Agentic Advertising Moves Closer to Real-World Adoption: Three Updates from IAB Tech Lab 

Share this article

Agentic advertising is beginning to take shape within the systems used to plan, buy and measure digital media. Three recent updates from IAB Tech Lab address different parts of that work: managing media transactions, deploying agents within the real-time bidstream, and using AI-generated audience signals in programmatic advertising. 

The updates provide a clearer view of how autonomous agents could operate across the advertising supply chain while supporting interoperability, governance, privacy and accountability. 

“For Canada’s digital advertising industry, the focus now needs to be on how agentic systems operate in practice: what they are authorized to do, which standards guide them and where human accountability remains essential,” said Sonia Carreno, President, IAB Canada.

 

1. AAMP 2.3 strengthens the foundation for agent-led media buying 

The Agentic Advertising Management Protocols, or AAMP, provide a common framework through which buyer and seller agents can discover inventory, negotiate deals and carry out media transactions. 

The release of AAMP 2.3 brings several improvements designed for enterprise use. These focus on three areas: interoperability and scale, governance and trust, and transaction-ready audiences. 

AAMP can now operate within common enterprise infrastructure and supports different AI models, including OpenAI, Google Gemini and other compatible services. New storage options allow agents to run across multiple instances, while integrations with Google Ad Manager Reporting and Meta Ads expand the tools available to them. 

The update also introduces stronger controls for transactions involving real media spend. A vendor approval gate can filter out unapproved partners before an agent considers them or completes a transaction. Pricing provenance allows organizations to trace where a CPM originated and prevents an agent from inventing a price when the necessary data is unavailable. 

AAMP 2.3 converts governance requirements into technical controls. This is particularly important when agents are authorized to manage budgets, evaluate vendors or negotiate commercial terms. Organizations need to be able to define those permissions, enforce them and review how each decision was made. 

The approach reflects IAB Tech Lab’s broader emphasis on shared standards. As Anthony Katsur, CEO of IAB Tech Lab, explained when the organization released its agentic roadmap: 

“Open, interoperable standards are what make that possible, and our focus is on scaling it responsibly.” 

AAMP 2.3 also connects agentic audience information with existing programmatic workflows. Audience vector embeddings can now be included in OpenRTB bid requests and supported through Prebid, giving buyer and seller agents a standardized way to describe and match audiences. 

2. ARTF v1.0 brings agents into the real-time bidstream 

While AAMP focuses on managing advertising transactions and workflows, the Agentic Real-Time Framework, or ARTF, addresses activities that must happen within the programmatic bidstream. 

ARTF v1.0 is the first stable release of the framework and is already being used in production. It allows companies to deploy specialized agents or other functionality inside a demand-side platform, supply-side platform or partner environment. 

The container-based model is designed to keep data within the host organization’s infrastructure. Each deployed agent operates in an isolated environment with no external network access by default. A partner can therefore provide its technology without requiring the platform’s data to leave its network. 

Potential applications include: 

  • Privacy and consent-signal verification 
  • Fraud detection  
  • Specialized bidding and bid modification
  • Audience augmentation  
  • Deal activation
  • Measurement and viewability calculations
  • Vector-embedding processing  

Version 1.0 also introduces orchestration between ARTF nodes. One node can call other nodes, including those operated by different organizations, and combine their responses before modifying a bid request. 

The framework uses established infrastructure such as Kubernetes, Docker, Protobuf and gRPC, with the Model Context Protocol providing an interface for agent interactions. 

ARTF gives the industry a common foundation for bringing agentic capabilities into real-time advertising environments. Its local deployment model will be especially relevant where data governance, security and processing speed are central concerns. 

3. Agentic Audiences makes AI-driven targeting operational

The third update focuses on how agents and programmatic systems can exchange audience intelligence. 

Agentic Audiences uses vector embeddings, which are numerical representations of interests, behaviours, content or other information. Embeddings allow systems to identify semantic similarities that may not be visible through exact keywords or manually defined categories. 

For example, a person engaging with content about electric vehicles and renewable energy could be matched with relevant advertising even if the same words do not appear in the campaign’s targeting criteria. Similar techniques can support contextual targeting, creative matching, lookalike modelling, brand suitability and performance analysis. 

The Agentic Audiences specification does not prescribe how an embedding must be created. It acts as a standardized envelope that identifies the embedding and the model used to generate it. Buyer and seller systems using the same model can then interpret and compare the information. 

Two additions make this approach usable within current programmatic infrastructure. An OpenRTB community extension defines the fields needed to pass embeddings in a bid request, while a Prebid module enables publishers and vendors to begin implementing and testing the standard. 

An open-source scoring tool also demonstrates how a bidder can extract an embedding, compare it with campaign targeting data and calculate the strength of a match. 

Vector embeddings may help systems exchange audience signals without exposing the underlying raw identity data. However, they do not eliminate the need for identifiers in every part of advertising. Embeddings can support decisions about likely relevance, while attribution still requires a reliable way to connect an impression or interaction with an outcome. 

Organizations will need to understand where each approach is appropriate and how it fits within their privacy obligations. 

What these developments mean for Canada

For Canadian advertisers, agencies, publishers and technology companies, these updates show how agentic advertising may begin entering existing operations. 

AAMP establishes protocols for agent-led planning and transactions. ARTF provides a framework for secure, real-time deployment. Agentic Audiences introduces an interoperable method for AI-driven audience matching. Each addresses a different layer of the advertising process, while all three build on established programmatic standards and infrastructure. 

Several questions will remain important as organizations evaluate these technologies: 

    • What decisions can an agent make independently?

    • Which actions require human review or approval?

    • How will pricing, audience and vendor decisions be documented?

    • What data does an agent receive, and where is that data processed?

    • How will organizations assess privacy, security and regulatory requirements?

    • Who is accountable when an automated decision produces an unintended result?  

    Canadian organizations can begin by identifying contained use cases, establishing clear governance and testing the available open-source tools. Early testing will help teams understand the operational requirements and determine where agentic systems offer measurable value. 

    These releases provide practical building blocks for agentic advertising. Their wider adoption will depend on whether organizations can use them to improve media operations while maintaining transparency, control and accountability. 

    Check out IAB Canada’s Agentic and AI Resource Centre for more information. 

    Related Articles

    Corus (NextScreen) | Shark Ninja

    Corus delivered a first-to-market shoppable innovation in Canada that extends brand storytelling through a custom second-screen experience - NextScreen. Built to support Linear, Connected TV, and Web, the format enables audiences to shop, engage, or learn more directly from video creative, reducing friction and helping drive conversion at the moment of inspiration. By combining upper-funnel impact with direct response capability, NextScreen allows brands to connect attribution to specific creative executions, supported by enhanced analytics and creative-level reporting through its custom portal.

    Captivate | Dell Product Launch

    Dell’s programmatic campaign delivered measurable gains in ad recall (68%), consideration (53%), and customers action (2 in 5). Captivate’s programmatic guarantee capabilities highlights DOOH’s ability to deliver predictable and reportable results. Watch the video here.

    Read the Case Study