Coca-Cola Rewires its Operating Model Around Agentic AI

Coca-Cola is about to put its roughly US$4bn global media account into review, but the real prize is far bigger than advertising. The drinks giant is rebuilding how it runs around first-party data and agentic AI.

The company’s central preoccupation now is fusing the data it owns with the data its partners hold, a theme its finance chief pressed at the DBAccess Global Consumer Conference.

“One of our core areas [is] how do we take the first-party data that we own, which is proprietary, and marry that with the customer data,” says John Murphy, President and CFO of Coca-Cola, describing the work as creating “a sort of new intelligence”.

A review that is really an operating-model rebuild

In its own words, Coca-Cola is evolving a “digital-first marketing operating system for future growth," Ad Age first reported. It wants to move from traditional media planning toward reaching consumers through technology, “including agentic tools”.

James Quincey and Murphy are treating data and AI as the spine of the business, with the review as the visible edge of a much larger restructuring.

The contest covers media, data and technology across most markets, while excluding North America, Japan and South Korea, and leaving global creative and PR with the incumbent.

Run by consultancy Mediasense, the review begins in July with a decision expected in the autumn. For any chief executive running a transformation programme, it is a live case study in using a routine procurement moment to force a deeper change.

Two holdcos, two theories of data

Publicis Groupe has spent years building a proprietary stack across Epsilon, LiveRamp and its CoreAI system, positioning itself as a technology company that also does media.

WPP, Coca-Cola’s global network partner since 2021, takes the opposite view. It acquired InfoSum to connect data sources without centralising them, a privacy-safe model now folded into its wider platform.

One holdco bets on owning the database, the other on owning the connective tissue between databases. Coca-Cola’s choice will signal which philosophy a marquee advertiser is willing to stake its growth on.

"The real battleground is how you show Coca-Cola how your data, your approach and the AI it plugs into actually matters,” says Robert Webster, Founder of AI marketing consultancy TAU, who rates the two holdcos evenly across regions.

The plumbing problem money cannot fix

The deeper challenge is one most large companies share. A Bain & Company survey of 951 firms found the leading reason AI programmes underperform is that organisations still cannot reliably reach their own data. That is despite a decade and hundreds of billions spent on modernisation.

Research from MIT points to the same wall, finding that 95% of corporate GenAI pilots stalled, mostly on tools that integrate badly. The Coca-Cola review, in that light, is a test of which partner can solve a plumbing problem the enterprise world has largely failed to crack.

The holdco that wins by locking Coca-Cola in hardest is not the one that serves it best. A single global architecture stitched across different agencies becomes a web of dependencies that is painful to unwind.

The lasting advantage will belong to whoever can deliver intelligence while keeping the client’s data portable. Coca-Cola is not really choosing an agency. It is choosing whose theory of data to believe, and wagering its next chapter on getting that answer right.