Still human: Why consumer closeness matters in an AI-driven world

Since the early large language models (LLMs), Google’s BERT and OpenAI’s GPT-2, were released five years ago, the world of automation has never been the same. These AI technologies and their relatives have found their way into the deepest corners of our online lives — so much so that “the algorithm” has become a common phrase even among non-technical folks to refer to the invisible hand of the more or less sophisticated models that feed our collective experience.

Widespread awareness and excitement about these kinds of models grew with the release of ChatGPT toward the end of 2022, touching every point of our online lives. It also has marketers asking two age-old philosophical questions, “What does it mean to be human?” and “What is authenticity?”

Here are practical pointers about those questions for those in business and marketing.

Navigating humanity and authenticity concerns of LLMs

We are all still human

No matter how connected humans are, how sophisticated a business’s targeting is, or how much data there is about us, we are still human.

For instance, marketers may have an idea of who they “think” Gen Z is based on stereotypes or even based on their own collected data. However, like all other generations, Gen Z isn’t a monolith, it is made up of complex humans with varying beliefs and perspectives that don’t always translate to behaviors.

With so much uncertainty about the future, brands need to remain grounded in what matters to humans on a real level — one that can’t be calculated by an LLM. 

Dig deeper: Artificial intelligence + human intelligence = success

Avoid using LLMs to generalize a target audience or demographic

This over-generalization, or reduction of a group to a stereotype, is not new. It has been around since the birth of marketing (or the beginning of war, depending on your viewpoint).

We have made great strides over the last few decades in moving away from it, and yet, ironically, the easy access to LLMs may cause us to take a few steps back if we are not careful.

For example, I asked ChatGPT, “What does Gen Z think about avoiding buying fast fashion (i.e., mass production, low-cost fashion)?” and saw this output: 

“A significant number of Gen Z members are actively advocating for sustainable and ethical fashion practices. They often prefer to support brands that prioritize environmentally friendly and ethical production processes… Keep in mind that these are general trends, and individual opinions within Gen Z may vary.” 

If we take a different view of this question, informed by consumer research, we asked the same question in our recent What Matters 2024 study and saw that 32% are currently avoiding fast fashion — the highest percentage among all generations surveyed. Still, we can see that two-thirds of Gen Z are not participating in sustainable fashion. 

One of the great strengths of LLMs is their ability to produce coherent, well-written text that is “plausible.” They are not trained to provide correct, discerning or insightful answers, but ones based on the mathematical average of the information available to them. When we use these models, we must bring that discernment and insight to them.

Dig deeper: Mitigating the risks of generative AI by putting a human in the loop

The importance of consumer closeness

You must still strive to understand the people you wish to connect with. You can’t assume their individual wants and needs are in the model’s training set. You still need to interact with, listen to, and research your consumers.

Once you better understand your audiences, use that insight — in conjunction with LLMs and other tools — to see them as humans, not just groups. This idea is known as consumer closeness, and it’s more important than ever as marketers explore the abilities of these new technologies.

Dig deeper: AI in marketing: How to balance automation and the human touch

So, how do we best use these models?

Recent techniques such as retrieval augmented generation (RAG) and parameter efficient fine-tuning (PEFT) enable us to have the best of both worlds. Coupled with the insights we gain from actual human understanding and taking a step back for a wider environmental perspective, these technologies allow us to supercharge our efforts. 

These models provide the power and flexibility of the AI models coupled with specific insight into the real-world behavior and attitudes of those they wish to serve. The companies that can balance consumer closeness with the incredible power of these new models will thrive in the years ahead.

It is likely to be a bumpy ride, as we are already seeing, but sometimes remembering the obvious is the most important thing — we are all still human.

Dig deeper: MarTechBot update: Personas, image generation and new prompts