Albanese asked the right question on AI. Here's the next one.

Artificial intelligence didn’t make me smarter. It made me understandable.

That may turn out to be the technology’s most important economic contribution.

When the prime minister outlined Australia’s AI agenda this week, the conversation centred on productivity, jobs and national capability. Those matter. But I suspect we are measuring only half of it.

Productivity asks how much more can be produced by the people and organisations already participating in the economy. It takes participation as given. For many Australians, participation is precisely what has never been given.

I am an autistic founder. Before generative AI, I could usually see the problem, understand the system and design the solution. The difficulty came between what I understood and what an institution was prepared to receive.

A government agency needed more than a good idea: it needed that idea expressed in correct procurement language. A grant body would only weigh community value once it had been compressed into a preferred outcomes framework. I could do the work. What I could not always do was perform the institutional dialect in which the work had to be described. Not without consultants, employees or exhausting personal effort.

Then generative AI arrived, and something specific changed. It did not give me ideas. It translated them.

Generative AI did not increase my capability. It increased what I have come to call my institutional intelligibility: the degree to which what I could do became visible, legible and actionable to the systems that decide who gets funded, contracted and taken seriously.

For the first time, I could move from my own thinking to a document an institution could act on, at close to the speed of the thinking itself.

My experience is hardly unique. Institutions are translation machines. They can only respond to capability once it has been translated into forms they recognise: tenders, business cases, grant templates, theories of change, the right register of professional confidence. Every entrepreneur therefore pays two costs: the cost of producing value, and the cost of translating that value into something institutions can process. We have come to treat the second cost as a natural part of the first. It is not.

The cost of translation falls heaviest on the people furthest from professional networks: migrants, regional founders, disabled people, sole traders, community organisations, experienced professionals starting again after redundancy or years of caring. Their problem is routinely misdiagnosed as a lack of capability. More often, they can solve the problem. They just cannot afford the translator.

AI does not create capital, customers, relationships or judgement. It can fabricate information, flatter weak arguments and produce polished nonsense, and it ties Australian founders to a handful of foreign platforms. Those are real risks. But none of them change the fact that, for many founders, AI has become something else entirely: affordable organisational infrastructure.

That should change the questions government asks. Not just: how much time can AI save a worker? But: who can now do work that institutional barriers previously prevented? Not just: how does AI help established businesses grow? But: can a sole founder or small social enterprise now operate with capabilities that once required an entire professional team?

Procurement is perhaps the clearest example. AI now allows one capable founder to perform analysis, drafting and strategic planning that once required a team. Yet many government tenders still reward organisational size, years of trading and layers of professional presentation. If a sole founder can produce the same work as a much larger organisation but is filtered out before anyone evaluates the substance, then the barrier was never capability. It was institutional design. AI has changed what a capable organisation can look like. Our procurement systems largely have not.

A new divide is forming, too. It is not between those who have AI and those who do not, but between people who can use it to convert capability into economic agency, and people limited to consumer tools that extract their data while they are told AI literacy is their personal responsibility.

The prime minister said the real issues raised by AI are economic, legal, social and moral rather than merely technical. He is right, and perhaps more right than the productivity framing allows.

For decades we have assumed the people missing from our economy lacked capability. AI raises a more uncomfortable possibility. Perhaps many of them simply lacked an affordable translator.

Australia’s AI debate isn’t really about technology. It is about whether our institutions have been measuring capability at all, or merely fluency in their own language.

First published on Substack, 15 July 2026.