Scope3 launches platform to manage and trade AI agents
Scope3 has launched a platform for partners to build and sell agentic media products that are “efficient and sustainable by design”.
The Agentic Media Platform is accessible to publishers, adtech platforms, curators and agencies, which can use it to develop AI agents and custom algorithms for media buying.
It hosts a centralised hub to create and manage agentic media products and access specialist AI agents.
The platform includes proprietary Scope3 media quality data on which agentic models can be based. This includes data on attention potential, “problematic placement” detection (defined by Scope3 as ad placements no user actually sees) and placement-level viewability.
Additional data partners on launch include LiveRamp, Classify, Sy.nexus and Compliant.
“What we’re doing is going into the plumbing and into the infrastructure of the adtech ecosystem, where there kind of aren’t humans anyway right now, making that better, providing transparency and adding a human into the loop,” said Scope3 co-founder and chief operating officer Anne Coghlan.
In a conversation with The Media Leader, Coghlan argued that because Scope3 retains control over what data inputs are added into AI agents developed and sold on its platform, it can guarantee that agents are designed to eliminate wasteful spend on made-for-advertising (MFA) sites, fraud, invalid traffic (IVT), non-brand-safe and climate-risk inventory by default.
“We’ll have partners that can create agentic media products using agents that are already in the platform,” she explained. “We’re vetting those third-party agents.”
A number of companies have already signed deals with Scope3 to become launch distribution partners, including Index Exchange, Equativ and Media.net. Additionally, MiQ, Elcano and Azerion are expected to initiate pilots using the platform in early April.
‘Show me the money’: Will business models be ‘redefined’ by AI agents?
Brand Standards agent
The first agent built on top of the platform, available on launch, is dubbed Brand Standards.
It purports to reimagine brand safety and suitability efforts by automating decision-making around compliance, responsibility and effectiveness based on models such as Ebiquity’s Effective and Responsible Advertising Curriculum.
Brands can customise models by feeding the AI agent brand values, standards and objectives to ensure all purchased media adheres to specific brand-safety needs.
Similar to Reach’s Mantis tool, the agent is capable of understanding the actual content in and context of a given webpage, allowing it to more accurately assess whether the page itself is suitable for specific brands, based on their inputs.
“[It] is a very precise, controlled job for an agent to be doing,” Coghlan (pictured, left) argued when asked whether AI agents could reliably decide ad placements given many large language models are prone to inaccuracy.
“It’s not like asking [AI] for facts or to write a poem or draw a picture. It’s literally doing checking.”
Notably, both the brand and the publisher or platform where the ad delivers via the Brand Standards agent can verify whether content meets or fails the brand’s standard, ensuring humans can check the agent’s decision-making and refine its accuracy.
The agent has also been made available on other agentic platforms, including both open web and closed ecosystems.
Integrated launch partners for the tool include Ebiquity, as well as all major demand-side platforms, and Dotdash Meredith. Multiple unnamed supply-side platforms, tech platforms and a digital publisher have also integrated ahead of launch.
Whacking two moles at the same time
The release comes weeks after an Adalytics report found a large number of major brands accidentally placed ads on a website that hosts child sexual abuse material (CSAM) over the past four years.
On a recent episode of The Media Leader Podcast, Responsible Marketing Advisory’s head of digital Emily Roberts explained how poor incentive structures led to ads being placed against such unsavoury content, including that agencies often expand limited inclusion lists in order to meet audience reach targets for campaigns.
This often occurs because trusted premium publishers will have articles blocked from receiving advertising because of blunt keyword exclusion lists.
Coghlan believes the efficiency and accuracy with which AI agents can be deployed will go a long way towards addressing the issue.
“Our idea is, if you’re a brand and you’re saying ‘These are my non-negotiables about my preferences — I want to apply that everywhere’, then if your agency, trading desk or independent partner is working on the Scope3 platform, anything that they do should inherit those brand standards,” she explained. “Then you don’t get that situation where [agencies] add 5,000 domains to an inclusion list because they need it for reach. That just can’t happen.”
Coghlan added: “If you have a very fast feedback loop of being able to classify content, the answer [becomes] kind of simple: don’t serve on it until you classify it.”
Not only can AI agents be used to avoid advertising against CSAM, fraudulent or MFA sites, doing so would reduce the amount of carbon expended on wasteful programmatic advertising, according to Coghlan.
While she conceded that AI tools themselves have large carbon footprints, she argued that “people are going to use AI, [so] let’s do it in a way that people are using the most efficient models”.
Built into Scope3’s platform are measurement and tracking tools to understand the carbon and water impacts of AI models. Given the high degree of waste and fraud in digital programmatic advertising presently, Scope3 believes the shift to more efficient AI agents will create a smaller overall carbon footprint for the sector.
Coghlan added: “You hit one mole and you actually whack two at the same time.”