The Rise of the Agentic Media Buyer: How AI is Reshaping Connected TV Advertising

The advertising industry is currently navigating its most significant technological shift since the advent of programmatic bidding. While the term "agentic AI" has become a ubiquitous buzzword in boardrooms and tech forums, its practical application in high-stakes media buying has remained largely theoretical—until now. A new wave of integration, centered on the Model Context Protocol (MCP) and sophisticated API orchestration, is moving beyond simple chatbots to create autonomous agents capable of managing, optimizing, and executing complex Connected TV (CTV) ad campaigns.

For agencies like InterMedia, a firm specializing in performance-based CTV, this is not a future-state prediction; it is a current operational reality. By leveraging the intersection of Anthropic’s Claude, the Vibe.co buying platform, and the Olyzon orchestration layer, the agency is effectively redefining the role of the modern media buyer.

Main Facts: The Architecture of Agentic Buying

At its core, agentic ad buying represents a move away from "button-pushing" manual labor toward high-level strategic oversight. The technical foundation relies on two primary frameworks:

1. Model Context Protocol (MCP): Introduced by Anthropic in late 2024, MCP provides an open-source standard that allows AI agents to securely connect to external applications and databases. It serves as the "connective tissue" that allows a Large Language Model (LLM) to "see" and "act" within platforms like Vibe.co.
2. API-Driven Orchestration: While MCP is the standard for connectivity, platforms like Olyzon use deterministic APIs to ensure that when an agent makes a change to a campaign—such as adjusting bids or shifting budget—the action is repeatable, verifiable, and consistent.

InterMedia’s workflow illustrates this synthesis. Campaign managers can now query Claude for real-time performance data across various publishers like Peacock or Warner Bros. Discovery. The agent synthesizes this data, identifies growth trends, and suggests optimizations. The human manager, acting as a "pilot in the cockpit," reviews these suggestions before authorizing the agent to execute the changes across Demand-Side Platforms (DSPs) and Supply-Side Platforms (SSPs).

Chronology: From Concept to Deployment

The rapid evolution of this technology over the last 18 months has been nothing short of a "Cambrian explosion" in ad tech.

• Late 2024: Anthropic launches the Model Context Protocol, providing the missing link for AI agents to interact with external enterprise software.
• Early 2025: Ad tech firms begin prototyping the "Ad Context Protocol," an industry-specific application of MCP, aiming to standardize how AI interprets ad demand.
• Summer 2025: Vibe.co, following its acquisition by Walmart, begins integrating its self-serve platform with Claude. This was partially born from internal necessity; Vibe employees were already using LLMs to manage their own internal ad spend, prompting the company to productize the capability for its users.
• September 2026: Vibe.co publicly announces its agent-access integrations, marking a turning point where MCP becomes a standard feature for over 1,000 brands.
• October 2026: Olyzon launches its agentic orchestration platform at Advertising Week New York. This tool provides a layer of consistency, allowing agencies to bridge the gap between different DSPs, which often suffer from fragmented terminology and data siloes.

Supporting Data and Market Dynamics

The adoption rate among small-to-medium businesses (SMBs) has been particularly aggressive. According to Vibe.co CEO Arthur Querou, more than 1,000 brands have already adopted the platform’s MCP integration. For these smaller players, the ROI is clear: they gain the sophisticated, round-the-clock campaign management capabilities of a large agency without the associated headcount costs.

InterMedia’s experience confirms that this is not merely a tool for small budgets. The agency has shifted the majority of its biddable campaign management to this agentic workflow within just a few months. The primary bottleneck, however, remains data standardization. Because different DSPs and SSPs use varying nomenclature for identical metrics, human intervention is still required to "normalize" the data before the AI can act upon it.

Furthermore, the "ownership" of these agents remains a subject of industry debate. In the Olyzon-led workflow, the agent is clearly proprietary to the platform. However, when an agent is accessed via a general-purpose model like Claude to interact with Vibe, the lines of accountability become blurred. This raises legal and operational questions: Who is liable for an "agentic error" that results in significant budget overspend?

Official Perspectives: The "Human-in-the-Loop" Consensus

Despite the excitement surrounding autonomy, the industry leaders involved are unified in their commitment to the "human-in-the-loop" model. Sydney Brower, Director of Digital Media and Strategy at InterMedia, draws a vivid analogy: "You can trust AI to fly a plane, but I still want a pilot in the cockpit to watch it."

This sentiment is echoed by Jules Minvielle, CEO of Olyzon. He argues that the industry must be careful to distinguish between agents talking to agents and agents executing commands via APIs. "An agent talking to an agent is not the same as having our agents writing what’s needed directly in the API," Minvielle explains. This distinction is critical for performance marketing, where deterministic outcomes are required for reporting and budget accountability.

David Nyurenberg, SVP of Digital at InterMedia, emphasizes that the skill set required for a "media buyer" is undergoing a radical shift. "Instead of being an expert in how to navigate a platform and push the buttons, you’re an expert in being judicious about what changes to make," he says. The value proposition of the agency of the future lies not in technical execution, but in strategic interpretation.

Implications: The Future of the Agency Model

The implications of this shift are profound for the advertising ecosystem at large.

1. The Death of Platform-Specific Specialization

For decades, agencies have built entire departments around expertise in specific platforms (e.g., "The Google Ads Team" or "The CTV Specialists"). As AI agents become proficient at navigating any interface via MCP, the technical barrier to entry for platform management vanishes. Expertise will shift toward cross-channel orchestration and brand strategy.

2. The "Holding Company" Dilemma

Large agency holding companies are currently racing to build proprietary agents for audience identification and campaign planning. However, they face a "build versus buy" dilemma similar to the one that defined the early days of DSPs 15 years ago. Will they successfully build a superior in-house "Agent Operating System," or will they become dependent on third-party SaaS orchestration layers like Olyzon?

3. The New Definition of "Scale"

With agentic buying, the relationship between campaign size and management time is decoupled. An agency can now manage 100 times the number of campaigns with the same number of staff. This suggests a future where hyper-niche, hyper-personalized CTV campaigns become the norm, rather than the exception.

4. The 18-Month Horizon

Jules Minvielle’s prediction—that every media buyer in the world will be equipped with an agentic orchestration platform within 18 months—is bold, but given the current trajectory, it is not implausible. If the past year has taught the industry anything, it is that once the "connective tissue" of MCP is established, adoption happens at a velocity that traditional software cycles cannot match.

Conclusion: The Strategic Pivot

We are witnessing the end of the "manual era" of digital advertising. The agents are not coming; they are already here, actively bidding on inventory and analyzing performance in real-time. For agencies and brands, the mandate is clear: the winners of the next decade will not be those who can optimize a campaign the fastest, but those who can most effectively train, monitor, and guide the agents doing the heavy lifting. The "expert" is no longer the person who knows the platform best—it is the person who knows which strategic levers to pull, and when to let the machine take the controls.