AI Agents Are Already Buying Media. Here’s What Advertisers Should Know.

AI has been part of digital advertising for years, helping advertisers optimize campaigns, adjust bids, find audiences, and decide where to spend budget.

Now, AI is starting to take on more of the actual media-buying process.

This summer, Butler/Till spent less than $10,000 testing AI agents to buy podcast inventory with iHeartMedia. Over four weeks, the agents delivered CPMs 42 percent below the client’s direct-buy benchmark and secured a better mix of mid-roll placements than the traditional plan had projected.

For a relatively small test, those are promising results. They suggest AI agents may eventually do more than make media buying faster—they could also change how efficiently media gets bought.

But we’re still very early, and the technology is moving much faster than the standards surrounding it.

AI is starting to do more of the buying

Most advertisers are already familiar with AI playing a role in their campaigns. Google, Meta, Amazon, and other platforms use algorithms to determine who sees an ad, how much to bid, which creative to serve, and where budget should go.

Agentic media buying takes that automation a step further.

Instead of simply helping a media buyer make better decisions, an AI agent can potentially take a campaign brief and act on it. That could include finding available inventory, comparing opportunities, negotiating pricing, creating deals, and making buying decisions.

In other words, AI is moving from helping media buyers do their jobs to actually taking on parts of the job itself.

That’s what makes the Butler/Till test interesting. The agents weren’t just used to save time or automate a few manual tasks. In this particular test, they bought media at a lower cost and secured a better mix of placements than expected.

That’s worth paying attention to. It doesn’t mean AI agents are ready to take over media buying tomorrow, but it does give advertisers a reason to start experimenting.

The industry hasn’t agreed on how this should work

The technology may be working, but the infrastructure behind it is still taking shape.

Right now, two competing protocols—AAMP and AdCP—are being developed to help AI agents and advertising systems communicate with one another.

Think of these protocols as the common language that allows an AI buying agent to interact with other parts of the advertising ecosystem.

The challenge is that we don’t have one common language yet.

The IAB Tech Lab identified 13 areas where AAMP and AdCP overlap, including campaign briefs, inventory discovery, pricing and negotiation, deal creation, and real-time decision-making.

Essentially, two different systems are being built to handle many of the same jobs.

Eventually, one may become the industry standard. The two may find ways to work together. Or something entirely different could emerge.

Right now, we simply don’t know.

Measurement is still catching up, too

There’s another important question: once an AI agent buys the media, how do we decide whether it actually did a good job?

The industry doesn’t have a complete answer to that yet either.

The IAB is developing a framework for measuring AI advertising, but that work is still underway. At the same time, the rise of AI agents is creating bigger questions about what advertising looks like when an agent starts making more decisions on behalf of a person.

If an advertiser influences an AI agent that ultimately influences a consumer’s decision, for example, what exactly should be measured—and how?

These questions will eventually get answers. In the meantime, they don’t mean advertisers should sit on the sidelines.

They simply reinforce an important point: we’re at the beginning of this shift, not the end.

You don’t need to pick a winner yet

Whenever a new advertising technology emerges, there’s pressure to figure out which platform or system will win and get there early.

We don’t think that’s the most important question right now.

Instead of trying to predict which agentic technology will come out on top, advertisers can start learning while keeping their options open.

Run a small test and compare the results against what you’re already doing. Try more than one platform or approach when it makes sense. Keep your advertising accounts and data under your control, and make sure your measurement can move with you.

That last piece is particularly important.

If every platform measures success differently, it becomes difficult to tell whether one is actually performing better than another. Advertisers need a consistent definition of success that allows them to compare results across platforms.

Do that, and you don’t need to know today which technology will be the best option two years from now.

What we’d do right now

For advertisers interested in agentic media buying, we’d focus on four things:

1. Start small. You don’t need a massive investment to start learning. Butler/Till’s test was less than $10,000, which shows that experimentation doesn’t necessarily require a major commitment.
2. Test more than one option. We’re still early enough that the first platform or technology you try may not ultimately be the best one.
3. Own your accounts and data. If a better option comes along, you should be able to move without rebuilding your media operation from scratch.
4. Measure consistently. Use the same business outcomes to evaluate each test rather than relying entirely on how an individual platform reports its own performance.

The goal isn’t to wait until every question about agentic media buying has been answered.

It’s to start learning now without getting locked into an answer too early.

Where Delve Deeper fits

A lot is going to change over the next 18 months. AI agents will get better, platforms will introduce new capabilities, industry standards will become clearer, and some of today’s approaches will inevitably gain more traction than others.

Advertisers don’t need to predict exactly how that will play out.

At Delve Deeper, we work across major advertising and media-buying platforms, helping advertisers test new opportunities while maintaining control of their accounts, data, and measurement.

That flexibility matters more as the market changes. If one platform performs better today, advertisers should be able to move budget there. If a new option becomes worth testing six months from now, they should be able to do that without rebuilding everything first.

The goal isn’t to correctly predict which platform will win. It’s to have the flexibility to use whichever platform works best for your business.

Want to learn more?

If you’re interested in agentic media buying, evaluating new advertising platforms, or figuring out where these technologies fit into your media strategy, reach out to Rich Holland at rich.holland@delvedeeper.com and the Delve Deeper team.

We can help you understand the options, identify opportunities worth testing, and build a media setup that gives you the flexibility to change as the market does.

Sources

• Digiday, Aug. 20 — Butler/Till’s agentic audio test with iHeartMedia
• PPC Land, Aug. 21 — AAMP vs. AdCP and the 13 overlapping functions identified by IAB Tech Lab
• Digiday, Aug. 24 — IAB’s work on an AI advertising measurement framework