In the 1990s, people worried that computers would take their jobs.
Today, they worry that AI will take their jobs.
But AI is also creating jobs that did not previously exist.
One of them may be the "AI Verifier" job.
The NSW Government is making AI governance an operational responsibility.
NSW agencies are responsible for governance and assurance of AI across its lifecycle, and the NSW AI Assessment Framework is mandatory for government agencies using AI. NSW guidance also says AI should augment rather than replace human decision-making where its use could create harm.
Writing the policy is the relatively easy part.
Software can also help identify where verification or human intervention is required.
But eventually the workflow reaches a human.
Who is that human?
Someone has to determine whether an AI-generated output can safely become an organizational record, recommendation, decision or action.
That person needs more than permission to click Approve.
The role carries responsibility.
It requires training.
It requires judgement.
And importantly, it creates accountability: who verified this?
That is the AI Verifier Job.
It may not yet commonly exist under that title, but the function is increasingly difficult to avoid.
In 2025, a Deloitte report prepared for an Australian Government department was found to contain nonexistent academic references and an invented court quotation. Deloitte subsequently agreed to partially refund the government.
Generative AI had been used during the work, although Deloitte did not establish that AI itself caused every error.
The important governance question is therefore not simply:
Did AI make the mistake?
It is:
Why wasn't the mistake verified before the output was relied upon?
Checking a report is one thing.What happens when AI can act?
NSW Government guidance (read more...) on AI agents already emphasizes setting clear ownership and guardrails before deployment.
Imagine an AI agent handling a telephone conversation with a bank customer.
It listens to the customer, interprets what was said, checks information and decides whether to progress an application.
But the customer has a strong accent. Speech recognition incorrectly transcribes an important answer. (read more...)
The AI does not know that the transcript is wrong.
The agent acts on it.
A verification problem has just become a decision problem.
And when AI moves from generating information to taking action, the question Who Verified This? becomes much more consequential.AI may eliminate some jobs.
But accountable AI will create others.
AI Verifier may be one of them.