What Marketers Miss in the GenAI Discussion

This content originally appeared in our XP² newsletter. Subscribe here to receive experience optimization insights like these, straight to your inbox.

If your LinkedIn feed is anything like mine, you’ve likely seen an avalanche of content about Generative AI. Is it the revolutionary tool poised to save our careers, businesses, and even humanity? Or is it the harbinger of doom straight out of a Mission: Impossible movie? While the potential impact of this technology is undeniable, the doomsday predictions might be a tad dramatic.

Instead of fearing the unknown, I took a proactive approach and talked to Oren Evron, Product Researcher at Dynamic Yield by Mastercard. In his 10+ years of experience in developing products to identify and service client needs in personalization, he’s also become quite the expert on how Generative AI has shaped the market. While he can’t predict the future, he did quite a good job at explaining how Generative AI might transform our jobs—and how we, as marketers, can not only adapt, but thrive in the evolving landscape.

Liz: We all know about Generative AI from ChatGPT and Dall-E—but what do you think is missing from the mainstream understanding of this artificial intelligence that might value marketers?

Oren: While the flashy demos and “magic” of AI has captured the public’s imagination, what truly empowers marketing is the ability to customize and control AI. Marketers need to understand the two key components of Generative AI:

• LLMs (Large Language Models): These are the engines that comprehend and generate text.
• Data sets: These massive collections of information train the LLMs and can assist it in performing a specific task.

The real power lies in training the LLMs on specific data sets relevant to the task you want to perform, to your target audience, and to your brand. (For instance, if you wanted to use your LLM to write a comedy but trained it with medical records, you wouldn’t get a funny script at all. But if you trained it with sitcoms and movie scripts, you might get better results.)

Imagine training AI with your unique customer interactions, brand voice, and previous marketing campaigns. This unlocks highly personalized content, targeted campaigns, and even predictive insights—all generated by AI—that understands your brand like no other and can assist with your marketing goals.

Currently, leveraging the full potential of Generative AI often requires dedicated R&D and data science teams which can be a barrier for many marketers. The future lies in user-friendly tools that empower marketers with limited technical expertise to:

• Select and curate data: Easily choose and prepare relevant datasets for training, without needing deep data science knowledge.
• Fine-tune models: Adjust pre-trained models or build new ones specific to their needs, without extensive coding.

So you’re saying I could choose an LLM, give it a data set of all my writing for the brand, and essentially create a “LizGPT” that would be able to execute my writing style and the brand guidelines? Wouldn’t I be making my own job redundant?

Yes, you’d have the ability to create a “LizGPT” or a “DY-GPT” with your data that would help you automate some of your more repetitive writing tasks. You’d have more time for higher-level strategic thinking, creative ideation, and relationship-building. It would make your life more productive and effective, but it wouldn’t eliminate the need for your human expertise.

You will, however, need to acquire some new skills to stay relevant. For example, prompt engineering: You’ll need to master the art of crafting clear and concise prompts that guide your Generative AI towards the desired output. This requires explaining to the model what outputs you expect and giving it proper examples/references. Plus, you’ll need to understand the capabilities of the model and how to interpret the outputs critically based on data input. Expect that you’ll need to interact with the GenAI chat over several iterations within a session, to perfect and optimize the result each time.

You’ll also need to maintain strategic oversight to decide which tasks to automate, which need to be done by a human, and which can be a human-machine collaboration.

Ok, now that we’ve dispelled the myth that AI is coming for our jobs, let’s talk about how we can better use it. Can you discuss some of the ways you’re seeing Generative AI being used in Personalization?

Generative AI has become a game-changer in the realm of personalization technology, offering marketers innovative ways to connect with their audience on a deeper level and creating more opportunities for 1:1 personalization.

For example, GenAI can generate variations of existing ad copy or product descriptions, helping optimize campaigns and personalize messaging for different segments. It can help translate your content into multiple languages instantly, and, with the right data set, account for the cultural nuances of each target market. This opens doors to global audiences while maintaining the local brand voice.

Need to adapt a brand-specific visual or design a product image that doesn’t exist? Generate it! It’s not hard to imagine a future where product recommendations feature personalized images that showcase the item in use by someone with similar demographics or interests.

Generative AI has also led to the development of chatbots that engage customers in real-time conversations, tailored to their needs and preferences. We’re offering Shopping Muse, a conversational assistant, to our retail and commerce customers. The experience will guide customers through large product catalogs from discovery to conversion, leading to simpler shopping, faster decisions, and increased conversion.

Together, these applications will incorporate customer data to create unique, dynamic interactive experiences that foster brand loyalty. And this is just the beginning!

How do you see generative AI evolving to play a larger role in personalized experiences?

I suspect video is the next frontier. OpenAI unveiled Sora, its text-to-video tool. Once it’s accessible beyond early testers, we’ll see more hyper-personalized video experiences. Imagine unique videos generated for each viewer based on their demographics, interests and past interactions. Serialized videos, like tutorials, might also become more dynamic and engaging. Rather than a one-size-fits all video, each video could adapt to each learner’s pace and understanding. Video will become more interactive as well, as quiz-style content and product demos change based on viewer inputs.

Beyond content generation, I think analytics and BI-tools will become even more accessible to personalization practitioners. For example, rather than manually sifting through data for insights, AI tools will allow you to engage conversationally, asking natural language questions about things such as audience data and receive personalized reports with key trends and insights. This will better empower marketers to make data-driven decisions and personalize campaigns with laser focus.

What are some potential challenges marketers need to be prepared for in the future?

Marketers need to be very conscious of their data sets, since this is what Generative AI models are trained on. Their outputs directly reflect the data they’re fed. This means that data quality really matters. Biases and inconsistencies in the training data can lead to biased and inaccurate outputs. Marketers need to be aware of and actively manage data quality in order to get the best outputs.

All the other potential challenges, really, stem from data: For privacy concerns, it’s about protecting sensitive customer data used for training and avoiding potential privacy breaches. The legal implications of using AI-generated content need to be clear, especially regarding copyright ownership around the datasets and potential infringement. In December, The New York Times sued OpenAI and Microsoft, claiming that millions of the publication’s articles were used to train chatbots that now compete with it.

I think one of the smartest things marketers can do is establish transparency with their customers. It starts by understanding how the models they use work and being able to clearly explain how it arrived at its outputs. This is crucial for building trust and avoiding unintended consequences.

GenAI Is Here to Stay—Embrace It

While GenAI may not take away every marketer’s job (or lead us all to utopia, for that matter) its inception marks the beginning of a new personalization era. Robust data sets can empower GenAI-based tools to craft and optimize highly personalized content, targeted campaigns, and predictive insights on the fly. Brands looking to take advantage of GenAI, however, should be transparent with their customers’ about how they use this data. In return, customers will reward them with their trust and loyalty.