where To Start

You’re not behind on AI. You’re buried in options. Every week there’s a new tool, a new thread, ten more people saying you’re doing it wrong. So you research instead of acting, and the research becomes another job you don’t have time for. The problem was never motivation. It’s that nobody handed you a simple way to point at one thing and say, start here.

That’s what this exercise fixes. It strips the decision down to a list and two questions, and it ends with a single task circled on a page. No tool comparison, no strategy deck, no boiling the ocean. The reason most owners stall isn’t a lack of AI, it’s a lack of a starting point they trust. We cover the bigger picture in the 2026 AI automation playbook for Australia, but you don’t need it to start. You need twenty minutes and an honest list.

Minutes 1 To 5: List What Eats Your Week

Open a blank page and write down every recurring task that eats your week. Don’t filter, just capture. The jobs you dread, the ones that pile up, the work waiting for you every Monday whether you want it or not. Aim for ten to fifteen lines in five minutes. Speed beats polish here.

Think about the must-do work, not the big strategic stuff. The admin that has to happen or things break. Replying to enquiries. Chasing invoices. Putting quotes together. Building the same report. Data entry between two systems that don’t talk. Scheduling, confirmations, follow-ups. This is the 80 percent that burns your bandwidth and leaves nothing for growth. If a task happens again and again and you’d never call it the reason you started the business, it belongs on the list. Write it down and move to the next one.

Minutes 6 To 10: Score Each On Two Things

Now go down your list and mark two quick things against each task. First, roughly how often it runs and how long it takes. Daily or weekly? Five minutes or an hour? You’re not after precision, just a rough size. Second, and this is the one that matters most, is it rule-driven or does it need real judgement?

Rule-driven means it follows a pattern you could explain to a new hire in a sentence. Same steps, same logic, every time. Sorting an enquiry, drafting a standard reply, chasing an overdue invoice, pulling figures into a report. Judgement work is different. Pricing a tricky one-off, handling a sensitive complaint, deciding who to hire. AI can prepare those, but the call stays with a person, which is the whole point of human-in-the-loop. So against each line write two notes: how often and how long, and rule-driven or judgement. That’s the entire scoring step. Two marks per task.

Minutes 11 To 15: Find The Candidates

Now circle the lines that hit all three. Frequent. Time-consuming. Rule-driven. Those are your automation candidates, and they almost always jump off the page once you mark them. The work that runs constantly, eats real hours, and follows a pattern is exactly the work a system can take off you.

Be honest about what doesn’t qualify. A task that happens twice a year rarely pays back the build, so leave it. Anything that needs real judgement stays with you, draft-and-approve at most. And if a task is frequent but the process behind it is broken or half-remembered, don’t circle it yet. Automating a mess just gives you a faster mess. Fix it by hand first, then it qualifies. What survives is usually three to five tasks, the ones that quietly became a job you can’t quit. That shortlist is the real output of the whole exercise.

Minutes 16 To 20: Pick The One

From your circled shortlist, pick exactly one. The single task with the highest volume, the lowest judgement, and the most hours attached. That’s your starting point. Not your top three. One. This is the layers-not-leaps principle in practice: you start with one task, prove a system can run it on your real data with you reviewing, then expand from there.

For most businesses this size, the winner is something like inbound lead replies or invoice follow-ups, frequent, rule-driven, and sitting close to money. But run it on your own list rather than copying ours, because the point is to find the task that’s costing you the most right now. Once you’ve got your one, you’ve done the hard part: you have a clear, defensible answer to “where do I start.” The deeper version of this, with a full scoring model that adds error cost and revenue proximity, is in what to automate first. That’s where you go when you’re ready to turn one task into a build order.

Frequently Asked Questions

How Is This Different From The Full Scoring Model?

This is the quick version you run today in twenty minutes to find one starting candidate. The full model in what to automate first scores every task on four factors, frequency, hours, error cost, and revenue proximity, and turns your whole list into a ranked build order. Use this first to get moving, then that one to build properly.

What If Everything On My List Feels Important?

That’s normal, and it’s exactly why you score on frequency and judgement instead of importance. Important and automatable aren’t the same thing. The annoying monthly report feels important but runs twelve times a year. A daily enquiry that you answer the same way every time is the quieter winner. Trust the marks on the page, not the feeling.

Can I Really Pick A Starting Point In Twenty Minutes?

Yes, because the value is in the ranking, not clever analysis. Most owners are surprised how fast the same two or three jobs float to the top once they’re written down and scored. The exercise is deliberately simple so you act instead of researching forever. Twenty minutes gets you a defensible first candidate, which is the whole point.

What Happens After I’ve Picked My One Task?

You scope it properly and build it, then prove it on your real data before touching the next. The deeper scoring model helps you turn your shortlist into a full build order, and you can weigh the spend against the payback using what AI automation costs in Australia. If you’d rather not build it yourself, systemising a business with AI covers the done-for-you path.

You don’t need more tools or another course to start with AI. You need twenty minutes, an honest list, and the discipline to pick one task instead of the whole business. Run the audit today, circle your candidates, pick the one that’s costing you the most, and you’ve got a real starting point. If you want that one task scoped against your actual stack and built around how your business runs, owned by you, Get In Touch.