Alexandra Balashoiu, Vice President of Product, Onetag
Agentic AI has quickly become one of advertising’s most talked-about developments. The promise is compelling. Give an AI system an objective, let it navigate complexity, and allow it to take on work that would otherwise require constant human intervention.
That idea has an interesting resonance as The Odyssey returns to popular culture through Christopher Nolan’s film adaptation. Homer’s epic is a story about navigating uncertainty, with Odysseus repeatedly forced to make difficult decisions on his journey home. Advertising is entering its own period of technological navigation as agentic workflows begin to reshape how programmatic operates.
The opportunity is significant, but so is the need for direction. How can AI help the industry move through complexity without creating more of it?
When the sirens start singing
The excitement around agentic AI can sometimes resemble the story of the Sirens. Their song was irresistible, drawing sailors towards something they could not safely reach. AI’s current wave of promises can have a similar pull.
Every new capability can sound like an answer to another operational problem. Fully autonomous campaign planning. Agents negotiating deals. AI making decisions across the media supply chain.
Some developments will prove genuinely useful. Others could add another layer of technology to an ecosystem that already has plenty of complexity.
An agent needs a clear purpose, access to the right information and defined limits around what it can do. Advertising decisions can also involve commercial commitments, privacy requirements, and legal or ethical considerations. Those responsibilities remain when an AI system takes over a task.
Knowing where to steer
There is already a useful distinction between automation and agentic workflows.
Programmatic advertising has relied on automation for years. Deal Sync, for example, can automatically move a deal between platforms using predefined fields and rules. If an API fails or something unexpected happens, the process can stop and return an error.
An agent could approach the problem differently. It could investigate what went wrong, gather information from connected systems, and determine whether another route could resolve the issue. The underlying automation still has a role, but the agent adds reasoning around it.
The same principle applies to campaign planning. A buyer could provide a brief and desired business outcome, with an agent developing an activation strategy for approval. During the campaign, it could monitor signals and surface opportunities that would otherwise require teams to work across multiple dashboards.
A seller-side agent could similarly assess inventory and demand signals to identify opportunities worth packaging into a deal.
The value lies in reducing operational work while leaving important decisions under human control.
Guardrails are part of the journey
Odysseus had to understand the dangers along his route. Agentic advertising needs a similar level of preparation.
People still need to establish campaign objectives, success measures, constraints, and risk tolerance. They also need to provide the knowledge bases and operational guidance that give agents the context to make reliable decisions.
This becomes particularly important as agents move from recommending actions to carrying them out.
IAB Tech Lab’s AAMP 2.3 reflects this shift, introducing approval gates and pricing verification that help determine when an agent can proceed independently and when human authorisation is required.
These controls provide the boundaries within which autonomy can be trusted, giving organisations greater confidence in delegating operational decisions.
A shared map for the journey
An agent may be capable of reasoning about a campaign, but it still needs access to the systems where relevant information lives.
Programmatic advertising solved a similar problem through shared standards such as OpenRTB. Buyers, sellers and technology platforms could communicate using a common language, allowing decisions to happen across the ecosystem at enormous speed.
Agentic workflows will require the same interoperability. An agent cannot navigate a fragmented ecosystem efficiently if every platform speaks a different language.
This is where standards such as AAMP and AdCP are becoming increasingly relevant. The industry is still working through how higher-level agentic advertising workflows should operate, but beneath those discussions sits a more fundamental connectivity problem.
The Model Context Protocol, or MCP, provides a common interface through which agents can discover and access tools and data. Its importance lies in creating a consistent connection between an agent and the systems it needs to use, regardless of how the wider agentic landscape develops.
That gives the industry room to refine agentic workflows without forcing every platform to solve connectivity from scratch.
Keeping sight of the destination
The most useful agentic workflows will help people navigate complexity more effectively. Their success will depend on reliable information, appropriate permissions and shared standards across the ecosystem.
Programmatic advertising became scalable because the industry agreed on common ways for systems to communicate. Agentic advertising faces a similar moment. Collaboration across the ecosystem will matter as much as innovation within individual platforms.
There will inevitably be plenty of noise around the journey. Some ideas will prove transformative. Others may become distractions.
The industry’s task is to build infrastructure that lets agents do useful work while keeping people accountable for the decisions they make.
Perhaps that is the more useful lesson from Odysseus. Reaching the destination depends on being able to navigate the journey, recognise when the course needs to change, and know which voices are worth following.
Originally Published on: LinkedIn
