The story is not that AI can write software. The story is that twenty years of accumulated judgement has suddenly become deployable.

It is 11:47pm on a Tuesday. My kids are asleep. I have just pushed a product update into production.

Not a slide deck. Not a concept somebody will assess next quarter. A working product being used by people with a real problem—built from a kitchen table in suburban Melbourne, around a full-time career and three children who have an inconvenient habit of needing a parent.

A few years ago, that sentence would have required a team, capital, months of planning and a forgiving family. Today, one experienced operator with the right tools can take a problem from conversation to production in days.

That is the change I do not think Australia has fully understood yet.

The problems we learned to tolerate

I have spent nearly 25 years in technology, transformation and delivery. Nearly every complex environment has the same uncomfortable feature: the truth exists, but it is scattered.

It lives across plans, risks, budgets, systems, contracts, conversations and the private concerns of the people doing the work. By the time it reaches the person making the decision, it is often late, sanitised or both.

That is why I built Lighthouse.delivery, Propli.ai and MarginShield.io. They serve different markets, but each performs the same underlying job: reveal the truth hidden inside a complex decision and give it to the person who needs to act.

What AI actually changed

For two decades, the cost of turning an idea into production software was itself a moat. You needed capital because you needed engineering capacity. You needed time because every decision became a queue.

That execution barrier has fallen dramatically.

But AI can produce a wrong answer just as quickly as a useful one. It cannot give you the career that tells you whether something is safe, commercially naive, incomplete or beautifully designed around the wrong problem.

The bottleneck has moved from execution alone to judgement: knowing what to build, who it is for, where the risks live, what must remain human and when another feature is simply easier to add than valuable to use.

The operator-founder

The people best positioned to build the next generation of useful Australian AI businesses may already be working inside agriculture, mining services, logistics, care, property, retail, government and financial services.

They are the people who have spent years watching the same failure recur. Until recently, many could identify the problem but could not economically build the solution.

Now they can.

Why the kitchen table matters

I build between 9pm and 1am, on weekends and in the gaps around gymnastics, tennis and whatever else has appeared on the family calendar.

I do not mention that as a productivity flex. Some nights nothing useful happens. Some nights a child wakes up. Some weeks the laptop loses.

The kitchen table matters because it is where the old constraints become visible. If a working parent with domain experience can build something useful without first constructing the machinery of a traditional startup, the pool of possible Australian founders has changed enormously.

Australia should build this time

Australia has become very good at adopting technology built elsewhere. AI gives us another choice.

The opportunity is not to recreate the frontier models. It is to combine those capabilities with the specific judgement Australians already possess: how our property system works, how a regional health service operates, how a retailer manages supplier funding and how a delivery leader recognises a program drifting before the dashboard turns red.

Those are not small opportunities. They are the raw material for a generation of specialised, exportable businesses.

To the person with the problem

If you have spent years thinking “someone should build a tool that does this”, do not begin by asking AI to generate an app.

Begin with the problem. Sit with the people who experience it. Find the evidence. Understand who carries the risk and what must remain human. Then build the smallest thing that proves the judgement.

The tools have arrived. The judgement was already here.

I started. I would love some company.