Learn Guide 11
AI agents in media buying
AI agents are starting to buy and sell media. Here is what they actually do, what they do not, and where people stay in charge.
What an AI agent is, in media
An AI agent is software that acts for one side of a media deal. A buying agent works for an advertiser or an agency; a selling agent works for a publisher. Each reads documents, calls tools and writes answers, within the rules its side has set. Built on AdCP, the open Ad Context Protocol, any buying agent can talk to any selling agent.
What they do
- Read a brief. Objectives, markets, dates, budget, in plain language.
- Build a plan. The selling agent proposes lines from the publisher’s packages and prices.
- Negotiate. Within the buyer’s budget and the seller’s floors, never outside them.
- Compare. The buying agent lays out every proposal in the same columns.
- Report. Delivery is read against the confirmed plan, shortfalls first.
What stays with people
Confirming a plan, committing a budget, accepting a make-good: every step that commits money or inventory waits for a person. The agents prepare the work; people decide. That is how Oderra’s agents, Fanera and Skopa, are built.
What they change
A brief that took a sales team days is answered while the buyer is still planning. Mid-sized budgets become worth answering. Publishers without a sales team can sell; buyers can reach publishers that are not on the exchanges. This is the third way between direct sales and programmatic, and it is described in Direct, programmatic or agentic?