Transparency issues around A.I.-assisted media buying might increase ad waste, not eliminate it.

Artificial intelligence has been heralded by many as a revolutionary technology in the media industry, with the power to drastically cut down on the time needed to create, execute, optimize and measure a campaign. All of those processes will be automated, return on ad spend (ROAS) will soar, wasted ad spend will near absolute zero and advertising will more or less be perfected.

Andrew Lipsman, founder of the digital media consultancy Media, Ads + Commerce, isn’t buying it, however. 

Automated buying platforms exacerbate longstanding issues about the lack of transparency in the ad-buying process, Lipsman says, and wasted ad spend will actually increase as more brands and agencies let A.I. handle their ad budgets.

It’s a startling take, as most of the conversation around A.I. is decidedly positive. But Lipsman believes the machines will prioritize short-term goals over long-term brand-building and driving incremental sales. In the conversation below, Lipsman lays out his contrarian stance on A.I., specifically as to how it relates to retail media networks, and makes the case that in this opaque media buying ecosystem, sales data is the most valuable currency.

In this edition of “How AI Will Change,” The Outcome talks to Lipsman about his more measured, if not pessimistic, view on the AI hype. 

‍

At a high level, how is A.I. changing retail media networks?

It’s not changing them that much.

‍

Really? Everything we hear about A.I. is how it’s going to revolutionize the industry.

There's been a lot of industry conversation about agentic commerce, which is a consumer prompting a large language model, such as ChatGPT, about some desired purchase. Then the LLM goes and does every step of the process, from price comparison shopping to completing the transaction. And I don’t believe that is going to happen in any way, shape or form. The best case scenario is that it will be on the extreme margins of how people shop and buy. (Editor’s note: since this interview took place, Walmart announced a partnership with OpenAI that allows purchases to take place within ChatGPT, so we’ll see how this plays out soon enough.)  

Look at voice commerce. When I started at eMarketer as an analyst, everyone thought we would soon be buying things through our Alexa — and it didn’t happen. We get so infatuated with what technology is theoretically capable of that we don't think about how consumers actually interact with it and what's required for habit formation.

‍

Fair point, but just because voice commerce failed doesn’t mean A.I. commerce will fail.

There are so many points of failure in your average agent commerce purchase that almost none of it is viable in the long run. 

Take the sell side: Certain categories, such as apparel, have return rates as high as 25 or 30 percent. If you completely outsource a purchase to an agent, the chances that you're going to have to return that item go up by a lot. This market can't exist with 60, 70, 80 percent return rates. It's just not economically viable. If I’m a retailer, I opt-out of that mechanism.

For consumers, they will very quickly learn agentic commerce is not all that useful at finding the products they want. Many people believe the standard web search experience has deteriorated, but they’re going to trust a shopping agent? If they don’t have a high degree of trust in these agents, they’re not going to allow them to make purchases on their behalf.

‍

But there are already people using agents to find products.

Yes, but through search, and search is definitely changing. Search engine optimization will evolve to generative engine optimization. Retailers will have to optimize product listings for the best visibility possible in A.I. search.

That to me is not a revolution. It's an evolution of how search works today. The idea that people are going to just set shopping agents and forget it and opt out of the whole purchase process to me is just patently ridiculous.

‍

What about for ad transactions? Surely A.I. is going to have an enormous impact there — right?

Somewhat. Buying and selling of media is already largely automated. Yes, there are parts of the process where you have human hands on the keyboard, but programmatic is infused into a lot of it. So again, A.I. is an evolution, not a revolution. And I don’t think A.I. is ever going to completely remove human intervention.

‍

So you’re not as bullish on A.I. for RMNs as the rest of the market?

A lot of people look at the upside of A.I. and ignore the downsides. I look at the downsides, and the most prominent one is the lack of transparency. Marketers don’t know where their ads are running, and the platforms pushing these agents are using them to sell their own inventory. Agentic media buying platforms are black boxes, and giving your money to them is wasting your money.

‍

I’m surprised to hear you say that considering that agentic media buying is supposed to more efficiently drive conversions.

These systems game the attribution models. They optimize to whatever makes the ROAS metrics look the best. That means buying a lot of cheap, low-quality junk impressions riddled with bot traffic.

Let's assume you are reaching some real humans with the campaign. Even then, the A.I. will optimize to the people already predisposed to convert. So it’s just retargeting people who have already demonstrated an intent signal. It’s not driving incremental sales.

Advertisers aren’t benefiting at all. The only efficiency is how the efficiency that's being created is allowing the platforms to capture more spend.

‍

How much of this problem is a matter of brands optimizing to the wrong metrics?

One of the best signals is outcome-based measurement, and my vision for the future is that transactions become a new form of ad performance verification. Because that’s the one thing that fraudsters can’t fake — transactions.