CNN starts building its own AI agents to handle future media deals

CNN has decided to build its own agent infrastructure instead of waiting for the industry to catch up. The move positions the network to handle deals where AI agents on the buy and sell sides negotiate and trade media inventory directly. Faisal Karmali explained that the company wants the system ready and tested before full trading begins in early 2027.

The plan mixes internal development with help from third-party vendors and demand-side platforms. CNN is also working with the IAB Tech Lab to make sure its protocols line up with what buyers will use. This includes checking standards like the Ad Context Protocol. Karmali noted that starting early means the system can learn how agents talk to each other even before money changes hands. 🤖

Timeline and testing steps

By the end of the second quarter CNN expects to finish scoping the agentic protocols. The third quarter brings tests on one or two properties to check how large language models interpret them. In the fourth quarter the focus shifts to buyer behavior and small budget experiments for agent-to-agent trading.

Karmali told Digiday that Q1 2027 marks the target for full trading as a new revenue stream, though the schedule could change as the technology develops. The infrastructure work is led from the U.S. but can roll out internationally with relative ease. For now the emphasis stays on consistent agent-to-agent communication, including requests for information, pricing talks, and task delegation.

What this means for publishers and buyers

Other publishers have started similar work. News Corp is building an internal agent for titles like The Sun. On the buy side, agencies have run early tests that cut costs significantly. Scott Messer from Messer Media pointed out that most current examples still focus on campaign setup rather than full trading, and early deals will likely stay in reservation or guaranteed formats.

Andrew Mole from PubX described a future of agentic abundance where publisher-specific agents and broader platform agents operate together. Publishers gain tools to see who is buying and then reach out for direct conversations or bigger campaigns. CNN sees agent trading starting with lower-funnel performance work that relies on real-time data, before moving toward branding once payments and commerce features arrive.

The network wants ongoing talks with the buy side to keep context protocols aligned. Karmali said the industry rarely gets to start a new trading method from scratch, so getting the basics consistent matters most right now.

source