TL;DR: AI agent negotiation in programmatic advertising means structured, mandate-driven deal-making between buy-side and sell-side agents, not simply faster bidding. This guide explains how it works and what US media buyers need to understand before adopting it.
Negotiation Is Not the Same as Bidding
In a real-time bidding auction, a DSP submits a bid in milliseconds. The process is automated but not negotiated. There is no back-and-forth between buyer and publisher, no discussion of deal terms, and no shared record of what was agreed. The winner of the auction simply pays the clearing price.
Agent-to-agent negotiation is different. A buy-side agent and a sell-side agent exchange deal parameters directly: price floors, inventory categories, targeting conditions, brand safety requirements, volume commitments. When both agents agree, the result is a deal record that both parties hold and can audit. This is closer to how a direct insertion order works than how a programmatic auction works.
How the Negotiation Process Works
The typical flow has five stages.
Mandate setting. The buyer defines the parameters within which their agent is authorised to negotiate. This includes maximum CPM, acceptable inventory types, required brand safety standards, and any exclusions. The agent cannot act outside this mandate.
Discovery. The buy-side agent identifies available supply. In a platform like Alkimi, this happens through the marketplace's inventory layer, where publisher agents have made supply available with their own deal parameters.
Negotiation. The buy-side agent proposes deal terms. The sell-side agent accepts, rejects, or counter-proposes. This exchange happens through a structured protocol, not free-form conversation. The Agent Communication Protocol (AdCP) defines the messaging format for this exchange.
Deal record. When both agents agree, the terms are recorded in a shared deal document, known as a DealSheet in Alkimi's implementation. Both the buyer and the publisher hold a copy. The marketplace holds the audit trail.
Execution. Campaigns run against the agreed deal parameters. Performance data is logged against the deal record.
What the Buyer Controls
A common concern among US agency trading desk leads is that agent negotiation means giving up control. The earned autonomy model addresses this directly.
In this model, an agent's authority is earned and bounded. An agent starts by observing and recommending. Over time, as it demonstrates judgement that matches the buyer's parameters, it is granted authority to act within defined boundaries. The buyer retains the ability to approve deal terms above a threshold, pause agent activity, set hard limits on spend, and audit every deal the agent has agreed.
The agent handles the operational work. The buyer retains decision authority over anything consequential.
What US Media Buyers Need to Check Before Adopting
Before committing to any agentic buying platform, US media buyers should verify the following.
Mandate controls. Can you define the exact parameters within which your agent operates? Can you change those parameters in real time?
Deal record access. Can you see, at any time, what your agent has agreed and on what terms? Is that record held by the marketplace only, or do you hold a copy independently?
Audit trail. Is there a timestamped log of every negotiation step, available to you for review?
Standard compliance. Is the platform building to the Agentic Advertising Marketplace Protocol (AAMP) or the AdCP standard? Platforms aligned with open standards are more likely to integrate with the rest of your stack.
Override capability. What happens if you need to stop or modify agent activity mid-campaign? How quickly can you intervene?
What This Means for Agency Trading Desks
For US agency trading desks, agent-to-agent negotiation changes the operational model. Instead of managing individual line items and bid strategies, the trading desk sets mandates and monitors agent performance against them.
This requires a different kind of oversight: less time spent on individual deal mechanics, more time spent on mandate design and performance review. Trading desks that establish clear mandate frameworks now will be better positioned as agentic buying becomes a standard part of the programmatic toolkit.