The Agentic Shift: How AI is Redefining the Programmatic Ad Landscape

By Digiday Staff

The era of “agentic” advertising is no longer a futuristic concept discussed in white papers or speculative tech forums; it is the current reality of the digital media ecosystem. This seismic shift in how ads are bought, sold, and optimized served as the primary heartbeat of the May 2026 Digiday Programmatic Marketing Summit (DPMS), held from May 6 to 8 in Palm Springs, California.

As the industry grapples with the transition from manual, human-steered bidding to autonomous AI agents, the consensus among attendees—ranging from legacy brand marketers to nimble agency strategists—is clear: the machine is here, but the human remains the architect.

The New Frontier: Why Agentic Ad Buying is Here

The 2026 summit moved beyond theoretical debates to practical implementation. Executives from brands such as Duluth Trading Company and agencies like Butler/Till detailed the integration of AI agents into their daily operations. Unlike traditional algorithmic tools that simply adjust bids based on historical performance, agentic AI functions with a degree of autonomy, interpreting intent, negotiating deals, and executing complex workflows without constant manual intervention.

During behind-closed-door town hall sessions, the atmosphere was one of cautious optimism. Marketers are already deploying these agents to streamline quality assurance (QA) processes, manage complex data sets, and even engage in direct negotiations. The summit’s interactive "Game Show" served as a litmus test for these technologies, forcing teams of professionals to compete against AI-driven seller agents in real-time. The exercise revealed a profound truth: while agents possess the speed and computational power to dominate, they require sophisticated "prompt engineering" and rigorous human oversight to produce meaningful business outcomes.

Chronology of the Shift: From Manual Levers to Autonomous Agents

The industry’s journey toward agentic adoption has been rapid, characterized by three distinct phases:

Phase One (Early Automation): Marketers relied on DSPs (Demand-Side Platforms) to perform basic automated bidding based on simple rules. Transparency was high, and humans controlled the levers.
Phase Two (The Black Box): The rise of proprietary platforms like Google’s Performance Max and Meta’s Advantage+ shifted control away from the buyer. Automation became a "black box," and transparency began to erode.
Phase Three (The Agentic Era): We have now entered a phase where agents possess agency—the ability to act on behalf of the marketer, make strategic trade-offs, and navigate the supply chain with minimal guidance.

The Rulebook: Guardrails for an Autonomous Future

The most pressing conversation at DPMS centered on the necessity of "guardrails." As AI agents gain the ability to make high-stakes campaign decisions, the risk of "runaway automation" has become a top concern for brand safety officers.

Emily Proctor, managing director of data and technology solutions at OMD, emphasized that while AI agents are slated to handle campaign optimization by the second half of 2026, the human-in-the-loop requirement is non-negotiable. "There is still a need for a human to sign off on that work," Proctor noted.

Similarly, Bayer has adopted a pragmatic approach, combining technological speed with strict organizational discipline. Glenniss Richards, senior director of digital media activation at the pharmaceutical giant, confirmed that Bayer utilizes spending caps and mandatory human sign-offs for all AI-recommended changes. "Everyone will need to define those guardrails," Richards stated. The message is clear: autonomy without accountability is a liability.

The "Homework Grading" Challenge

A recurring theme throughout the summit was the necessity of "grading the homework." AI agents are not infallible; they are prone to hallucinations, incorrect CPM reporting, and irrelevant outputs. Scott Ensign, chief strategy officer at Butler/Till, described the agency’s initial experiences with agentic deals as a process of "training wheels."

During the live interactive workshop, teams had to iteratively refine their prompts to communicate effectively with an AI seller agent. The exercise proved that if a prompt is vague, the agentic output is either useless or detrimental to the budget. Agencies are now building internal workflows dedicated specifically to auditing agentic output—essentially, acting as editors for their own AI employees.

Institutional Intelligence: Agencies as "Second Brains"

The true value of AI in 2026 lies in how it handles data. Agencies are now using agents to build "institutional memory" that survives staff turnover.

KSM Media’s Librarian: The agency has developed a “librarian” agent that codifies institutional knowledge, allowing new hires and existing agents to access client-specific insights instantly.
Huge’s Digital Twins: By modeling consumer behavior through digital twins, the agency can simulate marketing outcomes without the risk of real-world brand damage.
WPP Media’s Open Intelligence: This platform stitches together disparate, messy data sets into predictive models that allow for more secure and informed investment strategies.

Transparency: The Industry’s Open Wound

Despite the excitement surrounding AI, the industry remains deeply frustrated by the lack of transparency in programmatic buying. While The Trade Desk has historically been the target of such criticism, Google and Amazon faced intense scrutiny at DPMS.

Marketers argued that as these tech giants inject more AI into their platforms, the "black box" grows deeper. Attendees reported instances where platforms automatically added new video placements without clear notification, leading to wasted media spend. The prevailing sentiment among agency execs is that current platform "solutions" to these transparency issues—which often involve suggestions to spend more money—are fundamentally misaligned with the needs of the advertiser.

Implications for the Workforce: Who Wins?

The question of job security loomed large, yet the narrative has evolved. It is no longer about "AI vs. Human," but rather "The Human Who Uses AI vs. The Human Who Doesn’t."

Ellie Uberto, director of marketing at Duluth Trading Company, noted that while her brand delegates the bidding process to agencies, the core "ethos and voice" of the brand remains strictly in human hands. The role of the programmatic trader is transforming into that of a "systems architect"—someone who designs the environment in which the agent operates.

During a town hall session, an attendee summarized the consensus succinctly: "You’re not going to lose your job to AI. You’re going to lose your job to the guy who knows how to use AI."

Case Study: The Winning Strategy

The summit’s interactive game provided a concrete look at how to win in an agentic landscape. Team 1, the competition winners, succeeded not by giving the AI total control, but by providing specific, high-intent constraints.

By pivoting from a generic request for "300,000 impressions" to a highly specific, value-based request targeting "cord-cutters with HHI of $100K," they were able to negotiate a favorable deal. This confirms that the most successful marketers in the agentic era will be those who can provide the most precise context to their digital counterparts.

Conclusion: The Road Ahead

As the industry looks toward the next quarter, the focus will remain on refining the rules of engagement. Whether through better prompt engineering, stricter governance, or the development of more transparent, proprietary agentic models, the goal is to harness the efficiency of the machine while preserving the strategic intent of the human.

The 2026 Programmatic Marketing Summit proved that we are no longer preparing for the future; we are currently building the architecture of a new marketing paradigm. The winners of this era will be those who balance the relentless speed of the agent with the critical, discerning eye of the human architect.

Key Takeaways from DPMS 2026:

Human-in-the-loop is mandatory: AI agents require constant oversight to prevent hallucinations and brand-unsafe placements.
Prompt Engineering is the new SEO: The ability to communicate effectively with an agent is the most critical skill for a 2026 media buyer.
Transparency remains elusive: Despite technological advancements, the "black box" of major platforms remains a significant pain point for advertisers.
Agencies are evolving: Agencies are shifting from manual executors to curators of proprietary "institutional AI" and systems architecture.
Data Strategy is being rewritten: AI agents are successfully bridging the gap between disparate data sets, enabling better predictive modeling.