AI can’t make a good ad. Or at least, it can’t make one from scratch. Too many important details will be altered, distorted or missing altogether. Brands need their logos to remain intact, and they require their products to be accurately represented. The sizing must meet the brand’s requirements and the context in which the ads are displayed.
Generative AI is simply not up to the task.
The problem is that Commerce Media could greatly benefit from using AI to overcome creative barriers. Scaling a campaign is particularly costly when you consider the needs of commerce media advertising. In addition to the brand requirements mentioned earlier, you must also contend with the complexities involved in advertising across multiple retail media networks (RMNs).
There are over 175 of them, and each has unique requirements. Not only does the brand need to design ads with the retailer specifications in mind, but it also needs to submit the creative to the retailer or commerce entity for review. When you multiply this lengthy process by tens of different RMNs, you’re potentially stuck with a costly workflow.
While generative AI can technically make an image, it can’t meet the demands of a commerce media ad. But that’s not the end of the story.
In a LinkedIn Live interview with the CPG Guys, Jivox CEO and Founder Diaz Nesamoney explains how he has found a solution to commerce media’s scale problem that utilizes AI more strategically, reducing costs and boosting performance significantly. It sounds magical, but it’s actually just a clever use of the AI landscape.
Below we’ve shared highlights from the interview.
The Generative And Agentic AI Combo To Tame Commerce Media Challenges
In order to scale personalization efforts in commerce media advertising, some efficiencies need to be made. Campaigns are too slow to launch, the approval process takes an eternity, and ad variations take too long to produce. Peter and Diaz dove into these challenges in the context of innovations in generative and agentic AI:
You recently made what I consider to be a pretty exciting announcement with your DaVinci Commerce platform. It’s agentic. The AI powers creative generation, resizing, compliance, checking and reporting. (…) What are the key efficiencies you’ve unlocked for marketers through DaVinci Commerce?
Diaz: The way I think of generative AI is that it’s primarily assistive, meaning that you can prompt it, and it’ll give you everything from copy to actual content. But fundamentally, it’s assistive in nature. And then agentic AI was a big leap, because it actually does things for you.
We first took agentic AI and applied it to the problem of creative, (…) how to use audiences to personalize (…) how to optimize for conversions, (…) and how do we reduce a campaign launch—about 40 to 60 days, believe it or not—down to a couple of minutes. And I believe we did achieve that.
A Refresher On Commerce Media Challenges
- Advertising across multiple retail media networks (RMNs) adds complexity to commerce media advertising.
- Brands have to produce ads that meet their own specifications and the requirements of each RMN.
- The retailer approval process can slow campaign launch to a crawl with frequent back-and-forth between the brand and the retailer/commerce entity.
- Campaign latency makes deep personalization nearly impossible.
To produce a creative that passed a gauntlet of requirements, Diaz quickly pointed to the need to utilize a combination of generative AI and agentic AI.
What Is The Difference Between Agentic And Generative AI?
Generative AI can perform a task when prompted, but it can’t work independently of human requests. Agentic AI, conversely, is an autonomous agent that can work out the steps required to achieve a goal. It operates more like a human employee rather than simply responding to prompts from a user.
See more definitions in our Commerce Media Marketing Glossary >>
Agentic AI, much like a human, can analyze a creative brief, use existing assets, activate generative AI to produce images and then check against a retailer’s guidelines. You leverage existing assets and information that can be combined with generative AI images. Not only do you get ads that meet internal requirements, but they also have a much higher likelihood of passing retailer approvals.
How Can You Launch A Commerce Media Campaign In Minutes?
As mentioned above, the key to resolving many scale-related challenges with commerce media is by utilizing a combination of agentic AI and generative AI. What hasn’t been mentioned is the utility of using a templated approach.
Diaz: So the way we’ve solved this is through the use of templates. (…) The brand logo and the retailer logo are in the template. They’re baked into the template, and they’re not changeable by AI (…). The product image and the product content are not changeable.
Sri: So you work with commerce media providers, aka retail media networks, as an example, to have specific templates that brands can comply with, and brands are going to be aware of what those templates are, obviously. And then you build several agents to understand how commerce media networks work, so that when a brand is ready to scale personalization, the agents are already equipped with the necessary intelligence to build that level of personalization in minutes and seconds. Is that how it works?
Diaz: Exactly. The agents are taking care of all of the complexity under the covers that a human would typically have to do.
How To Avoid AI Ad Slopification
Using agentic AI and templates is also a way to avoid issues with generative AI output. As it turns out, generative AI is not particularly adept at creating ads on its own.
Brand Consistency: Generative AI can create lifestyle images and other visuals that alter the context of the ad, which is crucial for efficient personalization. However, the brand logo and product, for example, should not be changed by gen AI. This helps maintain brand consistency while meeting a retailer’s requirements.
Using templates and agentic AI, brands will be a lot more successful at producing compliant ads. The years and cost that went into careful branding won’t be undermined by generative AI’s inability to replicate existing assets perfectly.
Is True Personalization Possible In Commerce Media?
Commerce media introduced a few exciting opportunities for marketers that have helped it grow rapidly in the last few years. The initial offering focused on hosting ads near the point of purchase, typically on a retailer’s website. This is how commerce media began. However, the larger potential lies in the transaction data that commerce media has.
Diaz: Data is incredibly valuable and more precise and granular than anything we’ve seen before; the creative has to step up to that same level of personalization. (…) That’s why AI is so important.
When creative is run against the data, if you’re running category-level creative, or you’re just showing some generic product content, it doesn’t actually tap into the personalization power of that data. (…) You’re picking the right audiences, but you’re showing them creative that is not relevant to them.
That’s why AI is so important. Because if you had to generate skew level creatives, and make sure that you always kept up to date with what products were available, what SKUs, what packages and even offers (…) that’s an unmanageable problem through human power.
Granular Audience Targeting: Personalization at the category level is somewhat limited. Still, when transaction data is added to the mix, the level of personalization becomes much more detailed and more reliably relevant to the consumer. Having access to this data is a significant performance enhancer, allowing for the efficient generation of creative content with AI and templates. Without these key efficiencies, it wouldn’t be possible to utilize Commerce Media’s first-party data fully.
Is Handing Over So Many Marketing Tasks To AI Risky?
With agentic and generative AI working in tandem, many of the risks associated with AI use are heavily reduced. However, the technology isn’t perfect, and mistakes need to be prevented.
Humans in the Loop: To avoid AI risks, you have to keep humans involved at every stage. You need a human to check the work and approve templates, for example.
However, before personnel do their review, it saves a lot of time to have an agentic AI already conduct some checks of its own. The ads generated need to be reviewed by an AI agent to ensure they will pass retailer inspection, thereby increasing the odds of approval.
Diaz: Yeah, that’s a great question. One interesting thing that we are doing now is using AI to check AI (…) we have a separate compliance checker that actually can check what the AI has generated.
Now, once the creative is generated, let’s say the user asks the AI to generate a lifestyle image, or whatever. It’s possible that the image contains something offensive. We have a separate compliance checker that can actually verify what the AI has generated.
When you think about OTC medications, there are legal compliance issues, and you need to make sure that those are safeguarded.
But there’s always a human who is the last check.
Legal Compliance Bottlenecks: Some ads are subject to more scrutiny than others. There are legal compliance issues involved in the pharmaceutical industry, for example. The copy must be accurate and specific, in accordance with regional rules and regulations.
Whenever you need to involve a legal team, it significantly slows down the launch of a campaign. However, with a compliance checker agent taking a pass first, you can speed up the time it takes for a legal team to assess the ads. There will be fewer back-and-forths between teams, resulting in fewer errors.
The Results Of Using The Agentic And Generative AI Tandem In Commerce Media
In commerce media advertising, you have to produce ads in various proprietary formats and for multiple retail media networks. Some brands have dedicated entire teams to individual RMNs just to handle the workload. When you use AI to automate most of the processes, you save significantly on production costs.
Diaz: We are working now with a global Alcohol Beverage brand. They work across 75 different retailers (…) imagine, even for each one retailer, you have to build a creative in many different sizes and make sure it’s compliant for all the different channels.
For this particular global CPG company, it came out to about a 76% reduction in their actual production costs. This is a significant, immediate, and very measurable return.
They are seeing an immediate performance lift, ranging from 30 to 50% in engagement.
Personalization is Back: Since production costs were previously very high, advertisers largely skipped personalization at the SKU level. However, now that production costs are so low, this level of addressability poses no significant financial burden.
Are We Headed Toward A Future Of Fully Self-driving Campaigns And Agent Buyers?
AI Is Not Driving Itself Just Yet: While newer ad-tech platforms involve a lot of AI, Diaz maintains we’re still quite far away from self-driving campaigns. A lot of trust still needs to be earned from AI before it completely takes over the driver’s seat.
Diaz: Several years ago—at one of these WPP forums, I believe—they were asking us to brainstorm on the future of campaigns with AI, and I did use the term self-driving campaigns, but I think we’re quite a ways from there.
I believe that’s a good vision to have, but just as we discussed during our conversation about humans in the loop and the guardrails, I think it’s essential to keep that in mind and not rush (…).
Mark Zuckerberg said, ‘By 2026, you’ll be able to push a button and then it’ll just create the ad for you and launch it. That’s a bit of a scary proposition, frankly, if you ask me.
Some of the things that I’m excited about and watching closely is agentic commerce. (…) I think the idea of an agent automating the part of a purchase that you don’t like is exciting. (…) AI also provides that discovery phase.
The Gift of Shopping Agents
Agentic commerce is a promising concept for the time being. While it won’t entail shopper agents magically making all your purchases for you anytime soon, they will automate a significant portion of the buying process.
Shopper agents will also give consumers a sense of discovery through conversation, much like talking to a sales clerk in a store who has memorized their inventory levels. That will be a highly positive development for both consumers and brands.
Watch the full interview on LinkedIn: Finally: Agentic AI Drives Scaled, Contextual, Personalized Commerce Media