How to use generative AI in relationship marketing

By Michele Fitzpatrick (VP of enterprise strategy at Marigold)

Over the course of the last six-to-eight months, generative AI has evolved from a plot line in a sci-fi movie to a technology that is disrupting entire industries, for better or worse.

In the last two months of 2022 alone, the monthly page views for one of the biggest AI text generator services — OpenAI — increased from 18.3 million to 304 million - a gain of over 1,500%. As of March, it was getting over 1.6 billion visitors a month.

With momentum like this, AI stands to change the very nature of how companies communicate. It’s not a question of if, but when. What is worth discussing in greater detail, however, is how.

One of the greatest benefits of relationship marketing technology is its ability to personalize content and communications with customers at scale, leveraging zero-party data to manage large volumes of customer segments, communication needs, and content creation. Done correctly, AI technology can amplify these efforts even further.

But done wrong, AI could destroy customer relationships and brand reputation overnight.

What is generative AI?

The simple definition of generative AI is any form of artificial intelligence able to create new content, including text, images, videos, etc. It works by learning from existing examples and using algorithms and models to generate something new.

This is different from traditional AI, which historically has focused on detecting patterns, making decisions, honing analytics, classifying data and detecting fraud. Generative AI examples include LLMs (large language models) like ChatGPT, image generation models like GANs (generative adversarial networks), and music composition models (think David Guetta’s use of an AI-generated Eminem vocal in his live sets).

AI-generated engines for text, images, and music typically generate content based on user questions or prompts. They respond to those prompts by gathering answers based on whatever data the engine was trained on. This can include books, articles, a library of images or other content provided by the engine creator, or the user as part of the original prompt. In some cases, the engine may also utilize search engines in real time to provide the most accurate response.

Using generative AI in relationship marketing

Generative AI has the potential to revolutionize the way marketers approach relationship marketing by providing more accurate, personalized, and efficient ways to engage with customers. This can be applied in a number of ways.

• Customer segmentation: Analyze customer behavior, preferences, and other data to create groups of customers with similar characteristics.
• Personalized experiences: Analyze customer behavior and preferences to recommend products or services most relevant to each.
• Customer service: AI is central to chatbot features trained to answer customer queries, resolve issues, and even make recommendations based on customer preferences.
• Recommendation engines: Suggest products or services to customers based on their past behavior and preferences, improving customer satisfaction and increasing sales.
• Predictive analytics: Predict the best time to send marketing messages to customers based on their behavior and preferences.
• Content creation: Generate product descriptions, social media posts, or even personalized emails based on customer data and preferences.
• A/B testing: Create variations of marketing messages to test on different customer segments to identify the most effective message.
• Lead generation: Analyze customer interactions and identify those most likely to become customers.
• Social media monitoring: Identify customer sentiment towards a brand by analyzing social media data to identify customer complaints, feedback, and opinions.
• Sales forecasting: Analyze past sales data, customer interactions, and other factors to create accurate sales forecasts.

For instance, you might train a generative AI engine on branded content so it understands your brand voice and customer engagement style. This also provides source material to establish style rules, tone, reading level, and more, so that copy and images created are more on-brand and better matched to your audience.

All of these use cases are in practice with many brands today. For example, Klarna is applying ChatGPT to provide product recommendations to users who ask the platform for shopping advice and inspiration with links to shop for those products via Klarna’s search and compare tool. Spotify introduced a personalized DJ designed to know your music taste and choose what to play. Axe A.I. Body Spray was developed using 46 terabytes of data, 6,000 ingredients, and 3.5 million potential fragrance combinations to create one unique scent, in collaboration with Swiss fragrance company Firmenich.

Overcoming generative AI’s limitations

For all its potential, generative AI is not without its drawbacks. It is prone to biases and inaccuracies, carries risks of plagiarism and copyright infringements, and can create ethical dilemmas (not to mention, at times, limited creativity).

But all can be addressed through education, operational excellence, and proactivity. Here are a few things to consider when adopting and maximizing the potential of a generative AI solution for relationship marketing:

• Be clear and concise: Ensure that your prompt is easy to understand and provides clear instructions for the AI to follow. Avoid using overly complex language or unnecessary jargon that may confuse it. The construction of a well-designed prompt is one of the key ingredients to successfully using generative AI.
• Be prescriptive and helpful: For example — “The following is an article from our organization blog. Summarize the content into the most helpful bullet points for a person interested in purchasing a book.”
• Specify output: Distinguish between a social share text under 200 words for Twitter vs a 500-word blog post. Also dictate the tone you want — friendly and casual, or formal and authoritative.
• Specify formats: If you have a specific structure in mind, be sure to include this information in your prompt, such as generating a bulleted list or a numbered sequence.
• Provide examples: You can prompt generative AI to “remember” an example of content and create a response based on that example. AI is capable of analyzing text and learning it to generate content.
• Make adjustments: If the AI produces unexpected or undesirable outputs, try rephrasing your prompt or adjusting its parameters.
• Find a balance: Overly restrictive prompts can limit the model's ability to generate creative or nuanced outputs, while overly open-ended prompts may lead to ambiguous or off-topic results. Experiment with different guidance levels to find the sweet spot that best aligns with your project's objectives.
• Evaluate: Regularly examine the effectiveness of your prompts by reviewing the AI-generated outputs and comparing them against your desired outcomes. Use this feedback to refine and improve your prompts.

Remember, generative AI only plays a role in the overall creative process — either in the beginning to help with ideation or at the end to help adapt it to other audiences or channels — not doing all the work alone. And a well-designed prompt is perhaps the most critical step. AI also shouldn’t deprive your organization of the self-discovery that comes from the thinking and writing process critical to successful campaigns.

But while the future of AI may seem daunting for some marketers, and exciting for others, embracing AI can present opportunities to enhance relationship marketing efforts with customers both faster and at scale. The biggest question remains not why you’d utilize generative AI to enhance your relationship marketing efforts, but rather how and to what end.

At Marigold, we see endless possibilities for using data, technology, insights and creativity to design and deliver a dynamic and impactful customer experience.

Visit meetmarigold.com and follow @Marigold on LinkedIn.