I have seen the future, and it is Meta

AI is starting to profoundly disrupt the labour market and the tech companies currently aren't showing much interest in compensating displaced workers

Our bet is sort of that in the next year, probably, I don’t know, maybe half the development is going to be done by AI as opposed to people, and then that will just kind of increase from there.
Mark Zuckerberg, 30/4/25

We are now seeing a major AI acceleration. I expect 2026 to be a year where this wave accelerates even further on several fronts. We’re starting to see agents really work. This will unlock the ability to build completely new products and transform how we work… As we plan for the future, we will continue to invest very significantly in infrastructure to train leading models and deliver personal superintelligence to billions of people and businesses around the world… we’re investing in AI-native tooling so individuals at Meta can get more done, we’re elevating individual contributors, and flattening teams. We’re starting to see projects that used to require big teams now be accomplished by a single very talented person.
Mark Zuckerberg, 28/1/26

Meta has been in the news this week in my homeland of Australia. Well, all the tech behemoths have, in one way or another, but Meta has been the focus. So, I thought I’d look at three Meta-related news stories and what they portend for the future – a future that I fear is bearing down far quicker than most people realise.

The big news down under, as elsewhere, was Zuck’s restructuring of Meta.

Zuck has been trimming his workforce for some time, but he recently announced 10 per cent of Meta’s workforce is getting the heave-ho this month and 6,000 open roles will be left unfilled. Meta has been coy about further layoffs, but credible sources report more redundancies are planned for the second half of the year and the size of those cuts will depend on “developments in artificial intelligence capabilities”.

Meta is not merely asking its remaining workers to use AI. It is asking them to become training data for it.

Meta is installing software on US-based employees’ computers to capture mouse movements, clicks, keystrokes and occasional screen snapshots for AI training. The tool is called the Model Capability Initiative and is designed to teach Meta’s models how humans actually interact with computers – choosing from dropdown menus, using keyboard shortcuts, clicking through work apps and so on. As one internal memo put it, “all Meta employees can help our models get better simply by doing their daily work.”

The broader program has been renamed the Agent Transformation Accelerator. Meta CTO Andrew Bosworth has said of this program, “The vision we are building towards is one where our agents primarily do the work and our role is to direct, review and help them improve.”

Meta wants AI to do the work, while the remaining human employees become AI supervisors/editors/correction mechanisms.

Of course, as artificial intelligence capabilities develop, there will be ever less need for humans to supervise, edit or correct the work of AI systems.

Along with several other tech companies, Meta is on the cutting edge of replacing expensive human cognitive labour with AI. But what’s happening at innovative tech companies today will be coming to run-of-the-mill, late-adopter businesses tomorrow.

Any job done entirely through a screen, keyboard and mouse (i.e. a ‘KVM job’) is highly vulnerable to automation because AI agents are learning to operate the same interface humans use. Not just answer questions. Not just draft text. Agents can now look at a screen, click buttons, type into fields, move through apps, fill forms, edit spreadsheets, write code, triage emails, update CRMs, generate reports and coordinate workflows.

There’s still debate around exactly when a critical mass of KVM jobs will be automated, but most now agree the process has started to unfold.

In February, Microsoft AI chief Mustafa Suleyman predicted that AI would reach “human-level performance on most, if not all professional tasks” and that many tasks involving “sitting down at a computer” could be automated within 12-18 months, including work in accounting, law, marketing and project management.

Anthropic CEO Dario Amodei has warned that AI could wipe out half of entry-level white-collar jobs within one to five years, especially in technology, finance, law and consulting.

Those forecasts may prove too aggressive. But they are not coming from wild-eyed fringe dwellers. They are coming from people building the systems.

To summarise, AI is going to vaporise a lot of well-paying, mid-to-high status white-collar jobs. Nobody is sure how rapidly that process is going to play out, but those closest to the action predict profound labour market disruption between now and 2030.

Perplexity CEO Aravind Srinivas recently encouraged soon-to-be-made redundant workers to look on the bright side. During an episode of alpha-tech-bro podcast All In, Srinivas said, “The reality is most people don’t enjoy their jobs. There’s suddenly a new possibility, a new opportunity, to go use these tools, learn them, and start your own mini business… Even if there is temporary job displacement to deal with, that sort of glorious future is what we should look forward to… Either the other people who lose jobs end up starting companies themselves and make use of AIs, or they end up learning the AIs and contribute to new companies.”

The glorious future
I have my doubts about Srinivas’s specific claims, but he’s directionally correct about the possibility of a glorious future.

I’m sceptical of the idea that those automated out of a job are going to start their own business or find employment elsewhere. That will happen to some extent, of course, but there are all sorts of obvious issues with Srinivas’s rosy vision.

What kind of businesses are displaced individuals going to start? If AI rapidly automates cognitive labour away, they will have to involve physical work. In some cases, people will be able to retrain as, say, tradesmen and open their own business. But given how quickly robotics is now advancing, I’m not sure blue-collar (or pink-collar) labour is going to be safe from automation for long either.

Then there’s the issue of the market for what your business is selling. If there’s mass unemployment, are many people going to be paying someone else to mow their lawn, clean their house, cook their dinner or unclog their S-bend?

That noted, Srinivas makes a valid point about most people not enjoying their work. Most people don’t have jobs that provide a sense of personal satisfaction and raise their social status. Those people would be delighted to spend less time working and more time doing the things they enjoy. Such a glorious future is now in reach, if the benefits of automation and AI-driven productivity gains are shared equitably.

Worryingly, equitable sharing has been out of fashion for nearly half a century and tech companies have proven especially reluctant to pony up for the common good. Which brings me to the two other stories involving Meta that have broken in recent days.

We’re not all in this together
Wherever you’re based, dear reader, you’ve probably seen this kind of story many times before, so I’ll attempt to keep this brief.

Earlier this week, it was reported that Meta generated $1.74 billion in gross Australian advertising revenue last year but 87 per cent of that was sent offshore, leaving just $223.9 million in net revenue. Meta paid the princely sum of $61.16 million to the Australian Tax Office (ATO) in 2025.

Many ordinary Australian workers face marginal tax rates north of 30 per cent. Meta, by contrast, paid the ATO an amount equivalent to about 3.5 per cent of the gross ad revenue it generated.

Australia’s banks are minnows compared to the American tech behemoths. Yet as the Australian Banking Association recently pointed out, they paid $16 billion in taxes last financial year. In comparison, Alphabet, Apple and Meta collectively contributed $515 million.

Unlike Srinivas, plenty of tech industry heavy hitters have been honest about the fact many people are going to find themselves automated out of a job with no realistic prospect of finding other ways to generate an income.

Some, including Sam Altman, Bill Gates and Vinod Khosla, have argued that as AI (further) shifts economic power from labour to capital, tax systems will have to follow the money. Others, including Zuckerberg, Musk, Amodei and Hassabis, have gestured toward UBI, universal high income or broader redistribution.

Let’s leave aside the historical lesson showing that the losers of globalisation never received much compensation and assume it could be different this time around. It should be noted that tech companies, and their founders, employ armies of well-remunerated tax lawyers, accountants and transfer-pricing specialists to aggressively minimise their tax obligations.

But maybe tech company founders, especially nowadays, are all libertarians who prefer to dispense largesse directly rather than give money to inefficient bureaucracies?

Perhaps people like Zuckerberg quietly do good deeds, like throwing a few shekels to the legacy media organisations whose lunch they so comprehensively cut. After all, a free and independent press is vital to holding power to account, exposing wrongdoing and testing official claims.

A few days ago, the Australian government proposed an arrangement where Google, Meta and TikTok either provide funding to Australian media organisations or face a levy of 2.25 per cent on their Australian revenue.

Having already opted out of a similar earlier scheme, Meta reacted angrily to the latest proposal labelling it “nothing more than a digital services tax” and a “government-mandated transfer of wealth” from one industry to another with “no connection to the value exchanged”.

Things will move quickly from here on in
As I point out with monotonous regularity every Friday morning, AI is going to automate away a lot of jobs, especially the KVM ones. That’s not something that might happen in a few years’ time – it’s a process that is already gathering serious momentum. As Srinivas argues, that doesn’t necessarily have to be a bad thing – many people would be happy to see their job automated away if they no longer had to worry about generating an income.

But we’re only going to get a UBI, or anything like it, if governments make the companies currently capturing AI’s upsides contribute.

Given tech companies have long struggled to be good corporate citizens, it’s optimistic to assume they’re about to mend their ways, unless forced to by voters and their elected representatives.

The future Meta is building may well be technologically astonishing. It may free millions of people from dull and degrading work. But if the businesses automating labour also minimise tax and resist funding public goods, the future looks grim rather than glorious.