Will machines take over the advertising industry?
In November 2018 Lexus came out with an advertisement for the new Lexus car range. This ad was created based on award-winning advertisements. They knew exactly which shots to include in the video to create an astonishing commercial. Yet, when seeing the video at first you might be confused by the storyline. In short, the story is about a man who created a car that is going through a car crash test. The creator nervously watches the test with his daughter on the news and shows relief when the car is able to stop itself before crashing into a truck (Hammett, 2018). AI is integrated into the car which is why the slogan at the end follows: ‘The new Lexus ES – driven by intuition’. So how and why are these shots considered similar to the best advertisements of the industry? You already read about how AI is being integrated into cars, but Lexus also wisely used it to their advantages in their advertising. If AI is so advanced it knows when and how to stop a car, can’t it also make an advertisement?
And that’s exactly what the agency, The&Partnership, hired by Lexus thought and so arose the first AI-scripted advertisement (Faull, 2018). It is humanly impossible to know exactly what makes a certain advert successful. But an AI is particularly good at this. The AI can dictate the similarities in successful ads and provide this information to brands. This will increase the chance for a brand to create a successful ad and thus the AI drives business value.
Where IBM’s Watson was relatively futile in healthcare it is much more easily applied to advertising. With no life-or-death risks and privacy concerns blocking the way, Lexus and The&Partnership ‘fed’ Watson 15 years’ worth of award-winning advertisements in the industry and extra data related to intuition and how people make decisions1. Something that is humanly impossible to analyze, it would simply take too much time. But not for the AI. The AI provided a script for an advertisement video based on the ‘fed’ content. The production of the ad was directed by Oscar-winning director, Kevin Macdonald. For the creation of the ad, it was necessary for human and machine to collaborate (Clymo, 2018). For example, a human intervention was made at the end of the script. A kissing sound was proposed at the end of the video when father and daughter were embracing each other. By humans this is considered as uncomfortable.
It is hard for a machine to understand what is ethical and what is not. Even though the AI had the right intention with the sound, it could provoke controversy. People accept AI-created advertisements but perceive advertisements created by AI with an emotional intent negatively (Bakpayev, Baek, van Esch, & Yoon, 2020). Thus, Lexus took quite a big risk letting an AI create an advertisement without having a clue what the result would be. Luckily for Lexus the outcome had lots of potential. Since it was only the script that was created it was easily adjustable. What if the video itself was created by AI?
Legend that thing
‘Legend that thing’ was a Nike commercial video created with AI. However, it was not Nike that made the video but copywriter Jean-Baptiste Le Divelec. He trained an AI by feeding it 7 years’ worth of Nike’s video advertising content to create an advertising video for Nike (Divelec, 2019). The outcome included some weirdly structured sentences, but it did sound somewhat inspiring.
Jean-Baptiste however is in no way connected to Nike. He had the possibility to create Nike-styled content without them knowing or being able to intervene. It can become hard to moderate what content is real and what is fake if everyone can make content for any brand using AI. Certain content that is not even created or approved by a certain brand will get out and people might associate the wrong advertising with the brand.
As proven in the Lexus case, AI may produce content that can be wrongfully perceived by the public like the kissing sound. If anyone can create this content it will become easier to fool around with a brands’ reputation. Maybe competitors can sabotage each other. Are new laws needed to ensure a brands advertising style? But fake ads are not the only threat AI brings to the advertising industry.
Next to this external threat of the possibility of creating fake branded content, there are also internal risks. AI advertising is created by ‘feeding’ an artificial intelligence the necessary ingredients for composing successful ads. It establishes a result by comparing the similarities within the content of the fed ingredients. The AI might draw the wrong conclusions. I’ll elaborate this in an example.
A successful ad is the ‘Dumb Ways To Die’ campaign by Metro Trains in Australia. The video ad animates a bunch of dumb ways to die but in a cheerful manner with bright colors and happy music. In this ad it is very obvious that the happy format of the video is sarcastic. More of these sarcastic ads but less obviously sarcastic can cause the AI to use aspects for the creation of ads that are unethical. It might create a campaign where someone dies, and everyone is very cheerful and happy about it. As long as the AI does not automatically publish the content, these ethical issues can be resolved by human correction. That would mean that the job roles of content creator would shift to content monitor. Or in the cases of Lexus the content creator will keep existing, but they won’t have to invent the content idea, only execute it.
There is also another possible change in the work field of advertising. Internal marketing teams know all the ins and outs of a brand to create the right ads. Whereas external marketers don’t. A company that uses AI for the creation of ads won’t need as many internal marketing managers but can also use more external marketers because the AI tells them exactly what to do. Whether this is a positive or a negative change is probably personal. But there is also the possibility AI will eliminate the need for many marketers in general, which seems to me and probably a lot of marketers, like a negative.
A company might not need as many marketing employees anymore as it used to. But there is still the necessity to check whether the AI is doing things right. Because like the Lexus and Nike case, the content AI produces will not always be perfect. But it does get close to perfect once it is capable of adjusting and quickly spotting trends. A big trend in advertising is the shift from traditional to digital marketing (Statista, 2020). If AI were to make advertisements and use the data that traditional marketing is dying out, it might boost the killing of this marketing technique to go even faster. In that sense, AI can decide what is trendy and what is not. Because once AI determines that digital marketing is the trend, the AI will create content that is based on how this trend arose. When it starts to make these decisions, it will only create advertisements that have been proven to work. But doesn’t that constrain our own decision-making process?
AI will only create advertisements by learning from a human’s original dataset. On the contrary, creatives invent advertisements as new ideas. Giving the public the possibility to decide if they like it or not and this allows new trends to arise. AI is limited to the creativity of already existing advertisements. If humans don’t make advertisements anymore, there might not be any new trends. The public might get bored of AI advertisements. So, AI can create advertisements, but it will probably not kill the job of marketers and content creators. Humans will have to continue creating content and test what works and what doesn’t so an AI can determine why it worked or why it didn’t. This is a collaboration of AI and creatives with lots of potential.
Let’s take the example of the chocolate brand, Twix’s ‘Left or Right’ campaign. Twix come with two chocolate bars in one package. Around this fact, Twix launched a long-lasting marketing campaign pitching the left Twix against the right Twix (Lewis, 2017). The advertisement is very unique, and the idea is hardly applicable to any other brand but Twix. Before this advertisement, there was barely any other ad similar to this one on which AI could have based this content. But now that the campaign exists for a long time and Twix created more content, it would be useful if AI could improve this campaign. It can analyze what aspects of the Twix ads are also used in other successful campaigns.
After the creation, the ad has to be advertised to the right people at the right moment. An area that AI is already dominating. AI is often integrated throughout the whole consumer journey from need, to purchase and even post-purchase (Kietzmann, Paschen, & Treen, 2018).
Whether you interact with Google and Facebook or not they will always track you and create an online consumer profile. This is done with pixel technology. Almost all websites include codes that are connected to, for example, Facebook (Venkatadri, 2018). Once you visit this website, Facebook puts you into a certain audience (Venkatadri, 2018).
All the audiences and data points of you are combined and form an online consumer profile. Facebook uses the online data points to optimize their ability to provide businesses with the right target audiences for ads (Facebook, 2020). Making this a very valuable AI technique for the advertising industry. In fact, according to a survey by MemSQL, already 61 percent of the marketers say AI is the most important aspect of their data strategy (Memsql, 2018).
Everyday millions of data points are gathered about almost every activity online. And the people tracked? They are completely unaware. I suppose you did not read the cookies statement either. But every time you give cookies permission an AI is fed all of the information about what you do on that website. Or maybe you did not have to sign any cookies at all.
Amazon is a great example of a company that breaths your data. Amazon Alexa is a virtual assistant and Amazon does not openly say that your data is saved, they try to avoid this. instead of openly saying they record you, they say: “No audio is stored or sent to the cloud unless the device detects the wake word (or Alexa is activated by pressing a button)” (Amazon, sd). In other words, she does store your data when she is activated. Besides they don’t express what they use the data that is stored for. The fact that they try to sugarcoat the fact that they record you, makes it look like they know its unethical. It seems as if they rather have customers not know they record them. But Amazon uses AI in many more forms. They use AI for their personalized recommendation system, an anticipatory shipping model and they even own a security company, Ring, that enables Amazon to store video footage of security cameras (WILLS, 2020). You can only imagine what AI must know about you.
If those two AI techniques were to be combined it would create an AI that can create the best advertisement and distribute it to the right people. What will there be left to do for marketers. AI can create the best content based on award-winning advertisements. How can a human compete with that? There is a case where these two techniques were combined. M&C Saatchi, advertising agency, created the world’s first artificially intelligent advertisement (digital) poster (Still, 2015). They have created a poster that changes its content based on the attention it gets. It detects the attention through a camera integrated into a poster. The AI registers if people look happy, sad or neutral when they see the poster. Based on these findings the ad is adjusted and tested again until ultimately the perfect ad is formed.
This was all just about visual advertisements. But AI can also create content for written marketing activities. Marketers need to write too. Email is also often used for advertising. Phrasee is an AI tool that helps with the creation of subject lines for emails (Phrasee, sd).
Domino’s Pizzas is loved by many, but would you click on their email if it were to appear in your mailbox between all these other email ads? I know I wouldn’t. Domino’s told Phrasee they wanted to stop spamming people’s overflowing inboxes. The AI generated a subject line that probably had to make people smile when they looked at their spam. At least, it did make me smile. The subject line said: “Find us someone who hates pizza, and we’ll find you an elf riding a unicorn”. Turns out, the email was opened 26% more (Phrasee, 2018). I supposed it made people smile.
Another copywriting AI just like Phrasee, is Persado (Persado, 2019). It will not only generate text for companies but also do it in the company’s communication language. An AI might write text but if it starts writing formal sentences for Disney, it probably won’t do much good for the brand. The AI produces this language based on already existing content of that brand.
But these AI written content experience the same issues as the video advertisements. It has to be checked by people to see if its correct. It might create inappropriate content. It might not understand human concepts such as verbal irony. If AI learns what content exactly needs to made does a company still need a whole marketing team or will one marketing manager be sufficient? Will creatives still be needed if the AI creates the content?
People will probably continue creating content. Whether this is in the form of video, writing, image or any other format. An AI needs to learn from a dataset. A dataset that can only be created if there is original content. The AI needs marketers and creatives to learn from. But marketers and creatives, but especially marketers need AI by now. Some people might not realize it yet. But if I were to take away all the AI tools marketers use tomorrow, they probably wouldn’t know how continue their job.