Advertising AI roundup

Where are we with AI in advertising? Meta and Google are sprinting, but how is independent ad tech keeping up?

The Vendor Interview: Viant

Marketecture Vendor interviews are free for a week, then require a subscription. Read more about subscription options at marketecture.tv.

Podcast: Mark Jablonowski on programmatic politics

Advertising AI check-in

This week’s vendor interview is with Tim and Chris Vanderhook from Viant, following-up on their buzzy video launch of ViantAI. We are 18 months past the world-changing reveal of ChatGPT and this is the first thing we’ve seen in the advertising world that’s made people take notice. Let’s take a minute to evaluate where we are with AI and advertising, what’s still hype, and what’s reality.

Agencies

We’ve covered this ground before in this newsletter, but let’s quickly review some of what we know.

Agencies are investing in AI, largely at the execution layer on top of other company models and for specific use cases like creative, media planning, etc.

• Publicis’ CoreAI is investing $325 million over three years and explains how this has totally transformed everything it does in this press release that you might need some ayahuasca to fully understand.

• WPP’s iQ is investing $317 million annually and announced an AI-powered production studio at Cannes this year.

• Dentsu is “Integrating AI Into Everything We Do” which could mean a lot of things.

• And Omnicom just this summer launched something called ArtAIBot, which is based on something else called Omni, which, if you believe the press release, “assembles clients' digital assets, to create and deliver high-quality personalized experiences.”

So basically they are all investing in reducing the cost of creatives, using the tools created by big tech.

Big Tech

Unlike the agencies, the tech giants are taking the lead on both building AI models and using them to power advertising. They have the distinct advantage of operating closed loop systems where they both can control the creative parameters, and get a clean feed of outcomes back to the models.

Meta has productionized its creative AI tools and is right now perhaps the largest commercial AI use case at scale. Here’s what they said last week:

Not to be outdone, Google posted a blog last week highlighting how AI is in search ads, image creation and image editing and resizing. They also put all of PMax into the AI bucket, but it is hard to see if that is real or just marketing.

Pinterest will auto-generate backgrounds for your images and claims to have a “Performance+” product that is pretty much exactly what you think it is.

TiKTok offers a “Symphony Assistant” that helps you develop and refine scripts for TikTok ads, just in case a you ran out of ideas after the latest fad faded.

Use Case: Creatives

From the get-go, the most obvious use case for LLMs was creatives. This is borne out by many of the production-level systems that have already been discussed above, from both agencies and big tech. But where are the stand-alone creative AI products?

My opinion from the start has been that the use of AI to build the actual creative (image, video, etc) was going to be a commodity, and that the more interesting question would be in what workflows or systems that AI was applied. If you look at the history of ad creative tech, there have not actually been very many winners, and the ones who have done well do so with integrations and workflow rather than creativity, per se.

I did a quick tour around some of the ad tech vendors known for participating in the “creative” space using the scientific method of just seeing who we had interviewed at Marketecture previously then Googling the company name + “AI”:

• Smartly: “AI Lab” with features focused on DCO

• Flashtalking: Specific products using AI, including versioning, taxonomies, and insights

• Celtra: Nothing

• Clinch: Nothing

• Xtreme Reach: Nothing

• Brightline: Nothing

• CreativeX: Nothing

• Marpipe: Nothing

There are also start-ups doing just AI creative like the well-named AdCreative.ai. But overall, the pattern remains that the ad tech vendors are more focused on workflow than the actual creation.

Use Case: Planning

The Viant demo that got this all started was exciting because it seemingly automated one of the most people-intensive parts of media, planning. Planning is an area where there is a lot of diversity in approaches, and a lack of standardized workflows or technology. Many large agencies use Mediaocean’s Prisma, while smaller agencies might use FreeWheel’s Strata, or Basis or something else. But even with these technologies, the intelligence is often happening outside of the systems, in spreadsheets, with the end results cut-and-pasted in.

I also ran across something called Microsoft CoPlanner which appears to be CoPilot for ads?!? Has anyone used this?

If you Google “media planning AI” you’re going to get a heap of junky press releases from all kinds of companies, including Salesforce (does anyone know what they actually do?), but you’re not going to find ready-to-use SaaS solutions that are ready today. It is possible this won’t change that much given how many agencies and buyers like to have bespoke solutions anyway.

Use Case: Trafficking and Optimization

Ad optimization has been a staple of ad tech platforms for almost twenty years. These systems used to be powered by “machine learning” but it is a lot sexier to say they are now AI. The reality is that every single field in every auction, bid, and win, has already been optimized to the efficient frontier, and LLM is not necessarily going to change that.

There are a ton of companies that do automated ad ops, or automated trafficking. I am excited to see if and how MLMs can help with this, but am yet to see any real examples. Companies like Chalice.ai and Cognitiv.ai are staking out ground here, but I cannot personally vouch for where the line is between old fashioned optimization and “AI”.

Use Case: Analytics

This is an area where I think LLMs can be real time savers. While a trader might have the time and skills to look at detailed reports and optimize campaigns, the translation of that data into insights for executives is a clear opportunity for LLMs. I’ve seen demos of products that do exactly this, letting you query ads data with statements like “what happened last week”, or “why are sales increasing.” This is another exciting area for innovation.

I’m playing the role of an AI dilettante here, just reporting back what I see without a lot of specialized expertise. I want to hear from you — what’s the coolest ad-specific implementation of AI you’ve seen?

Reading list

• Lotame’s Monfried drops bomb on TTD on Twitter (link)

• Tiktok letting advertisers target its search results page (link)

• Rate Zuck’s drip (link)

• Meta shows tons of AI updates and enhancements (link)

• Meta previews Orion AR Glasses (link)

• Open internet programmatic has only grown 3% since 2021, according to Wolfe Research (link)

• Former “Justify Your eexistence” guest, UPAID, has gone under (link)

• Google offered to sell Adx to appease EU pubs (link)

• IAB issues an “exclusion notice” indicating potential patents (link)

• Equativ (Smart) acquires a Kamino Retail. Seems like a French RMN platform (link)