Selling shovels in an AI-disrupted economy

Thoughts on surfing – rather than getting held under by – the mass-automation wave

The cool kids – Meta, Microsoft and Nvidia – are up by an average of 22 per cent. Amazon and Alphabet shares have basically tracked sideways, while Apple and Tesla are both down 12 per cent…

Where Nvidia, Microsoft and Meta have clear strategies to monetise the artificial intelligence revolution, investors are doubtful about the path being taken by their rivals. Take Tesla, whose founder and chief executive Elon Musk told us once again is an AI company whose future is pinned to the robotaxi model of self-driving cars.

“Autonomy is the story,” Musk said on Wednesday night’s earnings call. “We need the physical product, without which you cannot have autonomy. But once you have a physical product, the autonomy is what amplifies the value to stratospheric levels.”
AFR, 24/7/25

AI is garbage. AI doesn’t work. AI always hallucinates.
It’s just repeating training data. It’s just predicting tokens. It’s not reasoning.
Fine.
Please show me your math olympiad problem-solving results.*
There are no realistic policy solutions or credible political coalitions that can slow down this AI super-trend…
The change management required isn’t just for us in our use of AI.
It’s a broader societal change management program that is presently being ignored or dismissed as science fiction.
Brennan McDonald, 23/7/25

If you’ve spent time in business circles, you’ll be familiar with the ‘sell picks and shovels’ adage. If you’re not the entrepreneurial type, here’s the tl;dr. Two types of people make money during gold rushes. (The Californian gold rush of the mid-1800s is the template here, but the dynamic scales across time, place and industry.)

The two demographics that make out like bandits on gold-mining sites are miners who strike gold and those who sell miners food, shelter, alcohol, sex and, yes, work equipment.

You typically don’t become fabulously wealthy overnight selling picks and shovels. But you do ‘get rich slow’ by selling goods and services, usually at an exorbitant mark-up, to those pursuing more of a high-risk-high-return strategy.

Roughly 175 years ago, a Bavarian immigrant arrived in San Francisco with a plan to sell tents and wagon covers to miners. He soon realised that there was far more market demand for sturdy pants suited to hard, dirty physical labour, so he created what are now called blue jeans. That migrant’s name was Levi Strauss.

About 120 years later, a couple of young nerds, Paul Allen and Bill Gates, became unimaginably wealthy by spotting another gap in the market hiding in plain sight.

Strange as it may seem in these SaaS-saturated times, tech industry types used to be obsessed with hardware. But while everyone else, including Steve Jobs, was obsessed with building more sophisticated computers, Allen and Gates’ world-changing insight was that computers required software and that they could make good coin selling software to the hardware manufacturers.

What we know so far
Once upon a time, we were all hunters and gatherers. Then almost everybody worked the land. Indeed, nearly half the population were still employed in agriculture at the dawn of the 20^th century, even in technologically advanced nations such as the US. And if you weren’t an agricultural labourer, you were probably working in a factory or as domestic help. There was a small class of individuals – doctors, lawyers, teachers and preachers, for example – who weren’t grinding away in a factory, farm or mansion. However, up until at least World War II, they were very much the exception rather than the rule.

We are about to experience an accelerated version of the agricultural society/industrial society/post-industrial society transition that many nations have traversed over the last 100 years.

There are two reasons to be concerned about this transition.

First, it’s likely to play out in years, or possibly even months, rather than decades.

Second, while previous technological tipping points generated creative destruction that saw old jobs being replaced with new and usually better jobs, we appear to be headed into a phase of straightforward job destruction.

This isn’t a prediction about what might happen in the future, it’s something happening – albeit with more of a whimper than an epoch-announcing bang – right now. There are a growing number of dark factories in China. They’re described as ‘dark’ not because anything sinister is going on, but because the lights aren’t turned on. They aren’t turned on because no humans work in these entirely automated factories.

It might not be too long before we have dark, or at least darkish, start-ups as well as dark factories. There has been a lot of excitement – especially among current and aspiring billionaires – about the idea of a ‘one-person, billion-dollar company’. To summarise, if you’ve got a clever business idea, you don’t have to hire anybody else to help you execute on it. You can utilise AI tools to automate coding, create an online presence, develop marketing plans and materials, make sales calls, handle customer payments and complaints, and more.

My digital research assistant informs that while nobody has yet created a billion-dollar business by themselves, there are a small but growing number of “AI-native solopreneurs with zero employees and six-to-seven-figure revenue.”

If you can’t be Levi Strauss, at least open a jeans shop
I don’t want to crush the dreams of anybody hoping to strike gold. But unless you already have substantial equity in a tech company, are a genius software engineer, or have the skill set to launch an (almost) fully automated start-up, you’re unlikely to get filthy rich overnight from AI.

That leaves the contemporary equivalent of selling picks and shovels.

To be clear, I’m not suggesting that selling picks and shovels is a viable long-run solution. To paraphrase Keynes, I fear that in the long run all our careers are dead. But if you’re trying to work out how to turn a dollar in the short to medium term, the following information may be of some use.

There’s no point applying for jobs that no longer exist
This seems self-evident, but having been around for the decline of the legacy media, I can attest that people often behave irrationally when their industry (and hence life plan) starts abruptly collapsing around them.

It now appears undeniable that much of the grunt work previously done by junior employees has been automated away. Nonetheless, it seems today’s grads are still taking much the same approach to securing an income stream that their predecessors did. That is, applying for grad roles in the belief they will eventually land one. Of course, many graduates are still managing to land roles. But it’s turning into a game of musical chairs, with a growing number of grads failing to seal the career-commencing deal.

The conventional advice for those in this situation is to apply for jobs even harder. But if entry-level roles are contracting while the number of grads entering the labour market is staying roughly the same, that suggests waiting around for a business to hand you a job may not be a winning strategy.

As anybody who has done it will tell you, the reality of self-employment is somewhat different to the wage-slave fantasy of it. It’s also sensible to spend some time learning from more experienced colleagues before hanging out your shingle. (Even Jeff Bezos paid his dues for a decade, working for the likes of Intel, Bell Labs and what’s now Accenture in his twenties before launching an online bookstore in 1994.)

But interesting times call for interesting measures. If they can’t get a business to hire them, younger people would do well to consider launching their own business.

There’s no point entering, or remaining in, highly automatable industries
AI excels at digital, repeatable and rule-based tasks. That’s bad news for those working in the following industries: admin and clerical services; customer service and support; media, content and design production; and transportation and logistics.

As is often observed, frequently with a side serving of schadenfreude, AI will soon be for white-collar and pink-collar workers what automation and offshoring have been for blue-collar ones. In short, if your job primarily involves following rules – and doesn’t require you to demonstrate taste, creativity, warmth or judgement – you’ve now got a target on your back.

That’s the bad news. The good news is that while AI can already perform many tasks far more accurately, comprehensively, quickly, and cheaply than humans, there are still areas where it struggles. These will be the growth, or at least resilient, industries over the next decade.

Skilled trades: One day, the robo-tradesmen will start rolling off assembly lines and the human variety will find themselves in the same position as many knowledge workers are currently in. However, nobody expects that day to dawn for at least five years, and probably not for ten. (When Geoffrey ‘Godfather of AI’ Hinton is asked for career advice, he always recommends training, or retraining, as a plumber.)

Even back in the pre-ChatGPT Beforetimes this was solid advice. Successful plumbers have long out-earnt all but the most high-flying professors. In a world where the cost of cognitive labour is approaching zero but lots of data centres and – dare to dream, fellow non-Boomer YIMBYs! – housing needs to be built ASAP getting a trade would seem to be a smart move.

Healthcare/caregiving: The outlook is slightly less bullish here, especially if it becomes feasible to affordably mass-produce humanoid caregivers.

Such caregivers – think those aged-care robots the Japanese are so keen on – already exist but are currently eye-wateringly expensive and not very capable.

Nonetheless, human caregivers, like human tradies, probably have 5-10 years before they need to even begin worrying about the droids cutting their grass. And while nobody is likely to care who unplugs their S-bend, they will presumably continue to prefer having a human look after their newborn baby, injured teenager and frail grandparent. So, our human healthcare heroes may end up having a significantly longer shelf life than their tradie counterparts.

Any industry where creativity, original thinking or judgement is essential: Once again, I’m somewhat tentative about making any long-term forecasts. I’m old enough to remember when Westerners would confidently proclaim that Asians lacked creativity and would only ever be able to fine tune technologies created by (more inventive) Caucasians. Whenever I see a pundit insisting that AI is great at remixing but incapable of originating, I remember all the times I was told the same thing about Japan in the 1980s and about China until five minutes ago.

However, while I wouldn’t rule out AI being able to develop good taste and display genuine creativity at some point, I suspect humans will maintain a competitive edge for a few years.**

The one shard of hope this content creator is desperately clinging onto is that while AI content-creation tools are near miraculous, they still seem to be most effectively wielded by those with a deep and intuitive understanding of the content being generated.

For instance, anybody can now use freemium or low-cost AI tools like HeyGen, Sora, Midjourney and Pictory to create ‘good enough’ video content. But if you want finely crafted, attention-grabbing video content that will stand out in a sea of AI slop, it’s still worth investing in a skilled videographer.

As my digital researcher points out, “AI hasn’t yet replaced the entire skillset of a professional videographer, but it’s automated, commodified, or undercut many key parts of the job.” Nonetheless, humans are still competitive when it comes to “high-end cinematography and location shoots” (i.e. live corporate events, weddings and docos). Carbon-based lifeforms also retain an advantage in situations where executive decisions need to be made. To quote Chatty one last time: “AI can churn out content, but it lacks cultural nuance, original storytelling instincts and on-the-fly adaptability. Clients still pay for vision, taste and trusted direction, especially in premium markets.”

That would seem to sum it up for all humans, whether or not they churn out video content for a crust. If you want to earn an income in a post-ChatGPT world, work out what AI can’t yet do well and do that. Ideally for deep-pocketed clients or employers.

* See here.

** Though the success of The Velvet Sundown, an AI-generated band, makes me wonder just how wide human’s competitive moat will prove to be.