Coca-Cola isn’t interested in AI pilots that generate headlines but don’t drive results. Under a directive from top leadership, including CEO James Quincey, the company’s technology team is under strict orders: no AI experiments unless they clearly boost revenue or cut costs.
“We don’t run hobby projects,” says Neeraj Tolmare, Coca-Cola’s Global Chief Information Officer. “Every pilot must be tied to a real outcome that moves the needle for us.”
Artificial intelligence at Coca-Cola is measured, disciplined, and deeply pragmatic. From generative content tools to demand forecasting models and early experiments with agentic systems, the company is applying a rigorous return-on-investment filter to everything it builds.
Tying AI to Tangible Gains
Coca-Cola’s AI initiatives are evaluated based on scalability, implementation cost, and relevance across its vast global network of more than 950 production facilities and 200 bottling partners.
A recent pilot aimed at improving demand forecasting used AI to analyze historical sales, weather patterns, and geolocation data. The system automatically sent WhatsApp messages to store managers, advising them when to restock fast-moving items like Sprite or Diet Coke. The pilot, tested in three countries, delivered a 7% to 8% sales lift compared to outlets not using the tool.
“The idea is to put intelligence directly into the hands of the people managing our retail execution,” Tolmare says. The success of the test has opened the door for broader rollout across Coca-Cola’s distribution network.
Content at Global Scale
Another major push involves the use of AI in marketing. Coca-Cola sells in over 180 countries and generates content in 130 languages, an operation that’s both massive and expensive. Using generative tools, the company recently produced 10,000 content variations from just 20 base assets.
“Consumers were 20% more likely to engage with AI-generated content compared to previous campaigns,” Tolmare says. “And we created it three times faster.”
With more than 200 brands and one of the most complex marketing ecosystems in the world, Coca-Cola has long faced the challenge of delivering culturally relevant creative at global scale. AI is now becoming a key lever in reducing cost while speeding up content creation across formats and languages. Automation hasn’t eliminated the need for human creatives, but it has allowed teams to move faster while staying aligned with local and cultural expectations.
Automating Brand Consistency with Fizzion
Project Fizzion, developed in partnership with Adobe, is Coca-Cola’s answer to the long-standing challenge of global brand consistency. It uses AI to embed dynamic brand guidelines directly into design files. Designers work in familiar tools like Photoshop or Illustrator, while the system interprets their decisions in real-time, creating machine-readable “StyleIDs.”
“With Fizzion, our design elements become smart,” said Rapha Abreu, Coca-Cola’s Global VP of Design. “This is about embedding AI at the heart of our brand system so creativity can move faster, without losing its soul.”
Fizzion reduces formatting time and allows creative teams and agency partners to produce assets up to 10 times faster without compromising brand integrity.
Despite its efficiency gains, Coca-Cola’s use of generative AI hasn’t been without friction. A holiday campaign attracted backlash, and another instance involved a fabricated quote wrongly attributed to author J.G. Ballard.
“These tools are powerful, but they can also make mistakes,” Tolmare says. “We’ve built guardrails to catch hallucinations and bias, and we’re constantly refining them. That doesn’t mean we’ll never stumble, but we have mechanisms in place to respond responsibly.”
Agentic Systems Still Under Evaluation
The company is actively exploring agentic AI tools that can operate autonomously across workflows. Partners include Microsoft, SAP, and Adobe, along with proprietary systems built on Coca-Cola’s internal data. But these projects remain in the pilot stage.
“We haven’t launched agentic AI into production yet,” Tolmare says. “We’re close, but only if the business case is solid.”
That may be true at the corporate level, but Coca-Cola Consolidated, the company’s largest bottler, has already integrated agentic AI into production. As Bhavin Shah, CEO of Moveworks, noted in a recent post: “Next time you drink a Coca-Cola product bottled by Coca-Cola Consolidated, you might taste their new, not-so-secret ingredient: agentic AI.”
“We haven’t launched agentic AI into production yet,” Tolmare says. “We’re close, but only if the business case is solid.”
At Coca-Cola Consolidated, the company’s largest bottler, AI adoption started with automating internal helpdesk tasks. “We saw that employees were calling the service desk for password resets, something that could be automated,” said Retha Summers, IT Director at Coca-Cola Consolidated. The team implemented Moveworks’ AI assistant within Microsoft Teams and Slack, which handled support requests at scale—even as the workforce tripled in size, the number of agents remained the same.
The team has since moved beyond automation into agentic AI, using Moveworks’ Agent Studio to build intelligent systems internally. “With Agent Studio, our teams can build fast and build right,” Summers said. During a recent hackathon, teams developed more than a dozen production-ready use cases in a matter of days.
Coca-Cola runs 80% of its workloads on Microsoft Azure, with the rest split between AWS and Google Cloud. This multi-cloud strategy helps the company stay flexible and optimize costs.
That same thinking applies to its generative AI stack. Coca-Cola works with OpenAI, Microsoft, Google, Meta, and Anthropic. “We don’t want to paint ourselves into a corner too soon,” Tolmare notes.
For Coca-Cola, AI is a tool for transformation not for experimentation. The company’s strategy is grounded in performance metrics, operational scale, and long-term value creation.
“AI at Coca-Cola isn’t guided by trends—it’s governed by outcomes,” Tolmare says. “Whether it’s shelf restocking, creative production, or internal automation, each initiative must deliver tangible business impact. That’s how we’re approaching this technology—one pilot, one outcome at a time.”
