The future of advertising

The future of advertising

16 January 2024

Originally published by Marketing Beat

Big doors swing on little hinges, and we’re at a hinge point in the advertising industry right now. As we look forward to 2024, everyone’s making their predictions, and a lot of money and a lot of jobs depend on which predictions turn out to be right.

It will be the details which make the difference, and isobel’s planning director James Appleby says anyone who claims to know what’s about to happen is lying.

But a few possible paths may be more likely than others, and thinking about them can help plan for whatever outcome eventually arrives.

So let’s begin with scenario one:

horse population

1. People are horses

Here’s a chart showing the number of horses working on farms in the USA since 1850 to today. It tells a simple story, where the horse was the essential engine of the economy, until right around the invention of the internal combustion engine – at which point it was surplus to requirements.

If you needed something heavy moving, you needed a horse. And then you didn’t.

The analogy is equally simple. For the entirety of history, if you needed something created, you needed a creative person. But what if you didn’t need a creative person any more?

This is the doomsday scenario a lot of people are worried about: a very large numbers of jobs being replaced in a very short amount of time.

And they have good reason to be worried. Wizards of the Coast, a billion dollar company owned by Hasbro and a premier employer for fantasy and concept artists, just laid off over 1,000 staff – roughly 20% of the workforce, many of them artist directors, artists and graphic designers – in a move to “Modernize our organisation and get even leaner” according to CEO Chris Cocks.

Observers have been quick to read this as the dead hand of AI coming in to cut costs. But one company’s actions do not act as a weathervane for all creativity.

So, how likely is it that humans are horses, and we’re living through our own internal combustion engine moment? Time will tell, but consider me a sceptic.

The way I see it, a lot would have to go right for this scenario to play out.

Most predictions have been based on the assumption that we will continue to see improvements in Generative AI stack up year-on-year. Between GPT-1 and GPT-3, each new model seemed light years ahead of the previous. But that progress has slowed down.

Behind the scenes, OpenAI has been reportedly having difficulty getting new models off the ground – cancelling projects and delaying announcements – hinting that we may have bumped against a lower ceiling than expected when it comes to the full potential of Large Language Models. They just can’t be relied on to operate without considerable supervision – they’re too ready to hallucinate or be led astray.

In this context, the doomsday prediction begins to look rather optimistic, as well as simplistic.

Engines replaced horses because they were better in almost every way. Generative AI in its current form has advantages, but not many. For lots of people to be replaced in short order, the technology is going to need a second revolutionary leap, and no one yet knows how that could come about.

So let’s consider the next scenario:

2. Generative AI is a damp squib (or a self-driving car)

It would have been too easy to build this section out of references to the metaverse or crypto – two very damp squibs of recent memory which were celebrated as game-changing as recently as last year but I think there is a fairer comparison. A technology which deserved to be taken at least somewhat seriously.

Throw your memory back to a little over a decade ago. It’s the heady days of 2012. The financial crash is still ringing in our ears, smartphones are a pretty big deal, and everyone has begun to obsess over self-driving cars.

Every trucker and taxi driver was about to get on the breadline, and people born after the year 2000 would never need to take a driving test ever again. Uber’s business model was understood to be “Capture the taxi market, then transition to driverless cars”. Speculation abounded that you wouldn’t park your car, you would just set it to Taxi mode and it would make money for you. It was a wonderful time.

In 2016 Elon Musk promised the imminent release of Tesla’s Full Self-Driving mode. Every year since then, he has made the same promise. Even in April last year, it was going to come out “by the end of this year”…

A decade is long enough for goalposts to have shifted considerably. Tesla has indeed released software that it has variously called Full Self Driving (*Beta) and Autopilot. It has also made it very clear that drivers should not expect this software to actually drive the car all of the time (it has been particularly keen to make this case when people have died). The actual product is much more a generic adaptive cruise control than anything which could replace a human driver, and Tesla currently has the highest accident rate of any car brand on US roads. Tesla currently has the highest accident rate of any car brand

This is not to pick on Tesla specifically. No-one has a self-driving car.

In the end, despite the hype, the self-driving car revolution has not changed any industry in a significant way. It still might, one day, but we’re a long way off that yet.

Is it going to be the same for Generative AI? A lot of heat, a lot of expensive start-ups, and a lot of nothing to show for it in the end? At the moment, the parallels are easy enough to make: a revolutionary promise which is dependent on a level of reliability which currently isn’t available or even feasible; hyped coverage of new announcements, limited follow-up covering underwhelming performance.

As well as the general slowdown in innovation, generative AI had a bad end to the last year with mounting evidence and building lawsuits, demonstrating the propensity towards accidental copyright infringement and plagiarism of transformer models like Chat-Gpt and Midjourney.

In this scenario, our workplaces remain as unchanged by Generative AI as our roads are by driverless cars. Ultimately it’s still you with your meat hands pressing buttons, going to meetings, having ideas and making them happen. And every year in Silicon Valley someone will stand on a stage in front of a screen and promise that by the end of next year, everything will be different, again.

As depressing at that paragraph feels, that’s probably what most people working today would like to happen – assuming they like their job and industry the way it is.

But let’s look at what I’d consider the most likely scenario:

3. Generative AI is Microsoft Office

There’s an old myth that if you put a frog in cold water and heat it up slowly, the frog doesn’t notice and will allow itself to be boiled to death. This is not true, but it is a good metaphor.

Many years ago, I started my working life as a secretary. It was a lot of filing, typing, diary management and sorting out mail. You will most likely notice that these are all things you do as part of your everyday work, despite the fact that you are not a secretary. It’s even an odd thing to say secretary these days, it sounds so old-fashioned.

Clerical office work has not disappeared, but clerical workers are projected to decline and continue declining by around a million jobs each year till 2035. Last year the US Bureau of Labor Statistics forecast thirty jobs likely to see the greatest decline in numbers and compensation, and ten of those thirty were clerical, administrative or support staff.

You will often have heard that “while tech may make certain jobs obsolete, it will also create new jobs that we can’t imagine” as the common reply to luddite fears about humans being made obsolete. However, in the case of clerical work, the new jobs created have often been “your old job, plus the stuff a secretary used to do”.

But crucially, this isn’t a change which happened quickly. It happened almost so slowly that no-one noticed. It started with the advent of computers, and really gained pace in the mid to late ’90s. This isn’t the horse getting kicked off the farm, where the glue factory is working overtime; this the fish getting depleted from the sea, a gradual, but constant change in the ecosystem.

At first it will look like nothing at all. The hype around AI will blow over, you’ll hear jokes about it the same way people joked about the Millennium Bug. But there will still be products getting released and improved.

So, it may feel like a damp squib scenario, until eventually it looks like we lived in scenario one all along.

In the spirit of the new year, let’s play this out: The first major change in this scenario will be an acceleration of in-housing. Improving AI tools and propagating familiarity with them steadily lowering the barrier which necessitates specialist production capabilities.

Greater competition drives prices down, leading to agencies closing shop or becoming more specialised, pushing and pulling business towards deploying more in-house creative. Eventually, that same force may well lead to the in-house creative functions dissolving to just be a part of everyone’s everyday life.

And so, at the level of the individual, the person who previously told other people to tell other people what they want, will find it easier to explain what they want by doing a bad version of it and then saying “this but polished please”. A few years after that, they might find there isn’t the budget for anyone to do the polishing for them, so they do it themselves.

In the week before Christmas around London there were many very-well-paid heads of departments filling in expense claims, wondering what happened to the people who used to do this as their job. And though it may not be the best use of their time, it has proven cheaper and more ‘efficient’ to put that work, along with the filing, typing, diary management and sorting of e-mail, all in the same job description as whatever their actual job is.

Life is what happens while you’re making other plans, and history is much the same.

We’ve mapped the upper, lower and middle bound of the likely future of generative AI. I wish I could say with certainty I knew the future – I’d be very rich if I could – but something’s going to happen, that much we can be sure of.

But the money at stake means it’s in almost no one’s interests to accurately represent the state of affairs, so we won’t know for a long time which scenario has actually come to pass. Until then, we live in interesting times.