Have a Coke and a… Simile?

For their recent Holiday campaign, Coca-Cola brought together three AI studios (Secret Level, Silverside AI and Wild Card) to work on three commercial versions, reportedly using the models Leonardo, Luma and Runway, with a new model, Kling, brought in near the end of production.

Our Magid tech execs weigh in:

1. The medium has overpowered the message on this one.

This shouldn’t be about AI. It’s not about whether AI can make good commercials. The question is, “is this a good commercial?” 

The means of production shouldn’t be an issue, as Coke has used digital and special effects in their ads for years. They’ve used animation for decades. What is the difference between these ads and the other digital/animated/sfx ads that have warmed the hearts of Americans during the holidays? These ads have some problems… 

The concept and scripts are solid, and harken back to previous campaigns that have set the tone for holiday advertising. But the production falls flat, leaving the consumer to feel that something is just a bit off… and it is!

• The closeups of the hands look funky and unnatural
• The wheels on the trucks don’t spin
• There are street light halos on the road where there are no street lights
• The people (though real models were used for likeness) just came off as a little flat and creepy at times

All of this is easy to miss in a typical one-time viewing of the ads (as most of us do). None of the shots are more than 2-3 seconds max, so they don’t linger on anything. But when you watch it over and over (as many will over the holidays as they watch sports, specials, holiday programs) these things jump out and are then easy to see.

2. Human creativity isn’t going to be replaced anytime soon.

Generative AI represents a monumental landscape pivot for many industries, not unlike disruptions that came before it like CGI, social media, and the internet itself.

Coca-Cola’s latest AI ad campaign is the biggest splash by a major corporation to embrace the technology in its creative efforts. Predictably, there was backlash. A lot of it came from the creative community – people who earn a living by working as graphic designers, editors, story producers, etc. who fear that the use of Generative AI in ads like these from Coca-Cola take jobs from artists, and lead to lazy art.

These concerns are valid, yet oversimplified. Yes, using generative AI in advertising and entertainment will result in lay-offs in one area and new hires in another. But generative AI cannot replace human creativity. Period. The idea of AI developing a creative project from start to publish is misguided. 

The concern around lazy art is probably valid. By design, AI will return the most predictable, suitable answer to a prompt. Humans designed this technology as an auto-regressive model that will provide the best fitting completion to the input based on the data available. It can only return information about what already exists and is expected. Anything outside the realm of predictability is not this model’s strength. 

Think about that in terms of advertising: if a campaign is derived solely from AI and void of the human component, the result is designed to be predictable. Right now, the AI technology is the shiny, new thing. It’s why Google searches for “Coca-Cola ads” skyrocketed after the AI ads launched.

But soon, the technology won’t be novel anymore. What will matter is how people creatively implement the tool into their workflows to design something that stands out. We’ve witnessed the local news industry struggle from a “sea of sameness” effect, in which consumers can hardly tell one news brand from another. Once more advertisers adopt AI technology, creative direction will become more paramount than ever to make a campaign stand out.

Our Magid technology team is the first to admit that AI is imperfect. It’s why we’ve developed content safeguards in our Collaborator tools to ensure that any flawed output gets immediately flagged. People are the key to creating content that stands out; AI is a helpful thought partner and tool to help them do more of it, faster.

3. Brand marketing – and AI – will continue to be a matter of Trust.

A recent Ad Age article interviewed Coke’s AI lead, Pratik Thakar, who brought up the consumer perspective on AI:

“Consumers don’t really look at it like creative directors look at it. Consumers, if the story makes sense, and it entertains them, and it gives information, then it’s good. And that’s what we saw: Consumers were not concerned about Al versus non-Al.”

Our consumer research on AI pushes back on Thakar’s statement here; many consumers are concerned about AI. 

People are split on how trustworthy they find AI to be. This is where transparency about AI usage in an organization becomes paramount. Coca-Cola did disclose – and even publicized – their use of AI to produce these ads. Continued efforts to explain how generative AI is utilized in their organization could help move the needle on the trust scale.

When we asked people who told us they “do not” trust AI, 50% of them told us they still “see the value in it.” And only about a quarter of all respondents told us that based on what they know right now regarding AI, their first reaction is negative. Consumers recognize the potential of AI technology, but we’re in a moment where transparency and education about it is critically important for companies wanting to implement it. 

Makes us wonder – is traditional ad testing really effective at predicting winners? The Coke comms team has done the rounds to talk about their intentions and satisfaction with the ads. They state that their intensive testing hit the success thresholds needed to be approved. They probably would with a single view in a focus group facility or on a mobile or laptop screen. But in the real world, we see a different reaction. As AI and other technologies proliferate, it might be time to update ad testing standards to better replicate real-world environments and exposures. It is also important to ensure emotional tonality aligns with the brand, brand intentions and goal of the ad. I don’t suspect that would be the problem with this ad. But if they had multiple exposure testing or put it through more rigor, they may have found that…

4. “All press is good press.”

It’s clear that these ads will have much more exposure than they would have through their campaign placements alone because of this. Will the questionable quality keep people from drinking Coke this holiday season? Only the sales will show for certain. 

Where the Coca-Cola ads took their biggest risk was the timing and focus on the holidays – a time when people prioritize in-person togetherness and community more than any other. It was a bold move to use AI-generated “people” for their annual holiday campaigns. There has been a strong consumer movement toward seeking quality in-person bonding time, with people increasingly willing to go out of their way or spend more for it, especially in the post-COVID era.

Arguably, a stronger campaign would have centered on that: real human togetherness. However, if the goal was to get people talking about Coca-Cola, mission accomplished.

5. AI just isn’t that great. Yet.

Don’t get us wrong. That AI can create even these commercials at this quality is remarkable. But as amazing and transformative as AI is, it is moving very fast and for some use cases, and it’s just not ready for primetime.

This is why Magid has been more measured as we enter multimodal AI inputs and outputs in our workflow tools. The technology is amazing and delighting. To see a demo of video creation in real time is mesmerizing! But after you see it 3, 5, 15 times, the cracks start to emerge. The quality is very inconsistent, and we just don’t trust it yet on behalf of our clients.

There are many video ad creation tools out there today that let you create spots for pennies. There are several broadcast groups that use these tools for storyboarding or prospecting new advertisers. But even in very small TV markets, these groups still rarely put those automated ads on the air. It’s just not there yet.

For Coke? They had 17 artists spend 3 weeks creating each of these ads, in addition to the AI. And the wheels on the trucks still don’t move? For a company that still puts a person on site at a print plant to approve the specific color red in every print piece they create, that miss is surprising. 

The Bottom Line

AI is intimidating to many because it is new. But Magid research shows that the more knowledgeable people are about AI, the more positively they feel about it. So there could be a level of guilt by association here, as people who are unknowledgeable and fearful of AI are knocking Coke regardless of ad quality. They are just scared of AI. But that will continue to shift, leading us back to the real question at hand: is this a good commercial?