Double Impact: Harnessing Generative and Evaluative AIs for Effective Marketing Decisions

Article published in the journal NIM Marketing Intelligence Review VOLUME 16 (2024): ISSUE 1 (May 2024)

Dr. Christian Scheier
Co-Founder & Chief AI Officer,
aimpower GmbH

Dr. Dirk Held
Co-Founder & Chief Strategy Officer, aimpower GmbH

The AI landscape: More than just generative models

In the realm of artificial intelligence (AI), the term has always encompassed a diverse range of technologies. Recently, the spotlight has been on generative AI models, known as “GenAI.” Their capabilities have captivated the marketing world, with their potential appearing boundless. This fascination with GenAI risks overshadowing another critical aspect of AI: its analytical capabilities. Previously, AI was synonymous with predictive analytics, classification and pattern recognition. The current hype around GenAI shouldn’t detract from these foundational abilities. For a holistic and sustainable AI implementation in businesses, it’s imperative to integrate both generative and predictive AI. While GenAI offers tremendous assistance in these areas, it’s not a standalone solution. Predictive AI can be highly useful for evaluating resulting marketing strategies.

Evolving processes, enduring marketing pillars

Technological advancements have revolutionized marketing processes, yet the core success pillars remain unchanged. Successful marketing requires, first, target audience insights – understanding the nuances of the target audience is foundational to any marketing strategy. Second, it is crucial to have a relevant and differentiating strategy that sets the brand apart from competitors and resonates with the audience. Third, the strategy must be executed in a creative and brand-cohesive way. Creativity isn’t just about originality; it’s about aligning with a brand’s identity and message. Finally, marketing activities need to be evaluated. Assessing the impact of marketing efforts ensures that strategies are effective and provides insights for future campaigns.

The effective utilization of both generative and evaluative AI in marketing represents a significant advancement in how businesses can approach their marketing strategies.

Previously, these pillars rested solely on human interaction and intuition. Now, AI serves as a complement and, in some cases, a substitute for human roles. This not only applies to generative AI but also to the evaluation process for which a growing number of solutions are available. However, the essence of these success factors remains unchanged. More than ever, formulating strategies and understanding consumer insights on the part of human experts are indispensable. On this basis, human experts need to guide generative AI but should not neglect the opportunities to also evaluate results by using evaluative AI. Operating without thorough evaluation is akin to navigating without a compass. The combination of generative and evaluative AI allows managers to make sound decisions and continually improve marketing processes (see Figure 1).

For a holistic and sustainable AI implementation in businesses, it’s imperative to integrate both generative and predictive AI.

While GenAI can inspire and guide the insight process, it cannot replace the need for thorough analysis and evaluation.

The irreplaceable human element: Idea and strategy formulation

At the heart of every effective GenAI application is a human-driven starting point and set of guidelines. GenAI, no matter how advanced, relies on human input to initiate and direct its processes. Without this human element, GenAI would produce generic, undirected output. The role of humans in idea generation and strategic guidance cannot be overstated. It’s the creative spark and strategic foresight of humans that set the direction for GenAI to follow. Furthermore, the formulation of brand and communication strategies, while informed by evaluative AI insights, remains a distinctly human decision. It is at this juncture where the true innovativeness and distinctiveness of a strategy are determined. GenAI can provide options and scenarios, but the strategic decision – what makes a brand unique and compelling – is a uniquely human endeavor.

Content creation: The thin line between plausible and true

One of the most apparent benefits of GenAI in marketing is its ability to rapidly generate a multitude of content, texts and visuals. This capability is a game changer in terms of efficiency and creative exploration. GenAI models are trained on vast datasets, learning the statistical likelihood of words or pixels appearing following a given prompt. This training makes their responses consistently plausible. However, plausibility does not guarantee accuracy or factual correctness. It’s important to recognize the current limitations and challenges in this domain: Currently, public models struggle with pixel-perfect generation of products, logos, iconic brand assets or fonts and accurately generating human features – even as basic as producing the right number of fingers or teeth. Training data biases are a known issue and individual models predominantly focus on separate domains, such as image, audio or text generation. Further, generative AI models often “hallucinate,” producing plausible but potentially misleading or incorrect outputs. This is particularly concerning when seeking consumer insights or evaluating campaigns. For instance, while GenAI can generate plausible reasons why a product might appeal to a particular demographic, these reasons might not align with reality. The challenge lies in distinguishing plausible from accurate insights. This is where the analytical prowess of AI becomes invaluable. By efficiently analyzing large datasets – whether from surveys or social media channels – AI can extract insights more quickly and accurately than ever before. While GenAI can inspire and guide the insight process, it cannot replace the need for thorough analysis and evaluation. Without this, insights might seem reasonable at first glance but ultimately lead to misguided decisions.

Content value: The thin line between nice-to-have and effective

Most importantly, the question of selection arises: How do we choose the most effective assets from a plethora of generated content? Without proper analysis and evaluation of their impact, the process devolves into an inefficient trial-and-error method. This not only negates the efficiency gains but it can also lead to significant financial waste or even possible backfiring of communication. Beyond short-term effectiveness, the long-term impact on brand equity is crucial. To maintain a distinctive brand identity, the executions must align with the brand’s core values and aesthetics. Here, evaluative AI can play a pivotal role by ensuring that only content that resonates with the brand’s identity and target audience is selected and utilized. In the absence of systematic control via evaluative AI, the use of GenAI by different individuals within a company can lead to inconsistent and disjointed branding efforts. Managing the sheer volume of content alone requires automation. If GenAI is added to the mix without intelligent analysis and oversight, the challenge of maintaining brand consistency and effectiveness becomes even more daunting.

Increasing availability of evaluative AI in the marketplace

While generative AI is already on everyone’s lips, evaluative AI and its potential are currently being less widely discussed. Nevertheless, the evaluative AI market is characterized by a growing number of vendors, many specializing in distinct aspects like predictive eye tracking, image persuasion analysis, image memorability prediction, text content analysis or social media click-through rate predictions. However, the landscape is not limited to these niche offerings. Holistic solutions are also emerging to address all relevant marketing touchpoints. These offer capabilities ranging from execution aspects, such as visual attention and simplicity assessments, to strategic evaluations that ensure alignment with the intended brand message. Looking ahead, the trend is moving towards fully integrated platforms that encompass the entire AI-based marketing workflow. These platforms promise to cover everything from audience insights and strategy development to creative execution and performance monitoring. With some solutions, such as brainsuite.ai, which are already available in this space, the market for evaluative AI in marketing is poised for rapid growth, underscoring its increasing importance in shaping effective, data-driven marketing decisions.

In the absence of systematic control via evaluative AI, the use of GenAI by different individuals within a company can lead to inconsistent and disjointed branding efforts.

The strengths of evaluative AI in marketing assessment

When it comes to evaluating marketing efforts, evaluative AI is superior to generative AI. Its strengths lie in its ability to make predictions specific to channels, brands and target audiences. Recent multimodal models like GPT Vision combine text and image understanding and can interpret and comment on image content. However, without further training these generic foundational models lack insights into target groups, brands and their competitors. They fall short of providing reliable and precise quantifications and analyses, let alone benchmark analyses. Given the level of nuance and detail required for accurate and detailed analysis, evaluative AI is more suitable – especially considering that most marketing campaigns are multimodal, encompassing text, audio, visuals and video. Evaluative AI is able to make pixel-accurate predictions about the effectiveness of all multimodal elements, such as brand presence or product placement within a medium. It allows for ease of processing checks which evaluate the simplicity of textual, auditory and visual content, including video, to ensure its accessibility to the intended audience. Evaluative AI is further able to ensure the correct tonality in created content, including checks for appropriateness and sentiment alignment. It tests whether the content effectively conveys the intended message and aligns with the overarching brand strategy while also being persuasive (see Figure 2).

Integrating generative and evaluative AI for the continuous optimization of marketing processes

The future of marketing lies in the seamless integration of generative and evaluative AI, guided by human expertise, as depicted in Figure 1. This integration involves a dynamic interplay where generated texts, audio, images or videos are evaluated and refined by evaluative AI, creating a continuous cycle of generation and evaluation. This process ensures that the final output is not only creative and diverse but also effective and aligned with marketing goals. The key to success in this integrated approach is intelligent collaboration between these AI aspects and human guidance. Effective and sustainable business implementation requires these systems to work in harmony, complementing each other’s strengths. Intelligent workflows can transform marketing from insight gathering to content creation, making the process both more efficient and effective. Such workflows necessitate an integrated, AI-based platform that encapsulates the entire brand management process. This platform should facilitate the seamless flow of information and insights between generative and evaluative AI systems. Additionally, integrating consumer research into this platform is essential to ensuring that the insights and content generated are grounded in real-world consumer behaviors and preferences. The effective utilization of both generative and evaluative AI in marketing represents a significant advancement in how businesses can approach their marketing strategies. By combining the creative and generative capabilities of GenAI with the analytical and evaluative prowess of evaluative AI, marketers can not only streamline their processes but also enhance the effectiveness and impact of their campaigns. The key lies in maintaining a balance between technological innovation and human insight, ensuring that marketing efforts are not only efficient but that they also resonate deeply with the target audience.

References
Jo, T. (2023). Deep Learning Foundations. Springer.
Search in Google Scholar

West, P., et al. (2023). The Generative AI Paradox: “What It Can Create, It May Not Understand.” arXiv.org.
Search in Google Scholar