Few brands have managed to stay as culturally resonant for as long as Coca-Cola — and as attendees learned at the second annual AI Deciphered in New York City, that longevity has as much to do with embracing change as it does with tradition.
The daylong conference, hosted by Campaign, PRWeek and MM+M, closed with a keynote from Pratik Thakar, Coca-Cola’s VP and global head of generative AI, who outlined how the beverage giant is using AI to fuel creativity, efficiency and consumer connection.
Moderated by Gideon Fidelzeid, VP of Haymarket Studio, the session came against the backdrop of recent buzz around Coca-Cola’s new set of AI-generated holiday ads, which have sparked conversation about the brand’s approach to AI-powered storytelling.
From lessons learned to bold experiments and future plans, here’s a closer look at what Thakar shared about the 139-year-old brand’s generative AI journey.
How Coke stays one step ahead
Thakar’s philosophy is simple: “We cannot be ahead of the curve if we wait for the solution.”
To maintain that edge, Coca-Cola is securing early access to tools. The company approaches tech firms across the world — from Silicon Valley giants to student-run startups — and asks for alpha and beta access so Coke can experiment while the tools are still being built, rather than waiting for commercial launches.
That early exposure fuels a second pillar of the brand’s approach: building a global creator-researcher ecosystem made up of independent PhDs, artists and entrepreneurs who prefer to challenge norms and work outside big corporations but bring breakthrough thinking. The company works with the big and small, mirroring how Coke products live everywhere from Walmart to mom-and-pop shops, Thakar explained.
Those collaborations feed into a pipeline of incubator projects — more than 20 at any given time — with the understanding that “some of them are going to fail,” Thakar said, “and some are going to take us ahead.”
Coca-Cola structures these experiments around a flexible tech stack and keeps legal and public affairs teams close, especially when partnering with small, fast-moving startups that may lack mature compliance systems.
Internally, the transformation isn’t owned by a siloed, standalone AI department: Only two people at the company, including Thakar, hold AI titles. His position came about after discussions with Coke’s global CMO, Manuel “Manolo” Arroyo, who also recognised the need for a role that could drive AI across marketing and beyond.
Using a bit of ChatGPT to help draft the job description, the team positioned the role within the marketing transformation group under the CEO’s office.
“We said let’s not park it in one particular category, brand or function of
“We said let’s not park it in one particular category, brand or function of creativity or media because it needs to transform everything,” Thakar said, who’s been in his current role since June 2023.
Coca-Cola spreads capability through strategic 6-to-12 month assignments and short-term experiential projects, where employees spend roughly half their time on AI initiatives tied directly to their market or brand responsibilities.
“Hundreds of associates are raising their hands across the company saying, ‘I want to be a part of this,’” Thakar said. That network, more than any single tool, is what Thakar said will keep Coca-Cola ahead as AI accelerates.
From experiments to impact
That future-focused mindset didn’t emerge overnight. Coca-Cola’s foray into generative AI began before GPT-style tools hit the mainstream.
In 2022, back when Thakar was Coke’s head of global creative strategy and content, the creative team faced a challenge while developing its Masterpiece campaign: matching the precise brushstrokes of iconic artists — from Van Gogh to Warhol — as the Coke bottle moved across their original works in the film.
To solve it, the brand partnered with Stability AI, the company best known for its text-to-image model Stable Diffusion, using AI to analyse and replicate the varied artistic styles.
“That’s how I understood how GenAI works — not by reading a book or white paper but literally using it,” Thakar said.
Around the same time, Coke deepened its partnerships with OpenAI and Microsoft, ultimately combining DALL-E and GPT to build a proprietary sandbox trained on archival brand assets. The result: Create Real Magic, an AI platform and brand campaign that allowed anyone to generate original Coke-themed artwork, with select pieces showcased on major global billboards, from Times Square in New York City to Piccadilly Circus in London.
Working months before tools were publicly available, the company co-engineered a real-time, 26-language “talk to Santa” experience, blending Azure speech-to-text, GPT-4 and a custom multimodal model to generate personalised snowglobe videos.
“You just need to know a little bit, like six months, in advance than the rest of the other people, and then you are ahead of the curve,” Thakar said.
“People always ask me, ‘Did it work?’” Thakar added. Millions of people spent an average of eight minutes and 20 seconds interacting with Santa, creating shareable social content and giving Coke first-party data and consent. “That’s a good example of good marketing.”
These projects — from Masterpiece to Create Real Magic to its latest work with MIT on Save the Orange — show how the company moved from testing generative tools to building a full creative and operational framework around them at scale.
Quoting former Coca-Cola CEO Robert Woodruff, Thakar said, “The future belongs to the discontented” — the people and brands willing to question the status quo and push for what’s next.
