Eli’s Thoughts 8.25: Marketers can lead AI into the Enterprise with Clean Rooms

Happy Monday! Before I get started, we will be hosting a Happy Hour at AdWeek NY on October 7th in NYC. RSVP here, I’d love to see you now that i’m back from paternity leave!

As this was my first child, I had a lot of new parent questions while I was out. Sometimes these were answered by my wife, sometimes by grandparents, but often by my close friend, the internet. My 2am queries were done one handed on my phone, asking Google’s Gemini how many ounces I should feed a 3 month old, what temperature should the milk be if the 3 month old is fussy, how to find a lost cat (we got him back!) etc. Out of curiosity, I began looking at the cited source data to the AI answers…a lot of the citations came exclusively from Reddit.

This is not an indictment of Google. If anything, I applaud Google as they made the deliberate choice to pay Reddit $60M for this data to use in their AI model. NY Times has signed a similar deal with Amazon AI. According to SEMRUSH, across Chat GPT, Google Gemini and Perplexity, 40% of LLM responses cite Reddit as the source. That’s a huge number and even more than Google owned Youtube!

Data is everything to these AI models, and unique data is clearly of value given the investments being made by the LLMs in procuring it and the amount of use (40% of queries) the consumers are getting from it. This is an area we as data driven marketers know extremely well. As LiveRamp CEO Scott Howe recently said on the Marketecture podcast, LiveRamp originally helped our clients with Data Onboarding: Connecting customer lists to media partners to make media targeting more efficient, a one way monologue. That then morphed into Data Collaboration: connecting data between partners to deliver better experiences to consumers, a two way dialogue. That leads us to today where Scott argues that just as Google’s LLM looked to proprietary data from Reddit to inform their models, brands need a way to collaborate with eachother securely in order to power their AI endeavors. This is where Scott makes the provocative claim that LiveRamp can now be considered an “AI Enablement” company.

Just as Reddit provides unique data to a Google LLM, LiveRamp’s Data Clean Rooms are helping brands access unique data from publisher and 2nd party brand partners by giving them a safe, secure and scalable way to collaborate with data. This data which today is used for attribution and audience creation, can also serve as the unique and proprietary data sets of enterprise AI models.

As an example, a large bank historically used LiveRamp Data Onboarding to make sure they were not wasting ad spend on current credit card customers by sending that data to your publisher partner. Then, using Data Collaboration that publisher could tell the bank which prospects viewed that acquisition campaign so the Bank could better attribute their new card applicants. In the future, you could see an AI model then using that enlarged and proprietary data for AI Enablement to identify the consumer behaviors amongst the combined publisher, 1st party, and credit bureau behavior that are most likely to signal a coming charge off on their credit cards, moving the use of the data from marketing to forecasting at the Bank’s broader enterprise level (while creating a great suppression audience for the marketing team!).

Data driven marketers and their partners, with their cross functional capabilities that touch cloud, data, collaboration, and relentless focus on performance, are by far the most equipped part of an organization to drive meaningful business value from AI. LiveRamp’s Data Clean Rooms are the tools that can help make sure functionality, compliance, and control are embedded in your organization’s journey from Data Collaboration to AI Enablement.

Your LiveRamp account team is standing by ready to help!

Have a great week!

Eli