Generative artificial intelligence (AI) remains at the top of the 2025 agenda for marketers as they work to determine how the technology will evolve and affect various functions in their organizations.
Among brand marketers, Coca-Cola has led the way in a space that saw a surge in attention and investment following OpenAI’s debut of ChatGPT in late 2022. But the beverage giant’s early moves actually pre-date the launch of ChatGPT, giving the company a strong base from which to experiment.
The results of those experiments, at least in the marketing media bubble, have been mixed, from the successes of its “Create Real Magic” platform to the contentious response to last year’s Christmas campaign. During it all, Pratik Thakar has been leading the way as the global vice president and head of generative AI for The Coca-Cola Company.
The following Q&A draws from a larger conversation that occurred during a virtual event hosted by Marketing Dive and has been edited for clarity and brevity. You can register here to watch a replay of the full event.
MARKETING DIVE: Coca-Cola was an early mover and has been a leader in the generative AI space the last few years. Where did that journey start?
PRATIK THAKAR: Pre-GPT launch, pre-hype cycle of generative AI, we actually started working with gen AI and we were working on a campaign called “Masterpiece.” That's where we connected with [text-to-video AI model] Stability AI in London, the makers of Stable Diffusion. That’s when we understood the importance and the utility of gen AI. One of the things was the perfection we were getting, especially when you are working on a very specific project with specific timelines.
That’s when we started our gen AI creative incubator, in late ‘22, and that's when GPT got launched and the whole hype cycle started. Bain is our consulting partner, and they came up with this proposal of collaborating with OpenAI, and we were the first one to raise our hand and said, “Yes, we want to be part of it.”
We announced that partnership two years ago, and everyone was so skeptical about AI, gen AI, OpenAI. It was with a relatively unknown company [but] we believed in that technology and we knew that it was going to be big. We put together a team of six people — including legal, public affairs and communications and tech people — and we created a sandbox.
We brought DALL-E, their image creation tool, and GPT together, and we launched “Create Real Magic.” It worked out very well.
How has that team grown and where does it live within the Coca-Cola organization?
There is no specific department of AI. Globally, we have a team of two: One is myself and another is a colleague looking after internal governance within the company. I focus on more consumer-facing ideas and experience development.
Other team members are all living in different geographies. They're part of different units and different functions, but they are early adopters and enthusiasts that wanted to be part of this. They have specific projects which I'm leading, so they work with me very closely. They give their 50% to 40% of time every day, and they have a commitment for long-term project involvement.
It builds their career, their capability, their skill set, and at the same time, it also builds the company’s ecosystem and capability overall, rather than creating a separate unit. The ideal scenario for everyone is basically that we don't need a person like me. Everything will be gen AI, like the way we use the internet and we use mobile; we don't really need the director or VP of internet and mobile nowadays.
How are you balancing the needs of both consuming-facing and internal uses with generative AI?
So let me give you an example. Gen AI is already distributed, it's democratized. Everyone has access to it, and people can come up with ideas. We have created an internal mechanism where everyone who wants to use gen AI fills out a very simple form and our small team looks at it and maybe we say, “That’s a perfect use of AI” — maybe just not that particular tool. We can recommend something else which is legally inline with the company's guideline prospective. Also, if we have done something similar, then they don't need to reinvent the wheel: They can learn from that or learn from some of the partnerships I'm doing.
Another part of my role, apart from creating experiences, is also reaching out to other AI innovators and tapping into their alpha and beta model products so we have early access to their products before anyone else and so we understand what innovation is coming in. We start building our products and our consumer-facing experiences with them while it's a hidden secret for the rest of the world. When they launch, we are ready with our products.
Can you walk us through a campaign that utilized generative AI?
Coca-Cola and Christmas was a global activation. It’s very ingrained with the Coca-Cola brand. We wanted to scale our journey and stress test all the things we and our partners can do.
We looked through two different lenses. One, we took one of our 1990s commercials, “Holidays are Coming,” and we recreated that using gen AI. That film required a lot of super realistic but fantastical imagery and storytelling, and that's what gen AI does very well. We ended up creating three versions of that film, and we put it out in front of consumers. All three of them worked very well with consumers.
We also created this platform called “Create Real Magic” and we took our Santa from 1931 and put together a video about how you can converse with Santa. The insight was that everyone has shaken hands with Santa, taken a picture with Santa or gotten a gift from Santa, but no one had that opportunity or time to sit down with Santa and have a conversation and ask questions.
We worked with OpenAI, Microsoft Azure and Mimic and we created a 3D digital twin of our historic 1931 Santa and then put a conversational AI into that with OpenAI's help. You can actually talk to Santa in 26 different languages across the world. It's a real time, unique conversation, and then, based on your conversation, it will create a snow globe for you for social media. That kind of stuff, which only gen AI can do, is very true to the Coca-Cola brand. We feel that's where the technology does justice to our ambition.
The TV commercial drew a lot of attention and became a lightning rod for some of the controversy over the use of gen AI. How are people getting gen AI wrong, and what did you learn from that experience?
First, it's not like you input one prompt saying “create an ad” and it pops out an ad. A lot of creative decisions are all made by humans. We worked with three different studios, one in L.A., one in San Francisco and one in Kuala Lumpur in Malaysia. All three studios have creative technologists. They are engineers, early adopters and storytellers, and we gave them as a brief our 1990s film because we knew that we had rights to use that as the prompt. That film was shot in Europe, in one village. I needed to take it to the next level. Gen AI opens up possibilities: you can show different parts of the world, it becomes international.
We also used multiple tools, and we had certain dos and don'ts in terms of what tools to use where we have our partnerships. We have a legal understanding of how those tools are built. We had our own guideline and we told our team to use those tools. There was also the music; we composed that with humans.
Once we had all three films ready, we researched it extensively with consumers in North America, Europe and Latin America, with at least three different research agencies and different sets of consumers, and it worked very well. It was not just an average success rate in terms of its acceptability, understanding and entertainment value, but it passed our own and competitions’ Christmas ad benchmarks.
Then, in North America, our team deployed with programmatic media. We had selected 12 different locations, and that's where you can solve content with geotagging. It's hyper personalization and a way of distributing content to different regions.
Some consumers loved it, a lot of people didn't like it. They had a different perspective. Journalists reached out to ask for our side of the story. Three or four different research agencies, independently, without even checking with us, they conducted studies and they published those results. Maybe some sectors of people didn’t like it. That's okay. We need to respect everyone's point of view. Not everyone will agree with everything we do; that's nature.
