What history tells us about jobs in the world of AI

One weird quirk of getting old is that you find you can remember stuff that other people never got to see.

The first job I had after university was organising and marketing conferences for an engineering institute in London. We’re talking 35 years ago. In the posh HQ near St James’s Park, there was a lady who used to ride the lift with a tea trolley in the middle of the afternoon. She’d stop at each floor and word would get around. We’d then scoot over from our nearby offices to grab a cuppa and maybe a scone or bun or something.

About 18 months or two years after my arrival, the tea lady was gone - her role as peripatetic purveyor of beverages deemed somewhat archaic even for the early nineties.

Another lady - who I think was a contractor for a photocopier company - paid regular visits to the offices to change the toner cartridges on the machines. Her name was Tina, so she was inevitably known as Tina Toner.

While the use of the photocopying machines at that time was pretty constant, as all documents were printed rather than distributed electronically, I imagine the need for frequent contractor visits diminished later in the decade. Technology improved, copiers got less heavy use and we started emailing everything. Bye bye, Tina. You were better than all the rest.

There was a third female employee who particularly sticks in my mind. Although we’d moved with the times and were busy faxing information to most prospective conference contributors, delegates and exhibitors, there were some parts of the world - such as China and the then-crumbling USSR - that either didn’t have fax or were relying on dodgy phone networks. So we had someone who operated a telex machine.

This was a technology developed in Nazi Germany and used widely in business after World War II, which was built out of telegraph technology. An operator would take your message and encode it on a special transmitter device. It would be sent down the line to another ‘station’, where the receiving operator would decode the incoming message and convert it into something that everyday people could read. It would then be shuttled as a printed note to the intended recipient.

The telex is particularly relevant our ongoing discussion on this Substack about the impact of AI on jobs. It was a technology that only a handful of people could operate and you needed special training to do it. It was largely superseded by fax, which was simple to use and became familiar to most secretarial and administrative staff. And fax, in turn, was replaced by email, which became something that everyone had.

Democratisation. That’s the key idea here. Technologies may often start out as something that only an elite group of people can operate, but in a relatively short space of time and with some tweaks, they quickly become accessible to all.

Slightly later in the 90s, I was working in a London ad agency.

There was one person in the whole building who had access to the internet.

I know to younger readers, this is such a brain-wobble that you might want to take a second’s pause to get your breath.

Back then, if we wanted to look something up on what passed for the worldwide web at the time, we had to go to Stuart. Only he was privy to its mysteries and gatekeeper to this virtual domain beyond our limited imaginations.

Think about the parallels to today.

You’ll notice how many people on LinkedIn will tell you that only those who understand the secrets of AI will have secure jobs in the future. If you learn how to operate and manipulate the tech, you’ll thrive while others languish on the scrapheap.

Let’s leave aside the fact that much AI will actually be agentic and autonomous, happily doing its thing without any day-to-day involvement from humans. When we are interacting with the technology, it will be entirely intuitive and easy. We won’t need any special qualifications or insights. We’ll simply say what we want the technology to do in everyday language - through a written or spoken prompt.

So let’s think through what this means in terms of jobs.

People who understand AI at a highly technical level will be much in demand for the next few years, as we race towards what’s called Artificial General Intelligence - or the point where the models are broadly better than humans at everything. These are engineers and scientists who understand deep learning or who can contribute to the physical infrastructure (data centres, energy production, chip manufacture etc) that support the AI revolution.

What about people who currently prompt AI, cajole it and have learnt the bot-whispering arts? Well, they start off being paid to do this, as their self-proclaimed skills have an allure and mystique about them. But it will very quickly become something that anyone can do.

Look at Veo 3 and the other highly sophisticated AI video production technology that’s emerged this year. Something that previously involved lots of money and people - and took many days and much effort - can now be accomplished by much smaller teams or even individuals in a fraction of the time and at a hugely reduced cost.

For the moment, the people using the technology still keep their jobs because there is human curatorial and technical skill involved manipulating the AI and refining its ouputs. But the next stage will undoubtedly be technology that you or I can give everyday conversational instructions and it will create complex and lengthy video content out of nothing. At which stage, no one will be paying anyone for their expertise, interpretation or ‘understanding’ of the AI.

The redundancy starts with the people who would have been involved in the production of real-world film and video - actors, sound engineers, camera operators, make-up artists, location scouts and so on.

The people who can do the prompting and direct the AI keep their jobs initially and are able to command decent pay for producing the non-human video content.

Then, as the tech becomes even more intuitive and many more people can use it, the bot-whispering job becomes devalued and the work gets outsourced to people who do it cheaper.

Eventually, it becomes apparent that anyone can do it and its value drops still further.

And the final stage is AI is not only generating the final output, but also making all the initial suggestions and prompts for the next feature film of piece of social content.

Your job won’t be taken by someone who knows how to use AI.

Someone who knows how to use AI will see their job taken by AI.