By Nicolas Bidon, CEO, Ogury
Advertising is entering a phase where AI affects not just tools, but how systems are organized. Beyond campaign optimization, AI is changing how agents increasingly handle discovery, planning, and execution, and, ultimately, how value is created. The question is no longer whether agentic systems will reshape adtech, but whether companies are preparing their data and infrastructure to be usable by them.
Agentic systems are not about more data, but better context
Agentic AI is often framed as an extension of big data, which is misleading. Agents do not operate well on raw, unstructured datasets. Without context, they do not interpret; they approximate. And approximation, in advertising, leads to inconsistency and inefficiency.
The issue is less about volume and more about how data is structured, translated, and made actionable. Well-structured datasets can be navigated efficiently by agents, but without the right contextual layer, even large amounts of data become difficult to use effectively. This requires building a layer that connects raw data to business logic, giving agents the context they need to make reliable decisions. This is where many current approaches fall short. Feeding more data into systems without improving how it is understood does not increase performance. It amplifies noise.
From interfaces to agent-to-agent interactions
For years, digital advertising has been built around human-operated interfaces: dashboards, DSPs, reporting tools. Agentic introduces a different model, where users can define an objective in natural language, and systems can translate that into audience selection, media planning, and activation.
The traditional separation between planning and execution starts to fade, as systems move from sequential workflows to continuous decision-making. What used to require multiple steps across teams can now be handled within a single flow, driven by an initial objective. As a result, decision-making shifts away from manual inputs and interface navigation, toward orchestration where the role is no longer to operate tools, but to guide outcomes.
Interactions between companies are already evolving. Emerging frameworks and integrations are enabling agency-side agents to interact directly with adtech platforms’ systems to define audiences, activate campaigns, and optimize performance.
This does not remove humans from the process, but it changes their role from operating tools to defining objectives, constraints, and strategy.
MCP: from technical layer to commercial interface
To enable this, a new type of infrastructure is emerging. Model Context Protocol (MCP) servers act as standardized access points to data and capabilities. In an agentic environment, they are not just technical connectors: they become the interface through which systems discover, evaluate, and activate partners.
MCP changes both how data is accessed and how relationships between partners are structured. Instead of relying on predefined queries, agents can navigate available resources dynamically, selecting and combining data based on context rather than fixed inputs. At the same time, this model has commercial implications. If agencies orchestrate multiple partners through their own systems, they gain greater control over how supply is evaluated, selected, and activated. This shifts part of the decision-making power away from individual platforms and into the hands of those who control the orchestration layer.
Frameworks such as AdCP or AAMP aim to structure these interactions. However, it is still unclear how standardized the ecosystem will become and how much fragmentation will persist. But whether MCP is standardized or not, it will function as a new layer of interaction between buyers and sellers.
Quality becomes the real differentiator
There is a concern that making data accessible to agents will commoditize it, while the opposite is more likely. Agents accelerate comparison and evaluation. As a result, low-quality or generic data becomes easier to filter out. What remains is data that delivers consistent, measurable value.
Agentic systems reward quality over availability. Data that was previously used because it was easy to access may lose relevance. Data that is differentiated, well-structured, and contextually meaningful becomes more visible and more valuable.
How to start preparing for agentic systems
For adtech companies and agencies, the transition does not require a full rebuild, but clear priorities.
- Build a semantic layer first
Before adding AI capabilities, ensure data is structured in a way that connects signals to business meaning. Without this, agents will not produce reliable outputs. - Rethink how data is accessed
Traditional APIs assume the user knows what to ask for. Agentic systems rely on layers such as MCP to make data directly usable in context.
- Treat MCP as a strategic layer
Whether through proprietary or standardized approaches, MCP will become a primary interface for activation. It should be considered not only as infrastructure, but as part of the commercial strategy. - Prioritize connections, not scale
Each integration carries a cost. It is neither realistic nor efficient to connect to hundreds of partners. Focus on high-value integrations that can operate autonomously and reliably. - Measure real usage and outcomes
Agentic systems require continuous iteration. Observability — understanding how systems are used and how outputs perform — is critical to improving them over time.
The move toward agentic systems will not happen all at once, but many of its components are already in place. As adoption accelerates, companies that have structured their data for context, not just storage, will be in a stronger position to integrate, collaborate, and compete. The transition is not theoretical. It is already beginning to redefine how advertising works.
