Young Donald King worked for PwC where his AI team automated about 30 per cent of manual administration tasks for a client. He used AI robots to pretend they were human. He was sacked and later felt guilty he was robbing people of their jobs. It’s a bit like being attracted to the Nazis because you love the arms bands and the outdoor rallies and then realised three years later, it has all gone terribly wrong.
Donald King worked as an associate consultant at the forefront of AI during his time at PwC, one of the biggest consultancy firms in the world, between 2021 and 2024 in the United States.
He says he was part of the firm’s “AI Factory”, which focused on building autonomous agents for companies, he told the ABC 7.30 report.
“But that shifted as I got into the weeds and understood really what we were doing.”
He says his team’s goal was to automate at least 30 per cent of the manual work involved in routine back-office tasks.
He also found himself grappling with a “moral dilemma” because he was training AI systems to perform work traditionally done by people.
“We were working with a large telecom company and we built 45 AI agents that were working together,” Mr King said.
“We made a Microsoft Teams agent that was pretending to be a human to update the team.
“That was a really big gut punch morally.”
Mr King left PwC in 2024 after being laid off.
Software developer and tech entrepreneur Shaon Diwakar has embraced AI.
After working for Meta for six years in Sydney, Mr Diwakar launched his own start-up company called InboxAPI, which is an email for AI agents.
His product allows users to have a standalone inbox for their AI agents – which can autonomously send, read, search, reply to, and forward emails – and keep their personal inbox, which may include sensitive personal information, separate.
Mr Diwakar has no human employees. Every facet of his work, except for what he does, is AI-generated.
“I use Claude to write code. I’m using ChatGPT for marketing. I’m using Gemini to create videos. I use Perplexity to do really deep research or competitor analysis,” he told 7.30
It costs Mr Diwakar “a couple thousand” dollars a month in AI subscriptions to run his company.
“It’s cheaper than hiring a human, which is the sad reality,” he said.
When he started working at Meta in 2018, he says AI was like an unreliable assistant.
“In the early days it was sort of gimmicky. It would hallucinate a lot. It would make a lot of really trivial mistakes that were easy to pick up for even a novice engineer.”
He says in that time AI has improved dramatically.
“In around November last year, Anthropic dropped a new model called Opus 4.5 and it was exceptional. It was exceptionally good at finding mistakes in code that I had written,” he said.
“There was a sense of fascination and curiosity with that shift. But also, holy cow, this may well remove the need for me to write code at all.”
Software engineer Shaon Diwaker worked for Meta for six years and witnessed the rapid rise of AI.
Mr Diwakar has studied and worked in technology and software engineering for more than two decades. He said he feels “grief” when he reflects on how AI’s ability to code is surpassing human ability.
“You spend a really large portion of your life getting really good at it, and all of a sudden you have a computer [that] kind of comes along and can do it better than you can.”
Although AI can do a lot, it still cannot outperform humans in several vocations.
Professor Clinton Free from the University of Sydney mapped Australian job listings and the potential for AI to automate tasks.
He found garbage collectors, carers and gardeners have the least number of tasks that can be automated – meaning they are the least at risk for being replaced by AI.
“As a parent I think about this a lot, what are the types of roles which are more resilient against AI?” Professor Free said.
“I think [it will be] human-centred roles and jobs which involve the real world, physical touch, like some work in trades and construction, physiotherapy, human connectedness jobs.
Mr Diwakar warns that a job wipe-out is on the way and Australians aren’t prepared.
“I don’t think we have the policy or infrastructure in place to support mass redundancies or displacement of jobs. I don’t think our policy makers have really thought through it,” he said.