The useful question is what changes for users, developers or buyers, and whether the announcement stays industry context or becomes something people can actually use.

The manual process of media buying relies on fragmented tools like emails and spreadsheets, delaying decisions as inventory and pricing shift. Agentic workflows aim to automate this coordination, allowing teams to focus on strategy and creativity. However, automation requires a shared foundation—standards like IAB Tech Lab’s Agentic Advertising Management Protocols (AAMP) provide a common language for inventory, audiences, and transactions, enabling agents to interact consistently across the industry.

Databricks has built an example of autonomous buyer and seller agents that discover each other, agree on price, and close deals using the open-source IAB Tech Lab SDK. The system runs on Databricks’ platform, which manages state, identity, governance, and models for each party. Agents from different companies interact only through the shared protocols, while their operational logic remains self-contained within their respective platforms.

The buyer agent is structured as a hierarchical crew of specialist agents, including a Portfolio Manager, Channel Specialists, and tactical workers. These agents use Databricks Foundation Model APIs and can switch between models like Claude via the Unity AI Gateway. Transactional data is stored in Lakebase, Databricks’ serverless Postgres, enabling fast, scalable reads and writes for agent interactions while maintaining governance through Unity Catalog.

Authentication and trust are handled natively via OAuth and a registry that assigns trust tiers to buyers, controlling access to pricing and transaction capabilities. The system logs agent reasoning and decisions using MLflow tracing, ensuring transparency and auditability. In a test campaign, a $200,000 budget was split across CTV and Linear TV, with agents mapping audiences, discovering sellers, pricing inventory, and executing deals based on predefined rules.