Published May 29
Agentic AI is already changing the web.
That was the theme of the annual IAB Tech Lab Summit on Thursday. Zooming out, that broadly consists of three evolving parts of the web: agentic advertising, consumer experience and publisher economics.
If you zoom out further, that means how media is bought and sold; how consumers are searching and shopping online; and how content is being licensed, managed and monetized.
On the latter, the Tech Lab released new recommendations this week for how content owners can strategize against AI crawlers.
During the event, The Current caught up with IAB Tech Lab CEO Anthony Katsur on a variety of topics. Here, he got candid on misconceptions around agentic that he’s sick of hearing; why we need agentic standards; what Publicis’ acquisition of LiveRamp says about agencies and agentic; and more.
On misconceptions around the role of DSPs and SSPs in the agentic era:
“Agentic will not break down the DSP and SSP paradigm. Tech doesn’t change those paradigms, business requirements do…. I do think it’s getting gray where decisioning is occurring.
The DSP is still ultimately the arbiter of performance and driving to outcomes. DSPs are the Formula 1 engines to do that….That said, some of the decisioning around content and audience could shift. But that means buy-side platforms could get better curated supply and inventory to make those decisions on driving outcomes. And I think that’s a shift [and agentic is one factor driving that].”
On agentic standardization:
“Right now, agentic standards are the Wild West…. Everyone is running off and developing agentic solutions. This concept that we may not need taxonomies or standards is bulls---. If we’re asking agents to discover, plan, execute, measure, reconcile on our behalf with some degree of autonomy, I would argue standards are more important because we as humans understand nuance. Those standards become guardrails and substitute for a human being in every step of the workflow. So I’m tired of hearing that the agents will ‘figure it out.’… We can’t leave microfinancial transactions up to chance.”
On the most common agentic use cases he’s seeing:
“The common use case we’re finding across the industry is people asking how they can use agentic to make programmatic smarter and more effective. I don’t see a tremendous amount of use cases around agentic direct buying because one would argue that agentifying the insertion order has already been solved. A lot of that workflow is automated. Is the juice worth the squeeze to agentify that?”
On the Publicis-LiveRamp deal (and what it says about agencies):
“I see a world where agencies, particular holdcos, become more like technical general contractors. I think this is an example of that. When we talk about agentic and things like hallucinations, one of the reasons an agent would hallucinate or make a mistake is a lack of data quality. The LiveRamp acquisition in that sense you could back into an agentic story because LiveRamp is the largest identity spine in the world. I think Publicis owning that spine could be woven into a greater agentic story; agents are probabilistic machines and that’s why they hallucinate. So having that robust data spin to have determinism around audience and context is critical to a reliable agentic workflow.”
On agentic commerce:
“There’s a real consumer impact there. How will agents change the shape of e-commerce? Tim Berners-Lee [inventor of the World Wide Web] envisions a world where agents will discover products on our behalf at the very least. It may not book your flights or hotel, but it sure will give you advice on where to stay, what the cheapest flights are. It will automate a lot of that.
Change of behavior is hard. I think we’ll be in a world of AI-assisted commerce for a while… AI will help with the discovery process. How do you advertise to an agent? The other issue is, how does AI remain objective? That’s another thing we need to be weary of. AI bias is a real thing. Maybe not intentionally, but there is an imprint.”