CES is a long-running technology trade show (Consumer Electronics Show) held each January in Las Vegas. Alongside its traditional consumer electronics focus, in recent years it’s become increasingly important to the advertising and media technology industry.

I did not attend the recent event, but reports indicate this CES was dominated overall by physical AI and, for advertising specifically, by agentic AI announcements (appropriate, as the first time I heard the term “agentic” was CES 2025). 

Today I offer a comprehensive overview of the agentic advertising developments from CES and what I think they mean for news media publishers.

What CES revealed

We’ve previously discussed the many ways agentic AI would reshape advertising. Over the course of the first week of the year in Las Vegas, ad agencies and platforms revealed their AI “engines” while ad tech’s biggest players unveiled systems and evolved protocols designed to let AI agents negotiate, plan, and execute advertising campaigns.

As I wrote in my November newsletter on emerging agentic advertising standards, many of these developments were “too early to mandate, too uncertain to predict, but also too significant to ignore.”

CES didn’t change that assessment, but it did provide substantial evidence that agentic infrastructure development is continuing across every layer of the programmatic stack and beyond.

From coverage on the ground, here’s what happened, what I think it means, and how I think news publishers should be thinking about all of it.

Before CES: the first agentic test campaigns

PubMatic and Butler/Till launched what they described as the industry’s first fully autonomous CTV campaign in December 2025. The campaign for Clubtails (Geloso Beverage Group) used PubMatic’s agentic capabilities with Anthropic’s Claude as the AI interface. Butler/Till submitted a natural language brief, and PubMatic’s agents interpreted it, generated the media strategy, and executed the buy autonomously. 

The results: 87% reduction in setup time and 70% faster issue resolution. Scott Ensign, Butler/Till’s chief strategy officer, described it as the agency’s “first end-to-end agentic campaign.”

Magnite and MiQ completed one of the first Ad Context Protocol (AdCP) test buys in December, with Scope3 serving as the buyer agent. Magnite embedded a seller agent directly into SpringServe, its video ad server, with additional pilots planned for early 2026. John Goulding, MiQ’s Global Chief Strategy Officer: “In 2026, we’re excited to start scaling agentic ad buying through MiQ Sigma’s trading agent.” 

Agentic infrastructure takes shape across the programmatic stack

The week’s announcements revealed that agentic capabilities are being embedded throughout the advertising technology ecosystem — from demand-side platforms to supply-side platforms to the standards bodies working to govern them.

Yahoo DSP launched what it calls a Yours, Mine, and Ours framework for agentic AI. Three capabilities went live immediately: a campaign activation agent enabling external AI connections through Model Context Protocol (MCP), a troubleshooting agent that proactively identifies and resolves pacing issues, and an audience exploration tool for AI-driven audience discovery. 

Adam Roodman, general manager of Yahoo DSP, framed it as fundamental workflow transformation: “Agentic AI changes how media buying actually gets done.”

PubMatic launched AgentOS, positioning it as the first operating system built specifically for autonomous advertising execution. The system runs on NVIDIA-accelerated computing and supports what PubMatic calls “three-layer Advertising Intelligence architecture.” Launch partners include WPP Media, Butler/Till, MiQ, and Wpromote.

Viant took the most aggressive stance with its Lattice Brain product, explicitly embracing a “no-human-in-the-loop” approach. CEO Tim Vanderhook told Ad Age: “From planning and setup to reading reports and making changes, AI is now making all those decisions.” In a test campaign for craft-ware company MacKenzie-Childs, Viant claims its AI achieved a US$15 cost per action compared to US$45 from human traders.

The standards battle intensifies

Perhaps the most consequential development from CES was IAB Tech Lab’s release of its Agentic Roadmap, which outlines how the industry can scale agentic buying and selling by extending established standards rather than introducing new ones.

Anthony Katsur, CEO of IAB Tech Lab, was direct: “The fastest and smartest way forward is to build on an existing shared foundation, not introduce multiple new standards that create fragmentation.” The road map extends established protocols including OpenRTB, AdCOM, and VAST with modern execution protocols like Model Context Protocol and Agent2Agent.

Notably absent from the announcement was any mention of the Ad Context Protocol (AdCP), the competing standard launched last year by a consortium including Yahoo, PubMatic, Scope3, and others. As I noted in November, AdCP faced questions about adoption given the absence of major demand-side platforms like The Trade Desk, Google DV360, and Amazon DSP from its founding members.

IAB Tech Lab is hosting a public Webinar on January 28 to explain its road map and beginning February 12, will launch monthly “Agentic AI Boot Camps” to provide hands-on guidance for implementation.

Premium TV goes agentic

NBCUniversal and FreeWheel announced they’ve completed their first AI-agent-led programmatic guaranteed deal, including live sports programming across both linear TV and streaming. Working with agency RPA and AI analytics vendor Newton Research, the system uses buyer and seller agents to automate what they called “some of the most manual, high-stakes” operations in advertising. The first major campaign will run during NFL (American football) playoff games in Q1 2026.

Mark Marshall, chairman of global advertising and partnerships at NBCU: “This step forward will redefine how inventory is bought and sold, and what better place to start than within our live sports inventory.”

Platform innovation: Reddit’s transparent approach

Reddit unveiled Max Campaigns, an AI-powered media-buying tool that automates targeting, bidding, and creative optimisation while pitching greater transparency than competitors. Over 600 brands participated in alpha testing, with Brooks Running achieving a 37% decrease in cost per click. Split tests showed 17% lower cost per acquisition on average.

What’s notable for publishers: Reddit is positioning its “community intelligence” — user-generated conversations — as the key differentiator, and brands are increasingly interested in how Reddit presence influences their visibility in AI chatbots. Jyoti Vaidee, Reddit’s vice president of ads product: “When brands lean into Reddit, both on the organic and the paid side, that sparks more conversation, and then that data powers the LLM outputs.”

Streaming giants: Disney bets on AI and vertical video

Disney used its sixth-annual Global Tech & Data Showcase at CES to unveil a suite of AI-powered advertising tools. 

The company introduced a new video generation tool that allows advertisers to create CTV-ready commercials using existing brand assets — with Instinct Pet Food among the first to test it. Disney also launched the “Brand Impact Metric,” a unified measurement framework synthesising attention, brand health, search, and attribution into a single score. 

The Disney Compass platform is expanding with a new Brand Portal offering AI-powered campaign summaries across all Disney properties. Perhaps most notably for format innovation, Disney is bringing vertical video to Disney+ following the successful “Verts” launch on ESPN — a clear response to TikTok and Reels viewing habits. 

President of Global Advertising Rita Ferro emphasised that Disney is on track to reach 75% automated advertising by 2027, positioning the company as “as much a technology company as we are a storytelling company.”

Agency transformation: the race to build AI “engines” 

CES revealed an intense competition among agency holding companies to build proprietary AI systems — variously described as “engines,” “operating systems,” and “platforms” — that promise to transform how agencies operate.

Havas launched AVA, a global LLM portal that provides centralized access to GPT-5, Claude Opus 4.5, and Gemini 3. CEO Yannick Bolloré framed it as infrastructure for an “AI-driven, human-powered future.” AVA builds on Havas’ nearly €1 billion AI investment commitment through 2027. 

Stagwell unveiled The Machine, built by Code and Theory and positioned as “marketing’s first agentic operating system.” Rather than replacing existing tools, it integrates with Figma, Slack, Teams, and Adobe to create a unified workflow. CEO Mark Penn: “We’re making a bet that we’re going to be an AI winner.” 

Horizon Media launched HorizonOS as an “open ecosystem” alternative to holding company proprietary systems. President Bob Lord drew a sharp contrast: “Holding companies are building closed systems of machines talking to themselves, eventually delivering the most reductive solutions. We’re taking the opposite approach.”

For publishers, these agency moves signal potential compensation model shifts. Ad Age reported that Havas is exploring moving from hourly/headcount basis to payment for results: “If AI lets teams do better work faster, then hourly rates punish efficiency.” The agencies publishers work with may soon operate under fundamentally different incentive structures.

Reasons to be skeptical

Amid the CES enthusiasm, I think it’s worth noting several factors that warrant caution.

• Limited pilot success so far. Gartner research from November found that among marketers using generative AI (but not AI agents), just 5% are seeing significant gains on business outcomes. For those further along with agent capabilities, they “aren’t yet delivering the promised business performance” either. Gartner sees agentic AI in the “peak of inflated expectations” phase of its hype cycle.
• Efficiency isn’t the same as effectiveness. Most CES announcements emphasised operational efficiency — faster setup, reduced manual work — rather than improved campaign performance. The business case for publishers and whether agentic systems will drive better CPMs or just faster commoditisation remains unproven.
• Standards fragmentation is real. The split between IAB Tech Lab’s approach and Ad Context Protocol creates genuine uncertainty. As one industry observer noted, “None of it works at scale without shared standards.” We’re not there yet.
• CMO skepticism persists. Janet Balis of Boston Consulting Group, reflecting on CES, noted that AI advancements “must work toward a business outcome — not just be an implementation of more AI and technology.” She added: “CES was a wake-up call this year for many. I think people are realising how far they are from that.”
• The “Year of the Agent” may slip. Industry analysts predict continued blockers around context fragmentation, governance gaps, and exception handling — problems that are “almost solved” but never quite resolved. Some expect the narrative to quietly shift from “autonomous agents” to “AI-assisted workflows” by year-end. 

What this means for news publishers

Nothing I’ve seen in the CES announcements changes my fundamental assessment from November. Of course you should be continuing to find ways to employ AI automation to processes and insights, while assessing efficiency and profitability.

But when it comes to agentic buying and selling, you should be paying very close attention to developments. It’s still too early to make big bets. If there’s anything to do, it’s to prepare (especially your data strategy) and be in a position to experiment.

What CES demonstrated is continued evidence of agentic infrastructure construction across the ecosystem. The blueprints are being drawn, the foundations poured, and early pilots are going live. But the critical question of if and how advertiser demand will materialise at scale in a way that matters to publishers remains unanswered.

Key takeaways:

• The infrastructure race is real. Yahoo, PubMatic, Magnite, Viant, and NBCU aren’t experimenting in labs but deploying production systems. Publishers working with these platforms should ask their partners specifically how they’re preparing for agentic workflows.

• Standards fragmentation remains a concern. IAB Tech Lab’s road map versus AdCP represents competing visions for how agents will communicate. Publishers should monitor which standards their technology partners adopt but shouldn’t bet heavily on either until clearer market signals emerge.

• Direct deals may transform before open auction. As AdExchanger noted, “Agent-to-agent direct ad deals have arrived. And it might be the biggest innovation the direct market has seen in two decades.” For premium publishers with strong direct relationships, this could be the first place to see impact.

• First-party data becomes even more critical. In an agentic world, buyer agents will seek inventory that can deliver against campaign objectives. Publishers with differentiated audience data, strong contextual signals, and proven measurement capabilities will be better positioned to surface in agent-driven discovery.

• The value equation may shift. As Havas signals with its move toward outcome-based compensation, the entire industry may be heading toward models where automation efficiency is shared rather than hoarded. Publishers should consider how their value proposition translates in a world where execution costs decline. Hence our Measurement & Effectiveness Master Class.

Practical recommendations:

• For all publishers: Stay informed but don’t panic. The infrastructure is being built, but widespread adoption is still ahead. Focus on fundamentals: data strategies, measurement capabilities, and technology partnerships with platforms that are actively investing in these systems.

• For premium publishers with technical resources: Consider joining working groups in listening mode. Map two or three signature products to emerging schemas. Allocate modest pilot budget for Q2/Q3 pending case studies from early adopters.

• For mid-tier publishers: Ask your current ad tech partners when they expect to support agentic workflows and what publisher action will be required. Document your differentiated offerings for potential future exposure to buyer agents. Conserve resources until market signals clarify.

We’ll be tracking these developments closely throughout 2026 to help you make sense of it all.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .