The newest shift in how media is traded: AI agents that buy and sell inventory on behalf of advertisers and publishers. This track builds the mental model before the deep dives. See where agentic buying sits next to header bidding, learn the agent protocol stack (MCP, A2A, gRPC), compare the two standards efforts (AdCP and the IAB Tech Lab AAMP roadmap), and understand how agents discover and trust each other. Everything here builds on the supply chain and auction knowledge from Tracks A to F.

1. L2G1The shift to agentic advertising

   Prereqs: A4

   Objectives

   • Place agentic advertising on the same timeline as the waterfall and header bidding: the next step in automating how media is bought and sold.
   • Distinguish a buyer agent (acting for an advertiser) from a seller or sales agent (acting for a publisher).
   • Explain what stays the same (the auction, ads.txt, sellers.json, ad servers) and what changes (a machine, not a person, runs the buy).

   Key concepts

   Agentic advertisingBuyer agent vs sales agentAgent-to-agent transactionHuman-in-the-loop approval

   Assessment

   In your own words, describe one media buy as a conversation between a buyer agent and a sales agent, and name which existing ad-tech pieces each agent still relies on.

   Read the lesson →

   LEARN

   Sets up Track H (AdCP, the protocol the buy actually runs on) and Track I (the IAB Tech Lab agentic standards).

2. L3G2The agent protocol stack: MCP, A2A, and gRPC

   Prereqs: G1

   Objectives

   • Explain MCP (Model Context Protocol): how an AI model calls tools exposed by a server, and why ad systems expose their inventory as MCP tools.
   • Explain A2A (Agent-to-Agent): how two independent agents negotiate a task without a human in between.
   • Place gRPC as the high-throughput transport some agentic systems use, and see that these are general protocols an ad-specific layer sits on top of.

   Key concepts

   MCP (Model Context Protocol): tools, clients, serversA2A (Agent-to-Agent): tasks and negotiationgRPC as a transportGeneral protocol vs ad-specific layer

   Assessment

   Match each protocol to its job: a model invoking a publisher tool, two agents agreeing on a deal, and a fast binary transport. Explain why none of them is advertising-specific on its own.

   Read the lesson →

   ANALYZE

   Proposed AdCP request validator: paste an MCP tool call and see the fields an agent must send.

3. L3-L4G3Two standards, one goal: AdCP and IAB Tech Lab AAMP

   Prereqs: G1

   Objectives

   • Summarize the Ad Context Protocol (AdCP): an open, task-first layer over MCP and A2A, with founders across the sell side and a Prebid.org reference implementation.
   • Summarize the IAB Tech Lab agentic effort (AAMP): a roadmap that extends existing IAB standards with agent SDKs, a registry, and real-time execution.
   • Explain how the two overlap and differ, and why both build on the standards from Tracks A to F rather than replacing them.

   Key concepts

   AdCP (Ad Context Protocol)AAMP (Agentic Advertising Management Protocols)Open protocol vs standards-body frameworkBoth extend ads.txt, sellers.json, and OpenRTB

   Assessment

   Write a short brief for a publisher: what AdCP is, what the IAB Tech Lab AAMP roadmap is, and why adopting agentic demand does not mean throwing away their header-bidding setup.

   Read the lesson →

   LEARN

   Track H takes AdCP apart in detail; Track I does the same for the IAB Tech Lab standards.

4. L3-L4G4Trust, identity, and discovery for agents

   Prereqs: G3, C2

   Objectives

   • Explain how a buyer agent discovers which sales agents are authorized to sell a property, using adagents.json as the agentic counterpart of ads.txt.
   • Connect agent authorization to the supply-chain transparency you already know: the same questions of who is allowed to sell, and how deep the chain runs.
   • Describe where human approval, brand safety, and fraud control sit in an otherwise automated flow.

   Key concepts

   adagents.json (authorized sales agents)Agent discovery and registriesAuthorization as the agentic version of ads.txtHuman-in-the-loop, brand safety, and fraud control

   Assessment

   Given a property that lists two sales agents in adagents.json, explain how a buyer agent decides which to trust, and what would make an unlisted agent a red flag.

   Read the lesson →

   ANALYZE

   The same trust question as the live audit: who is authorized to sell this inventory. Run an audit and read the authorized-seller checks, then picture the agentic equivalent.

   Run it live →

   PRACTICE

   Proposed lab: author an adagents.json that authorizes a sales agent for a property.