A communication standard that lets one AI agent hand off work to another AI agent, across companies, without a human relaying the message in between. In advertising, it is what makes it possible for a buyer’s agent and a seller’s agent to negotiate a media deal directly.
Also known as A2A
Agent-to-agent protocol, usually shortened to A2A, defines how one autonomous AI agent discovers, authenticates with, and exchanges tasks with another agent it doesn’t control. Instead of a person copying output from one AI tool and pasting it into another, A2A lets agents call each other the way two web services call each other’s APIs, with structured requests and responses instead of a chat transcript.
In advertising specifically, A2A style handoffs are what sit underneath frameworks like the Ad Context Protocol: a buy-side agent representing an agency’s client can query a sell-side agent representing a publisher, get back available inventory and terms, and execute a transaction, all without either party’s humans typing anything in real time.
A2A is the layer that turns “we use AI tools” into “our systems transact with other companies’ systems,” which is a different level of responsibility.
It moves trust from people to configuration. When an agency’s agent can transact with a publisher’s agent directly, the important safeguards, budget caps, brand safety rules, approval thresholds, have to be built into the agent’s setup, not enforced by a human double-checking every line item.
It compresses timelines the agency has to plan around. Live tests from MiQ’s Sigma trading agent and NBCUniversal’s agent-sold media show deals moving from discovery to execution in a fraction of the time manual trading takes, and campaigns built on human-speed assumptions will feel slow by comparison.
It raises the stakes of a misconfigured agent. An agent that can autonomously transact with another company’s agent can also autonomously make an expensive mistake with another company’s agent, so testing and guardrails matter more than they did when a human was the last check before anything went live.
An agency running upfront TV and streaming buys for a CPG client configures its media-buying agent to handle A2A negotiations with two broadcasters’ sell-side agents for standard inventory under a preset budget and frequency cap. The agent completes three routine placements overnight that would normally wait for a trader’s morning. Anything outside the preset rules, a new broadcaster, an unusual rate, a bundled sponsorship, gets flagged and routed to the agency’s media director instead of executed automatically.
The automations and agents module of the workshop teaches you how to build AI workflows that compress the busywork without taking the craft out of the studio.