WPP opened Devon’s Point hub, which uses AI-powered tools to automate parts of the development and production process.
AI agents are beginning to move beyond experimentation and into execution across the advertising workflow. But the technology is still a long way from running advertising on its own.
In the short term: ad buys are ramping up
Apostra has already handled automated media buys for some “very large” brands totaling around $1.2 million over the last three months—admittedly a tiny spend, but a sign that advertisers are putting in real dollars behind agentic buying.
Red-light therapy brand Rouge Care is crediting AI agents with a fivefold return on ad spend, after spending $25,000 on CTV ads and driving over $125,000 in attributable sales. The awareness campaign was facilitated through PubMatic’s AgenticOS in the second quarter of 2026.
Still, no one is doing this at scale. Highly autonomous deployments represent less than 5% of all marketing use cases in 2025, according to Gartner. It predicts that by 2028, only 15% of agentic AI deployments will be highly autonomous.
For now, the question isn’t whether advertisers will put money behind agentic buying, but how far that spending can scale—and what infrastructure will ultimately support it.
In the long term: who is responsible when agents go rogue?
Apostra uses AdCP, a protocol developed by a group of adtech companies, including Scope3, that provides a common language for agents to interact. It’s among a handful of protocols within the ad industry, including IAB Tech Lab’s AAMP (Agentic Advertising Management Protocols) and Amazon’s Model Context Protocol.
The rise of multiple protocols could create fragmentation issues, making it harder to scale agentic advertising. If buyers and publishers end up backing different standards, the systems will need a way to communicate, or risk creating another layer of interoperability problems for an industry that is already notoriously fragmented.
