Markets tend to discover the second problem only after solving the first.

Right now, Australia is still solving the first AI problem: getting to production. Silicon Valley is already dealing with the second one.

I spent time at HumanX in San Francisco in April, then AWS Summit Sydney last week. What stayed with me was not the difference in technology, ambition or even adoption speed. It was the difference in where each market sat in the lifecycle of the conversation.

AWS Sydney was focused on operationalisation. The message running through the keynotes, technical sessions and hallway conversations was consistent: move beyond pilots, get workloads live, stop treating AI as experimentation. For the Australian market, that is exactly the right push. Most Australian organisations are still trying to bridge the gap between interest and implementation.

HumanX felt further down the curve.

The companies there were already deep into production. Agents, orchestration layers, evaluation tooling and operational infrastructure dominated the discussion. The assumption underneath almost every conversation was that AI was no longer a future capability. It was becoming part of the operating model. Some were even already turning it off because they weren't able to quantify the value. Sounds familiar?

But the most interesting conversations were happening around the edges of the conference. AI was starting to be treated like any other operational investment, and the questions on the floor reflected that shift.

Which workloads are actually creating value?

Which ones deserve to scale?

Which ones are quietly compounding cost faster than outcome?

What happens when the board stops viewing AI as innovation spend and starts viewing it as operational spend?

That is the second problem.

The first board meeting approves experimentation. The second asks for accountability. Mature investment always reaches that moment, regardless of how well the technology is working.

That was what connected both conferences for me. AWS is correctly pushing the market towards production. But production is not the finish line. It is the point where AI stops being primarily a technology conversation and starts becoming a business one.

The companies that build measurement and operational discipline alongside adoption will move through that transition smoothly. The ones retrofitting answers later will not.