First published on Travis's LinkedIn, where he shares his take on agentic advertising, ad tech, and the business of building in this space. A marketer turned entrepreneur, Travis is co-founder and Executive Director of Adzymic, with over two decades across adtech and digital media in Asia Pacific — including a stint leading SAP's regional digital media practice. Follow him on LinkedIn for more.

Over the past few months, I have been talking quite a bit about agentic advertising with publishers, agencies and brands.

I started with a fairly clear idea of where this is going: buyer agents talking to seller agents, with much of media planning, buying and optimisation eventually happening autonomously.

After presenting this many times, I noticed I was spending a lot of time explaining how it works. What is a buyer agent? What is a seller agent? How do they find each other? Where does the DSP sit? Who gives an agent permission to spend?

Then we get into protocols, MCP, A2A and how all of this connects with the programmatic stack we already have.

Seems the same complexity we had with Programmatic is now reappearing in Agentic.

Though publishers would ask a much simpler question.

Where are the buyers?

Some of the most useful conversations I've had have been with publishers.

The seller agent concept is relatively easy to understand. Make your inventory, audiences and advertising products discoverable to AI buyers, and potentially open up a new source of demand.

Then comes the obvious question: where are the buyer agents?

It's a fair question. Publishers don't need another piece of technology. They want buyers and revenue. We can build seller agents and protocols, but someone still needs to show up with demand.

With agencies, the conversation gets practical quickly. They already have DSPs, trading teams, planning tools and processes that work. An agent has to make that job easier or produce a better result. Otherwise, why change?

I probably underestimated this part when I first started looking at agentic buying.

Why end-to-end still matters

An AI agent could take a brief, build a media plan and use MCP or APIs to execute through the platforms an agency already uses. It could generate creative, analyse campaign data and take on more of the execution.

End-to-end agentic buying can go much further. Give a buyer agent an objective, budget and guardrails. It can discover opportunities across multiple sellers, decide where to spend, execute and continuously move budget based on what is actually working.

Today, media buying is fragmented across platforms, channels and workflows. Each platform sees its own piece of the campaign and optimises within it. A buyer agent could sit across those boundaries and optimise against the advertiser's objective instead. And it should get better over time. With enough context and performance history, the agent can learn from one campaign and apply that learning to the next. Combine that with creative performance and insights such as BLS, and you start to build a much more powerful feedback loop between media, creative and brand outcomes.

It could also discover inventory the buyer may not have considered, compare direct publisher opportunities with programmatic ones and react faster to performance. For publishers, seller agents could also make more of their inventory and unique products discoverable.

That's what interests me about end-to-end agentic buying. It's not just automating what a trader does today. The agent can learn across media, creative and outcomes, and bring those learnings into the next campaign.

End-to-end agentic buying is still early, so the practical question is how we get there from the way media is bought today.

Maybe the first buyer agent doesn't replace the DSP

For many agencies, the starting point may be an agent that works with the tools they already use.

Give the agent a campaign brief and budget. It builds the plan, generates the creative and looks for media. If a publisher's seller agent has what it needs, buy there. Otherwise it can fall back to the DSP, or an API, whatever gets the job done.

The existing programmatic world and the emerging agentic world can work together. As more publishers expose seller agents and the technology matures, more buying can happen directly agent-to-agent.

We don't need everyone to change at the same time.

The technical work is moving fast

There is a lot of work happening around agentic protocols and infrastructure. We will need it if independent buyer and seller agents are going to discover, communicate and transact with each other at scale.

Getting a protocol to work is one thing. The harder part is making agentic buying useful enough that buyers actually want to use it.

Here is my objective. Here is my budget. Here are my guardrails. Help me get the best outcome.

Behind that instruction, the agent could compare direct and programmatic supply, generate creative, shift budgets and learn from performance. Some of that may happen through existing platforms, some directly with seller agents.

Protocols will matter as more independent agents start talking and transacting with each other. But they should make things simpler for the industry, not create another layer buyers need to understand.

Where we are heading and how we get there

The first buyer agents will probably work with much of the infrastructure, platforms agencies already use today, while connecting directly to seller agents where it makes sense. As more sellers become accessible to agents, more of the buying can happen directly between them.

My take is that full agentic buying will not happen overnight. It will happen gradually through pilot campaigns, testing and learning. Buyers need to see the value and become comfortable letting agents make more decisions, while more of the supply side becomes ready to transact with them.