Real Impact Beyond Buzzwords: AI Methods in Marketing and Sales

Sebastian Seibel & Frank Ohnesorge

An Abundancy of Data Sources and Types

With the current momentum and computational potential, Artificial Intelligence (AI), especially generative AI, is becoming an essential tool for marketing and sales, expected to reshape and disrupt operations and customer engagement. OpenAI’s ChatGPT stands as a prominent exemplification of recent AI advancements and its growing adoption, as the large language model demonstrates the diverse potential of AI in various downstream tasks to a large audience.

Particularly for the marketing and sales domain, the combination of advancements in varied data collection, computer science, and computational power has brought about great opportunities. AI informs more effective marketing strategies while supporting a myriad of content generation tasks. A large corpus of digital products and tools creates input data for AI to leverage. Think only the wealth of insights gained regarding one’s ‘digital footprint’ on social media as individuals share opinions and feelings on various topics across platforms. The surge in further digital activities and interactions such as web searches, online shopping, or product reviews, complements the ‘digital exhaust’ of consumers and presents an opportunity to collate meaningful data points in real time. The nature of such data is diverse and may be numerical (e.g., like-count), textual (e.g., user comments), and/or visual (e.g., photos). After all, AI-methods can leverage merely all these types of data to inform decision-making and streamline operations in the marketing and sales domain.

This article illustrates the usefulness and multifunctional nature of various AI methods with precise application examples and further resources across a variety of marketing and sales tasks. These methods are particularly powerful as they can cope with large data sets and may be adapted, combined, or extended to fit respective contexts and firm-specific tasks. Our selected examples span across Deep Learning (DL), Natural Language Processing (NLP), and Computer Vision (CV).

AI Supporting Market Intelligence Tasks

Market intelligence insights through AI enhances marketing activities and decision-making processes by offering faster, cost-effective, and more accurate analysis and predictions about customers and the market. Particularly, the automated process of identifying, structuring, and prioritizing customer needs drawing from vast amounts of data streamlines marketing tasks that profit from granular market intelligence.

Sentiment analysis refers to the process of eliciting the customers’ positive or negative perception of, e.g., content, a brand, or specific product attributes from both textual and visual data. Employing AI allows marketers to track brand conversations, derive communication insights, monitor trends, and gain a comprehensive understanding of consumer perceptions towards their own brand and competitors’ brands in real time and with higher accuracy. As sentiment analysis often requires a nuanced understanding of dependencies across words in a sentence, NLP-based text classification is an advantageous approach for this task. Given the complexity of language, employing advanced deep learning models (e.g., Transformer-based models like BERT or RoBERTa) provides a more refined and accurate output, as contextual relationships between words can be inferred. Similar to text classification but with a different data underlying, CV-based image classification can help to retrieve customer sentiment from visual data.

Customer segmentation, also known as customer clustering, is a prevalent marketing task that involves classifying customers into distinct groups based on demographic factors, psychographics, social identity cues, or other relevant consumer characteristics. Accurate and meaningful customer segmentation based on AI results in improved targeting and an enhanced approach to customer relationship management (CRM), among other benefits. Once again, multifunctional NLP-based Transformers such as BERT or RoBERTa may be leveraged to segment and categorize customers based on similar attitudes toward products, attributes, etc. Traditional approaches like k-means clustering and novel artificial neural network-based approaches such as Self Organizing Maps are also possible approaches.

Marketing Campaign Design is another task in which the advancements in AI methods, particularly NLP and CV, can provide marketers with substantial support. By eliciting content elements that increase engagement metrics campaign success can be optimized. To elicit such campaign success drivers, marketers can review typical performance metrics such as likes or comments with specific features of the respective content (e.g., The Power of Brand Selfies).

Personalized content curation is another marketing task benefitting from recent advancements by AI. Curating content is dependent on a firm’s ability to personalize, i.e., the ability to differentiate between consumers, a major strength of AI. Employing AI, deep network representation learning in particular, can enable one-to-one marketing at scale, as it allows to detect preference subtleties on the user level by e.g., crawling social media activity. Implementing those learnings into content curation, AI aids in maximizing a particular outcome, such as increased user engagement, revenue, or clicks. Exemplary AI-based outputs are content recommendation systems and personalized ad sequencing models that can be solved through a collaborative filtering approach (e.g., Python Library Surprise) or Deep Q-Networks, respectively.

Automated pricing and demand learning through AI allows to charge customers varying prices in response to fluctuations in supply and demand or depending on the respective customer and product consideration set. By utilizing AI, firms can overcome traditional uniform pricing, yielding increased profits. While Bayesian, Reinforcement Learning, or Decision Tree models are well suited for automated pricing, an LSTM architecture or the Python library PROPHET may be used to learn about customer demand. Given the significant role of the price in a consumer’s purchase decision, personalized pricing might not be suitable for all firms, especially in light of high price transparency online. However, employing NLP and CV enables the extraction of pricing and demand details from data which can provide valuable insights for planning and forecasting endeavors.

Promotion optimization is a marketing task where individually personalized policies are frequently implemented. Hence, this task is particularly prone to leveraging AI-based personalization to make data-driven decisions regarding promotion targeting. Beyond Collaborative Filtering and Reinforcement Learning techniques, the Python libraries scikit-learn for LASSO methods or PyPi for XGBoost provide robust, customizable optimization tools to execute the marketing task of promotion optimization upon fine-tuning to firm-specific needs.

Comprehensive Market Intelligence forms a basis for many decisions in Strategy, Marketing and Sales organizations. AI-based market intelligence enables a thorough understanding of the market in addition to individual-level insights. Understanding market opportunities, threats, and the competitive landscape offers valuable advantages for marketers and salespeople, allowing them to operate more competitively. Similar to learnings from customer segmentation, community detection algorithms enable the creation of market maps, while the Python library NetworkX sets the basis for in-depth network analysis by providing insights into complex relationship sets.

AI Supporting Content Generation Tasks

The efficiency gains of employing AI in content generation are rooted in the automated creation of textual and visual content across marketing tasks such as advertising and CRM. Given its versatile applicability, generative AI can support the entirety of the customer lifecycle, from acquisition to relationship cultivation to implementing successful retention measures.

Large Language Models like Generative Pre-trained Transformers (GPT) have proven to be particularly powerful, as they can create persuasive, natural-sounding, and coherent textual passages that closely resemble natural language patterns. Hence, they provide rich possibilities in marketing and sales such as question-answering (e.g., for research tasks), automated conversational agents (e.g., for creation of a customer service chatbot), or text summarization and translation tasks (e.g., for in-house campaign content creation).

In addition to text-based natural language generation, creating visual content is also made possible by generative AI. It empowers marketers and salespeople to automatically create synthetic images that will communicate information to customers by complementing existing text, enriching email communication, providing visual content for social media campaigns, or supporting design tasks (e.g., product, packaging, logo, etc.). A prominent open-source software used in practice to enable the creation of photorealistic images based on text prompts is Stable Diffusion.

Managerial Implications

After all, marketing and sales managers must think strategically about AI employment concerning commercial excellence. As AI rapidly evolves, it can hold substantial sway over a firm’s market position, necessitating an iterative and continuous approach to method selection and deployment. For marketing managers, in particular, AI advancements can be leveraged for enhanced personalization of the marketing mix and the automation of various tasks, thereby improving efficiency, reducing costs, and elevating decision-making. These advances in commercial excellence are accompanied by a more nuanced understanding of consumers and their feedback, again increasing overall marketing effectiveness. Hence, managers should implement taskforces, initiate experimentation and, if proven functional, incrementally establish AI methods. The abundance of freely available resources on open-source platforms reinforces that notion, as capital constraints in adoption are reduced. Lastly, managers should be open-minded and experimental toward AI. For certain marketing problems, an optimal ready-made algorithm may not exist, yet creative model design and rigorous fine-tuning of customized models to match individual requirements will yield gains in both efficiency and financials.

Key sources and further recommended readings
  • Ding, MengQi (Annie) and Avi Goldfarb (2023), “The Economics of Artificial Intelligence: A Marketing Perspective,” in Review of Marketing Research, Vol. 20, K. Sudhir and O. Toubia, eds, Emerald Publishing Limited, 13–76.
  • Hartmann, Jochen and Oded Netzer (2023), “Natural Language Processing in Marketing,” in Review of Marketing Research, Vol. 20, K. Sudhir and O. Toubia, eds, Emerald Publishing Limited, 191–215.
  • Hauser, John R., Zelin Li, and Chengfeng Mao (2023), “Artificial Intelligence and User-Generated Data Are Transforming How Firms Come to Understand Customer Needs,” in Review of Marketing Research, Vol. 20, K. Sudhir and O. Toubia, eds, Emerald Publishing Limited, 147–67.
  • Hartmann, Jochen and Oded Netzer (2023), “Natural Language Processing in Marketing,” in Review of Marketing Research, Vol. 20, K. Sudhir and O. Toubia, eds, Emerald Publishing Limited, 191–215.
  • Rafieian, Omid and Hema Yoganarasimhan (2023), “AI and Personalization,” in Review of Marketing Research, Vol. 20, K. Sudhir and O. Toubia, eds, Emerald Publishing Limited, 77–102.
  • Schweidel, David A., Martin Reisenbichler, Thomas Reutterer, and Kunpeng Zhang (2023), “Leveraging AI for Content Generation: A Customer Equity Perspective,” in Review of Marketing Research, Vol. 20, K. Sudhir and O. Toubia, eds, Emerald Publishing Limited, 125–45.
  • Sudhir, K. and Olivier Toubia (2023), “The State of AI Research in Marketing: Active, Fertile, and Ready for Explosive Growth,” in Review of Marketing Research, Vol. 20, K. Sudhir and O. Toubia, eds, Emerald Publishing Limited, 1–12.

Leave a Reply

Your email address will not be published. Required fields are marked *