The ad tech market at Cannes Lions 2026 is running two parallel conversations: how much AI to put inside media buying platforms, and how much is too much. A France-based startup called Concord is planting its flag on the “too much” side, and just closed a $3 million seed round on the thesis.
Concord describes itself as an agentic media execution platform. The distinction it draws is between execution and optimization: Concord’s agents handle campaign activation across Google, Meta, and Amazon via direct API integrations, but they do not determine what to optimize. If an algorithm determines that mobile ads outperform desktop, that insight flows to the agent for execution. The agent does not run the analysis. Founder Nathan Venezia argues that conflating execution with decisioning is why most agentic media platforms underperform: the more an agent tries to do, the less precisely it does any single thing.
The practical architecture involves direct integrations with each platform’s full API, including hundreds of endpoints, rather than wrapping those APIs in Model Context Protocol. Venezia is explicit that MCP “most of the time does less” because it adds abstraction where precision is needed. Concord’s model suggests that the agentic media buying category is splitting into two schools before it has fully formed: platforms that hand AI the full brief and hope it optimizes, and platforms that give AI a narrow, well-defined job and measure it on execution quality alone.
The deeper signal here is not about Concord specifically. It is about what the next 18 months of platform evaluation will look like for media buyers. As agentic buying moves from experimentation toward operational infrastructure, the performance question sharpens. OpenAI’s move into self-serve advertising adds a new supply-side variable to the mix; platforms that can execute cleanly across an expanding, fragmented inventory set will matter more, not less.
Source: Concord (PR Newswire)
