AI in advertising: more misses than hits, so far

With the possibilities offered by AI, it was eventual that the advertising industry was going to have fun with it.

Think of the most basic AI use case: to replicate someone’s voice to say anything (based on your input). Even better, make their face, in a video, say that. This is ‘deepfake’, once difficult as a process, now made easy with AI.

In the pre-AI era, the advertising industry had to do things the hard way.

Consider the famous ad campaign for Old Spice Red Zone After Hours body wash, by Wieden + Kennedy Portland in 2010, featuring actor and former football wide receiver Isaiah Mustafa. Beyond the smart Super Bowl TV ad, the agency also produced a series of ‘response’ videos, all featuring Isaiah mouthing lines written in response to quips and questions sent by audience members after watching the ad on TV. There were a total of 185 response videos, using questions and quips gathered from Facebook comments, Reddit, Twitter, and YouTube replies… all produced in under 3 days, close to the original ad airing on TV.

All of this was in 2010. There was no AI involved. These videos were shot the old-fashioned way with Isaiah mouthing all the lines and acting them out.

Today, they could produce a million such ‘response’ videos using AI, in no time, without needing Isaiah. Just enter text and you can make Isaiah say anything you want. The technology is now ready.

So, what have ad agencies done with this technology so far?

For starters, there’s the much-awarded Cadbury’s campaign featuring Shah Rukh Khan, by Ogilvy India.

I have severe reservations about that campaign and have written about it extensively.

The crux? Getting Shah Rukh Khan to mouth specific store names and locations, through AI, at the end of the video, is merely a curiosity-inducing wonder, not a compelling call to action. In fact, the Cadbury’s ‘Shah Rukh Khan My ad’ going viral produces the opposite result – if people think they have already seen it, they wouldn’t even care that Shah Rukh Khan is uttering the name of a local store called ‘Pooja Electronics’ in Chinchpokli at the end of the ad. They have seen the actual ad, and what he utters, via AI, in terms of specific store names and locations wouldn’t stand out and wouldn’t become a compelling call to action.

A more recent campaign on these lines is by the agency VMLY&R, for Virgin Voyages, featuring Jennifer Lopez + AI.

Unlike Cadbury’s campaign, Virgin Voyages’ new campaign has a more purposeful use of AI in service of an actually useful goal – personalized invites with the voice and likeness of J.Lo!

Virgin Voyages campaign’s goal is to make J.Lo say something specific about the trip you are planning and send that as a personalized invite. Because the invite lands in your email through a known/trusted person, you are more likely to open it and view it. And because J.Lo is the person fronting the invite, presumably uttering your name and your friends’ names (the microsite asks for these details), you are more likely to be excited. So, more purposeful use of AI, unlike Cadbury’s which used the same technology to append the digital equivalent of print ads’ dealer panel in video.

The promo for Virgin Voyages’ campaign is spot-on! It is engaging, humorous (in showcasing the different ways the ‘crew’ uses J.Lo’s likeness in how each person believes the invite should be), and adds the necessary googlies as the ad progresses while everything is being fronted by J.Lo!

Sadly, that’s where things stop. The execution of the idea by the agency VMLY&R and the startup AI agency Deeplocal is disappointing. When they claim ‘personalized’ invite and ask for details like my name, and the invitees’ names, my assumption was that I’d get to see a custom video where J.Lo says, ‘On behalf of Karthik, I’d like to invite you to…’. But all J.Lo says is ‘Hello Sailor…’ (regardless of the name you input!) and mouths generic sentences. The names you input only appear in text, in the email!

The Cadbury’s campaign execution at least went the whole way even if the goal was banal. This one has a clearer, more exciting goal, but the execution, after a specific promise, is both misleading and banal.

And then there’s this brand new campaign for Oreo, by the agency Leo Burnett, featuring Farhan Akhtar.

Honestly, I’m befuddled by Oreo’s campaign.

Or, to quote the campaign terminology, I’m at a loss of words 🙂

Leo Burnett’s idea is simply this: using AI, they send you a voice note based on your text input, in the voice of Farhan Akhtar.

The context for this is ‘if you are at a loss of words to say something to someone, tell Farhan your predicament, and get a voice note in his voice that offers a solution, in terms of what you can say’.

I have so many questions! But let me stick to just 4.

1. Why Farhan Akhtar? Is he the kind of person who is known for smart responses, either in his movies or in his off-screen persona? I hardly think so. I mean, I like him as an actor and a director, but he is hardly someone who can carry this ‘Are you at a loss of words? Let me help you!’ drill. There’s no context as to why he’s fronting this campaign even in the promo! If it was say, Arshad Warsi, and his Circuit persona, the link makes some sense. Or Sanjay Dutt’s Munnabhai persona. Or even Prabhas’ wise Baahubali persona!

2. What am I supposed to do with the voice note from Farhan (or the one generated by AI using the likeness of Farhan’s voice)? If, for instance, I input into the microsite that I want to tell my friend not to send me any more WhatsApp forwards, simply informing me of the smart response in plain text is enough. I don’t need that in Farhan’s voice as a voice note since I’m not going to share that voice note with my friend for the simple reason that Farhan is responding to me, addressing me (and not my friend)! That’s severely overdoing the technology use. Rather, it is a solution waiting for a problem/gap. Or, in other words, the agency thought of the AI use-case first and then attached a need in hindsight.

3. The campaign execution is a massive putdown. After seeking input about who the message is intended for and what my predicament is, it asks me for my name, phone number, and email ID. And I have to agree to ‘Yes I would like to receive messages via WhatsApp/email’. If I leave it unchecked, it forces me to accept it! Why not send it via email alone? Why both WhatsApp and email? Data harvesting?

4. Still, I persisted. Shared all details. And got an email reply, and on WhatsApp too. The result was a voice note that was utterly banal. When even plain text would suffice, they sent a voice note, that too with middling, pointless rhyme and wit of a 10-year-old!

“Tell your pal, with a smile and a wink, no more WhatsApp forwards I kindly think
Let’s stick to chats that are fresh and new, it’s just a preference, it’s nothing against you”

I understand the AI/voice cloning technology provider is Resemble AI.

In all these four campaigns (Old Spice, Cadbury’s, Virgin Voyages, and Oreo), the basic construct is very simple.

1. There is an input.

This is what is fed to the ‘system’ where it is taken and articulated by someone, and in these examples, the someone is a celebrity.

The input for the Old Spice campaign was comments on social media.

The input in the Cadbury’s campaign was specific store names and their locations.

In the case of Virgin Voyages (despite the poor execution) and Oreo, the audience provides the input as to what the celebrities should utter.

2. There’s technology.

In the Old Spice campaign, there was no technology. It was manually, and painstakingly done!

In the other 3 campaigns, they used AI and voice and video/face cloning (to get the lip sync right, in case of the Cadbury’s campaign), among others.

This technology can now be safely taken for granted. It just exists and can be used by anyone. That is, any celebrity can be made to utter anything, voice or face, or both, even without their involvement.

3. There is an output.

This is the manifestation of the input, after going through the technology intervention.

In the Old Spice campaign, even without any technology, people posted questions, quips, and comments, and the scripting team at Wieden + Kennedy Portland wrote clever, quick-witted reactions that were performed by the celebrity.

The Cadbury’s campaign execution is a bit like mail merge in Microsoft Word. There is a database with store names and locations. That database is matched with Shah Rukh Khan’s script – or, his digital likeness’s script. The AI does the merging and produces multiple videos that end differently, with each store name and location.

The intent of the Virgin Voyages’ campaign was supposed to be a customized video where Jennifer Lopez (or her digital likeness) invites someone using their name. But, based on what I ended up seeing, they have skimped heavily on technology and have merely stopped with email text-level customization.

With Oreo, they have gone overboard in terms of technology. That is, the execution of the technology is spot-on, but the reason for using it is flimsy.

Taking a cue from the Old Spice campaign, creative magic happens when:

  • the input is interesting and topical. That is, people want to participate and look forward to the output with excitement.
  • the technology can now be taken for granted.
  • the output is fun and shareable, fulfilling the excitement promised at the input stage.

Cadbury’s had functional inputs, and the output was functional too. A bit too functional, even though it was positioned, for the purpose of awards, as a panacea for all of marketing’s problems.

Virgin Voyages had everything in place, though it was a misdirection, and the execution was a complete misfire.

Oreo’s input is hardly thought-through, and barely interesting (rather, it seems forced). The technology works really well here. And the output is an overkill. It’s not shareable, and the creative execution is boring and pointless.

I’m still waiting for the AI-based equivalent of the Old Spice campaign, with the kind of interactive excitement it whipped up and became a digital cultural landmark.