Last year, AOL laid off about 100 salespeople because their jobs were made redundant by programmatic advertising. Such automated selling can obviate the need for human salespeople.
If others make the same value equation as AOL, more job losses may be in the offing. Magna Global, for instance, predicts that by 2017, 80% of advertising will be bought programmatically. Even that remaining bastion of three-martini lunches and late-night schmoozing, TV ad sales, is beginning to be eaten up by programmatic.
While proponents argue that programmatic won’t steal jobs but will rather free up human employees from the grunt work involved in manual insertion orders, the rise of programmatic on the sales side highlights the possibility that some advertising jobs may become automated. Even creative.
That’s right, artificial intelligence is accelerating at such a rate that some argue within 15 years or so, traditional creative advertising functions like dreaming up taglines and ad campaigns may be done by computers. While that’s still a ways off, already some creative is already being taken over by the bots.
The rise of the machines
By now we’ve all heard the claims that the 2020s will usher in the era of self-driving cars. Uber CEO Travis Kalanick has expressed interest in having a fleet of such cars in the next few years, potentially putting thousands of human drivers out of work. Those Uber drivers aren’t alone. An Oxford University study estimates that 47% of jobs could be taken over by computers in "perhaps a decade or two." Even IT jobs are succumbing to automation as artificial intelligence and other innovations are prompting self-regulating machines that don’t require human intervention.
At first blush, creating ad campaigns doesn’t seem ripe for similar cooptation, since it is neither manual labor nor work that relies on brute processing power and specialized knowledge. Instead, creative work, it is assumed, requires a human touch. Right?
The short answer is yes, at least for now. However, something called "programmatic creative" may be the start of the automation of creative work. Several firms, including PaperG, Celtra and Spongecell offer programmatic creative, which mixes and matches different creative concepts to create personalized ads. For instance, a programmatic creative agency might work from 10 taglines and then discover that one works well with one group and another works better with another.
"It’s true that digital media has caused marketing to undergo a paradigm shift from big ideas to long ideas," says Rob Lennon, senior product manager at PaperG. "Instead of defining their brand with one massive campaign, companies are forming their relationships with customers across so many smaller touch points over time."
But why stop there? If a creative agency can come up with 10 taglines, why not use an algorithm to create 1,000 and use programmatic creative to find the best one? "A thousand variations for any creative is completely feasible," Lennon says. "You can easily optimize those down to the best performers, and do it by audience segment, as there is likely going to be some subtle differences.
Lennon says PaperG ran a campaign with a "top five automaker" that used 250 total ad variations. The best campaigns performed seven times better than the worst ones. "Now, who is to say which campaign they would have gone with if there had only been one variation? It could have worked out very well, or terribly," Lennon says. "But this way, they were able to optimize toward the absolute best combinations of audience and creative and shift budget to bet big on what was working once it was proven out."
Ben Kartzman, CEO of Spongecell, says that agencies tell him that they no longer have to manually create separate campaigns. "Even if they were creating five or 10 or 20 [executions], they don’t have to even do that anymore," he says. "They literally just create the one template and technology takes care of the rest." For example, for a recent campaign for Absolut, Spongecell created 12 banner ads featuring different cocktails. The company then targeted each execution at different users depending on their personal tastes and previous engagements online.
Though most people associate programmatic with digital media, this summer M&C Saatchi erected a billboard in London for a fake coffee brand called Bahio that used Microsoft’s Kinect technology to read the emotions of passers-by. The agency tested thousands of iterations of ads. Machine learning helped discover the most effective. Dave Cox, the chief innovation officer at M&C Saatchi, told Campaign that he foresees a day when "artificial creativity" will supplant the traditional version.
The human advantage
There’s some debate about whether machines will ultimately be able to churn out truly creative work. Though they can help refine existing ideas, as we’ve seen, coming up with something from whole cloth isn’t possible — yet.
Andrew McAfee, an MIT professor and co-author of the book "The Second Machine Age," says that he doesn’t think creative work will ever be replaced by machines. However, he sees a future in which "Don Draper dreams up an idea and then we use technology and algorithms to find out if it’s any good or not." Such testing can also determine if one person is more persuadable by one method than another. "The eureka moment doesn’t go away at all, but I think instead of that being the end of the journey and we ramp up the campaign that we get a lot more intelligent about stress-testing it and modifying it."
Vivek Wadhwa, a Duke University professor who studies occupational automation, has a different view. "People think that AI will come in and take every human there is. Eventually, it will, but what we have now is brute force," he says. "It’s very narrowly focused and takes a given problem and solves it with brute force."
For the moment, Wadhwa says we’re "nowhere near" the point where computers can be creative the way humans are. That’s because human creativity is very broad, Wadhwa says. "It’s not deep, it’s broad," he says. "Right now, AI can’t go broad."
By "broad," Wadhwa means the ability to make associations and apply a human touch. For example, think how Mad Men’s Don Draper was able to compare Kodak’s Carousel viewer to a time machine that elicits the delicate and somewhat painful emotion of nostalgia. (It’s also hard to envision a machine coming up with something as random and weird as this Skittles ad that features a man regretting opening a pawn shop that trades unwanted holiday items for Skittles.)
Nevertheless, Wadhwa doesn’t see that human advantage lasting very long. As he sees it, we have about 15 or 20 years left until the machines overtake even this ability.
So what happens then? "No one really knows," Wadhwa says. "People speculate about it, but no one really knows."