Personalization Pros 2024 Meeting

Meeting Themes

“Air of crisis in the digital world.”

The group discusses the convergence of loyalty and personalization platforms, how to personalize for B2B, the rise of LLMs and AI, and more 

Loyalty Personalization and… Affinity Modeling?

Colin brought up the ongoing trend in the convergence of personalization and loyalty platforms. 1 in 3 businesses without a loyalty program today will establish one by 2027 to shore up first-party data collection and retain high-priority customers. (Gartner) As a result, loyalty programs have become a sort of default hub for a customer’s personalization interactions with a brand. The loyalty stack faces the same old integration challenges we see in any other MarTech platform, however, i.e. where does the data live, creating the single view of customer, data ownership and integrity, etc.

Amanda discussed Marigold Loyalty and how they are working in this space. Marigold makes it easy to deliver and manage coupons, certificates, gift cards, discounts, cashback, sweepstakes, donations, physical rewards, and more. She also mentioned trends that have received notoriety, but haven’t fully played out yet, such as the intersection of NFTs and loyalty. The group discussed how Generative AI is having an impact but new format hasn’t really landed.

Jaina talked about an interesting new research project for universities exploring can you track user behavior and correlate with affinity. For example, they looked at metrics like time at the university or pride in rankings of university. In terms of weightings, they found most indicators came out to be the same.

B2B Personalization and Personas 

Kevin discussed a new AI trend seen more and more in the B2B space, which is how they can work within multiple systems in which content lives. It’s hard to maintain integrity for brand, customer experience unless there is structured data and content. He worked with Scott Abel on a technical content inventory. Look at Salesforce or Adobe as leaders for technical documentation and support.

B2B Sales Personalization is another issue. It’s not just a company, but where various audiences are in the customer journey. For example, the Procurement Team (one persona certain part of process), Engineer, etc. “The B2B customer journeys we see there are generally three types of personas depending on the product: influencers, evaluators, and decisions makers.” Community platforms and engineer zones can be important. Another persona is KPMG / Deloitte who are often helping them make decisions.

Cruce gave an example of documentation content modeling over 1M items, with a to provide answers based on contextual info acquired through the session. “Normally we would do through remixing of structured content objects but now adding in an LLM layer. It allows to start writing content on the fly — experiment with Chat GPTs JSON Schema (just a simplified schema portability format). LLMs are being trained to work with a provided schema.” He plans to walk a case through on Kevin’s show later this month.

The group discussed some implications and risks of exposing data to AI engines. “Don’t train on this option,” is provided, but not always fool proof. Typically they use open source model and use specialized applications.

John reported that he’s “out of the trenches” of actual practitioner concerns, more overall solutions, particularly in healthcare. He has been focusing on is using Optimizely CMP as a way of orchestrating LLM interactions in a customized way for larger companies, like a car manufacturer. There are SaaS products like StoryQT

https://www.qt.io/resources/qt?content-type=Success%20Story

The group also discussed the emerging effects of AI on personalization, particularly around content generation. “There’s a one-two punch,” says Berndt. “One of the things that always held back personalization was incremental cost of content development. But now generative AI workflows can create different versions of messages in channels. The second is governance and control of content fragments for example. The static website case, all types of content you wouldn’t have written you would now do to create SEO benefit. Could you use LLMs to look at all the content and flag for freshness, etc.”

Growing concerns about AI and plagiarism are real, although relative to the coming AI paradigm shift they may be missing the forest for the trees. “They miss the enormity of the point of what abouts to happen to human self-image,” says Berndt. “It should be an extremely weird moment for human beings. I grew up in a Chompskyan household, my mom was a cognitive neuroscientist. He was focused on the plagiaristic aspect of it, but the convergence of the simulation of comprehension with actual comprehension is astounding.”

A number of the fellows reported the use of AI platforms. Valtech has built Valtech AI to live between workflow engines and LLMs for managing prompts. You can push same prompt to different Gen AI platforms. Phaedon is also working on an AI solution based on the Microsoft stack that it plans to release to clients this year.