Agentic AI in Programmatic: Which Way Now?

By Leo Oliveira, Director, Client Strategy

Agentic AI has quickly become one of the defining conversations in media buying – and one of the least understood. For some, it represents a fundamental shift in how media will be planned and bought. For others, it remains more marketing narrative than market reality.

Based on our work for advertisers, covering agencies, technology providers and publishers, our view is more measured. Agentic AI is real, momentum is building, but the market remains in its infancy. As with most structural shifts in advertising, the technology itself is only part of the story. The more important question is how value moves across the ecosystem and whether agentic AI accelerates or reshapes that movement.

AI in media buying is not new, but its application is

AI has been embedded in media buying for well over a decade. Rules-based automation and machine learning have optimised bidding, pacing and audience selection since the early 2010s. Today, products such as Performance Max and Advantage+ are used by around 90% of leading performance marketers, although the depth of adoption varies.

What is changing is not the presence of AI, but the level of autonomy it is being given. Agentic AI can independently interpret a brief, determine an appropriate course of action and make decisions within predefined parameters, all in pursuit of a specified objective rather than a fixed set of rules.

That may appear to be an incremental change, but it fundamentally alters the questions advertisers should be asking. The discussion moves beyond whether a platform optimised effectively to whether the right objectives were set, who controls the agent, and who remains accountable for the decisions it makes.

Protocols are coming and advertisers will have a say in them

The industry has quietly been developing a new generation of protocols and standards that allow AI agents to connect with tools, data and, increasingly, with one another.

Within advertising, the Ad Context Protocol (AdCP) has the greatest potential to reshape the programmatic ecosystem. Unlike technical standards such as the IAB’s ARTF, AdCP has been designed specifically for advertising workflows. It provides buyer and seller agents with a common language through which they can plan, buy and measure media directly, agent to agent.

While real-time bidding established a protocol for transactional execution, AdCP begins to establish a protocol for intent-driven decision-making. The emphasis shifts from speed and volume towards objectives and outcomes.

Advertisers should pay close attention because AdCP has the potential to redefine who sits within the value chain and who captures the value created. A more direct, agent-led route could reduce the role of intermediaries that currently extract significant value. For now, however, adoption is being driven primarily by independent sell-side providers and publishers, organisations with the most to gain from a shorter supply chain. The major walled gardens have yet to participate, which may ultimately determine the pace at which agentic buying scales.

Results so far are encouraging, but they come from controlled environments

Several live campaigns have produced results worth noting. One agency reported an 80% reduction in campaign activation time through automated deal-making with an SSP. Another campaign, executed without a traditional DSP by combining buy-side and sell-side capabilities within a single vendor, reported an 82% reduction in supply-chain costs – although much of that improvement reflected the removal of DSP fees rather than agentic efficiency alone. On the buy side, an advertiser working alongside a specialist partner and its in-house agency reduced costs sufficiently to reinvest savings back into working media.

These results deserve attention, but they should also be viewed in context. Most remain self-reported, few have been independently audited, and almost all originate from tightly controlled pilot environments rather than open-market deployment. One publisher recently described the volume of live agentic campaigns flowing through its platform as being “measured in dollars.” That remains the reality today.

The ambition, however, is significantly larger. The IAB expects agentic AI to become a major driver of US advertising growth this year, while UK forecasts suggest AI-driven buying could account for around one-third of digital advertising by 2030. Expectations are running well ahead of deployment. Closing that gap (not proving the technology) is now the industry’s primary challenge.

What this means for advertisers

Three implications stand out.

Ways of working: Agentic AI is most effective at compressing repetitive execution tasks including campaign setup, optimisation, pacing and reporting. These are high-volume operational activities where automation can deliver meaningful productivity gains. Much of the value created, however, is likely to be reinvested into oversight, governance and strategic decision-making rather than removed entirely.

Agency models: As execution becomes increasingly automated, value shifts from labour to orchestration. The conversation is no longer centred on who performs the work, but on who owns the agent, controls the decision-making process and accepts accountability for outcomes. That transition will require different commercial models between advertisers, agencies and technology partners, with traditional time-and-materials approaches becoming increasingly difficult to justify.

Governance: Delegating decisions to autonomous agents introduces the classic principal-agent problem at machine speed. Advertisers need confidence that an agent is acting in their interests, within clearly defined boundaries and with appropriate transparency. Governance frameworks, audit rights and accountability mechanisms should therefore be built into commercial agreements rather than assumed.

Handled well, the opportunity is significant: lower operating costs, faster execution and greater control over programmatic investment. But early advantage will not come from adopting agents first. It will come from establishing the governance, commercial models and operating principles that allow them to be deployed with confidence.

Where to start

• Do you know where AI is already operating across your media ecosystem – and who controls it?
• If you are planning a pilot, have you established clear governance, ownership and operational guardrails?
• How will you measure both efficiency and marketing effectiveness?
• Do you understand the agentic capabilities of your agencies and technology partners today—and their roadmap for tomorrow?
• Is your commercial model fit for purpose in an increasingly agentic market?

If you are considering what agentic AI means for your programmatic investment, operating model or commercial model, we’d be delighted to discuss how mediasense can help.

Talk to our team