What do doctors, bookkeepers and media practitioners all have in common? There’s a great chance that automation and machine learning will replace the humans in these jobs in the not-too-distant future. And while this may seem alarmist, the reality is that quite a lot of jobs as we know them today will see humans replaced by technology. A 2020 report by the Committee for Economic Development of Australia (CEDA) and The Australian Computer Society outlined how 40 to 50 per cent of local jobs may disappear completely by 2035 due to advances in technology. Whether a result of full automation or augmentation by robots and artificial intelligence, one in five workers today have roles which could be completely replaced within the next 15 years. Studies from the US report automation technology may eliminate up to 73 million jobs by 2030 - which equates to a staggering 46 per cent of the current workforce.
While there are still limitations and setbacks, automation and machine learning are revolutionising practically every industry. Just over a century ago most of the world’s population worked in farming but today a tiny percentage of farmers can produce greater yields. Technology not only replaces human labour, but it undertakes tasks at scale, to a consistent and uniform standard. There is a very good reason why Jeff Bezos wants to replace people with drones in his warehouses, and while he may claim it’s to liberate people from the monotony of low-value tasks, the reality is robots never take days off and never need to leave early to pick their kids up from swimming class.
“One in five workers today have roles which could be completely replaced within the next 15 years. Studies from the US report automation technology may eliminate up to 73 million jobs by 2030″
The rise of digital within the media industry has seen an incredible amount of automation in the last 15 years. In its infancy, digital campaigns were booked just like one would book a print ad in a newspaper, the buyer would call the sales team, negotiate a rate, book the space, and send the copy. AT&T are widely credited as running the first digital display ad in 1994; their 468x60 banner ad, a true classic in the early days of digital media, generated an incredible 44 per cent click through rate thus proving the power of novelty. As spends in digital increased, the volume of work required to traffic and execute campaigns grew accordingly and click-through rates fell seismically.
The first stage of the automation revolution, automated bidding, happened around 1997. Agencies now had the ability to set budgets and bid-ceilings, the maximum price you are prepared to pay for an ad, for digital campaigns. This removed the requirement for humans to horse trade advertising space. Fast forward to around 2010 and we see the second stage of the automation revolution - the rise of so-called programmatic advertising.
Programmatic ad buying is the use of technology to buy and place digital advertising, again removing the human input of calling a salesperson, checking availability, negotiating price, and sending ad copy, also referred to as trafficking. Programmatic advertising also meant that the same ad space could be used to serve a different ad to different people at the same time.
In the newspaper analogy, everyone saw the same ad, but in a digital world the headline banner ad at the top of The Irish Times homepage could now be carved up and that same space sold to different advertisers, so that one ad could target men while another ad targeted women with a completely different message, and a completely different advertiser, all at the same time.
Digital data reporting
But newspaper advertising does not generate direct clicks. There is no consistent way to measure, report or analyse the performance of the ad. With digital media, advertising generates clicks, our interaction leaves a digital footprint and creates data. Creation of data leads to an increased requirement for reporting and analytics, which makes digital media highly labour-intensive. The by-product of all this data creation meant a crippling amount of reporting and analytics work for agencies. Available data is an itch that must be scratched but just because something can be measured doesn’t mean it should be. An unintended consequence saw agencies, and ultimately clients, burning hours on low-value work as most people spent their time doing as opposed to thinking. And so, the third stage in automation was borne, the automation of reporting, that is real- time dashboards and data visualisation, which became popular and prevalent around 2018.
We are now entering the fourth phase of automation, the automation of creative, or dynamic creative delivery. Like most things, early iterations of dynamic advertising were, at best, poor, but it has moved on. The data sitting behind campaigns, coupled with data generated from user behaviours, mean that you can now see a very nicely designed digital ad, personally tailored to things you are interested in.
The concept is simple. Create a suite of copy lines, prices and supply a bank of images, and the result is a near infinite number of ad units, constrained only by the number of available permutations and combinations of copy, images and prices that were pre-determined for said ad. Now with just five copy lines, five different images and five price points you now have 125 different combinations of an advert.
Artificial Intelligence (AI) pieces the component parts together in real-time based on the targeting parameters the agency sets. If you take Ryanair as an example, that’s the reason why you see, in my case, an image of Barcelona with copy that reads ‘seats are selling fast’ and a price point relative to the dates I was hoping to travel.
Optimising in real time
But it gets even more sophisticated. AI can use decision engines to increase bids for routes that have more capacity and reduce bidding on routes that are almost full. Similarly, when you know that one person is interested in football while another is interested in opera, technology can be used to further personalise messaging in the form of imagery used it the ad.
And everything can be optimised in real time. AI and machine learning can use eye tracking technology, coupled with media campaign data such as clicks and shares, to understand what versions of the creative are performing best.
Creative thinking abides
The question is how far we can take automation and machine learning in creative thinking. Can a machine really be capable of original thought? AI is already actively producing some of the articles you read, art, books, music and even a television commercial or two, but I am sceptical.
While a machine can iterate and reiterate to solve the problem of making a diesel engine cleaner, it would never suggest an electric engine as a solution to the problem due to the parameters set by humans. It is confined by all that has gone before.
Automation is making a real difference to our business, in a positive way, and if it strips out low-value tasks from the day job then that’s a good thing”
A machine would never have come up with Alfred Hitchcock’s Psycho because up to that point the monsters in popular culture were fantastical beings, werewolves, ghouls, and demons. They lived in the shadows. There was simply no precedent for the monster being the normal-looking guy next door, hiding in plain sight. It was unimaginable, truly original.
Automation is making a real difference to our business, in a positive way, and if it strips out low-value tasks from the day job then that’s a good thing. The industry can get back to spending more time thinking and less time doing. I remain hopeful that, for now at least, we will still need humans to interpret learnings and add a sparkle of creativity and original thinking.