Brands and agencies are using AI agents in advertising mainly to turn a written brief into a live campaign, to choose and buy inventory within rules a person has set, and to test agent-to-agent buying directly with publishers. Documented examples include a Movistar connected TV campaign in Spain developed by Havas, Omnicom's agent-to-agent buys for several clients, and what AgenticAdvertising.org describes as the first agent-to-agent media buy, using LG Ads inventory in October 2025. In almost every case a person still sets the brief and approves the commitments that matter.

TL;DR. The ten examples below come from company newsrooms, IAB Tech Lab and trade press. Most brand examples are agency-run campaigns in which agents interpret a natural language brief, select inventory and adjust bids inside client rules, while agencies, holding companies and publishers test direct agent-to-agent buying on shared standards such as IAB Tech Lab's AAMP and the Ad Context Protocol (AdCP). The common thread is that agents draft and negotiate, people approve, and the record of what was agreed matters as much as the buy.

What counts as a brand or agency using an AI agent?

An AI agent in advertising is software that takes a goal, such as a campaign brief, and decides which actions to take to meet it, within limits a person has set. That is different from automation that follows rules written in advance, and from an assistant that only recommends. The examples in this article qualify because a published source describes agents selecting, negotiating or buying media, not just generating copy or reports.

Agent activity sits on a scale. At one end, agents set up and optimise a buy that a trader would otherwise configure by hand; at the other, a buyer's agent connects directly to a seller's agent so that the two can agree terms. The second kind raises a question the first does not: who holds the record of what each agent agreed.

That record is a practical concern. A marketplace such as Alkimi, for example, writes the agreed terms into a DealSheet at the moment of agreement, timestamps it, and lets buyer and seller each retrieve it independently. Several of the examples below point to the same need from different directions.

Brands: which advertisers have run campaigns through AI agents?

In the public examples so far, brands mostly appear as the client of an agency or trading desk rather than as the operator of the agent. The three below are named in their partners' announcements.

Movistar (Telefónica), with Havas: June 2026

In June 2026, PubMatic announced a connected TV campaign in Spain developed with Havas for Movistar, part of Telefónica Group (PubMatic, 9 June 2026). According to the announcement, a natural language brief was processed by the Claude large language model, and agents then selected CTV inventory, applied frequency controls and brand safety settings, and adjusted bids in real time while following the client's own rules. Esther López, Head of Media at Telefónica, said the work showed "that agentic AI can operate with the level of rigour and performance that brands require."

INTERBEV, with Amnet: March 2026

In March 2026, PubMatic and the trading desk Amnet announced what they described as the first agentic advertising campaign in France, a video campaign for INTERBEV (PubMatic, 31 March 2026). The announcement says Amnet's team entered a natural language brief into Claude and deployed the campaign within minutes, with agents translating the objectives into targeting and data segment choices.

Geloso Beverage Group, with Butler/Till: reported March 2026

PubMatic has described its first autonomous campaign in the United States as one run by the agency Butler/Till for Geloso Beverage Group, and cites it as the model for later activations in Europe (PubMatic, 9 March 2026). Butler/Till was also present at the start of agent-to-agent buying: its chief marketing officer, Amanda DeVito, opened the live presentation of the first agent-to-agent media buy in October 2025 (MediaPost, 16 October 2025).

Agencies and holding companies: how are they using AI for programmatic buying?

Agencies are using AI agents for programmatic buying in two broad ways: to compress campaign set-up and optimisation inside existing buying platforms, and to buy directly from publishers through agent-to-agent connections. Holding companies are also helping to shape the standards those agents use.

Omnicom: April 2026

On its first-quarter earnings call in April 2026, Omnicom said it had executed live media buys for several clients through an agent-to-agent framework built around the Ad Context Protocol (Digiday, 29 April 2026). "We've actually executed real media buys for several clients using our agent framework, doing agent to agent buying, which is all in service to shortening the media supply chain," said Paolo Yuvienco, Omnicom's head of AI.

WPP Media: September 2026

WPP Media is among the supporters quoted in IAB Tech Lab's announcement of AAMP 3.0, the latest release of its Agentic Advertising Management Protocols (IAB Tech Lab, 22 September 2026). Michael Palmer, Global Vice President, Emerging Technologies at WPP Media, said the release would aid "our industry's movement from proof-of-concept tests towards the wide-spread use of AI Agents in media transactions." AAMP 3.0 introduces OpenProposal, a specification for how sellers respond to buyer briefs in a format agents can read.

PMG: September 2026

Mike Treon, Head of CTV Strategy at PMG, described agentic planning in the same AAMP 3.0 announcement as a way to evaluate a much broader range of media opportunities when building a plan (IAB Tech Lab, 22 September 2026). "What's new in AAMP 3.0 is that the reasoning travels with the proposal," he said, explaining that a seller can now set out why a package fits the brief, not only what it is. Treon added that this context "is what our planning models need, and now it arrives in a form they can actually read."

Rise, a Quad agency: August 2026

In August 2026, PubMatic named Rise, a Quad agency, among the first agency partners to pilot a governance framework for autonomous campaigns on its AgenticOS platform (PubMatic, 5 August 2026). Under the framework, an account administrator sets budget thresholds, creative approval requirements and approval structures, and an agent that meets a decision beyond its authority escalates it to a human approver. Every agent action is logged, including what it did, when, under which parameters and whether a person was in the loop.

Publishers and platforms: what are they building for agents?

Agent-to-agent buying needs a seller on the other side, so publishers, ad tech companies and industry bodies have run their own tests.

LG Ads, Scope3 and Swivel: October 2025

AgenticAdvertising.org dates the first agent-to-agent media buy to 16 October 2025, describing "real money, real inventory from LG Ads, real AI agents handling creative, targeting, and approval with human oversight" (AgenticAdvertising.org). MediaPost reported that the order was placed live between a buyer agent built by Scope3 and a seller agent at the publisher ad server Swivel, using LG Ads programmatic inventory, with fine-tuning and approvals from the two people running it (MediaPost, 16 October 2025). The advertiser in the demonstration was a brand agent created for a drink that MediaPost noted appeared to be a made-up product.

LiveRamp: November 2025

In November 2025, LiveRamp donated its User Context Protocol to IAB Tech Lab, where it became Agentic Audiences, one component of the AAMP initiative (IAB Tech Lab; IAB Tech Lab on AAMP; PPC Land, November 2025). Agentic Audiences defines how buyer, seller and measurement agents exchange identity, contextual and response signals as compact embeddings rather than long text exchanges (GitHub).

IAB Australia: September 2026

IAB Australia has opened expressions of interest for a seller-side pilot using the AAMP reference implementations, according to IAB Tech Lab's AAMP 3.0 announcement (IAB Tech Lab, 22 September 2026). Jonas Jaanimagi, Technology Lead at IAB Australia, said the aim is "to help Australian publishers test discovery, proposal and approval workflows in a sandbox before any live spend is involved."

What do these examples have in common?

Agents draft and negotiate; people approve. In the Movistar, INTERBEV and Rise examples, agents interpret the brief and act inside rules a person configured, and decisions beyond that authority go back to a human. The first agent-to-agent buy and IAB Australia's planned pilot both place human approval inside the workflow rather than after it.

Shared standards are doing the connecting. WPP Media, PMG and IAB Australia are working with IAB Tech Lab's AAMP, LiveRamp's donation became part of AAMP's audience layer, and Omnicom's buys run on AdCP.

Records matter. Once two agents agree terms, buyer and seller each need to be able to show what was agreed, which is why logging, approval records and audit features recur across these announcements.

Alkimi builds for that point: its Marketplace works within IAB Tech Lab's AAMP framework today, is being built to be AdCP compatible, and keeps human approval at the deal level, with each agreement recorded in a DealSheet held by both parties. In its Agree. Transact. Verify. simulation, 95.3% of deals had mismatched records when each side kept its own, against 0.19% with a shared record.

Frequently asked questions

How are brands using AI agents in their advertising?

Most brands using AI agents today do so through an agency or trading desk, with agents turning a natural language brief into a live campaign and choosing inventory within the client's rules. Named examples include Movistar in Spain and INTERBEV in France, both reported in 2026.

How do agencies use AI for programmatic buying?

Agencies use AI agents to set up, optimise and report on programmatic campaigns, and some are testing agent-to-agent buying directly with publishers. Omnicom said in April 2026 that it had executed live agent-to-agent buys for several clients.

When was the first agent-to-agent media buy?

AgenticAdvertising.org dates the first agent-to-agent media buy to 16 October 2025, using LG Ads inventory. A buyer agent built by Scope3 and a seller agent at Swivel carried it out live, with people approving the steps.

Do AI agents buy media without human approval?

In the published examples, agents act within limits that people set, and decisions beyond those limits are escalated for approval. Approval workflows and action logs are a recurring part of these announcements.

Which standards do agents use to buy media?

The two most cited are IAB Tech Lab's Agentic Advertising Management Protocols (AAMP) and the Ad Context Protocol (AdCP), published by AgenticAdvertising.org. AAMP 3.0, announced in September 2026, adds OpenProposal for the stage between a brief and a buy commitment.

Examples of Brands and Agencies Using AI Agents in Advertising