A strategic framework for artificial intelligence in marketing

How would privacy and data security affect marketing’s use of AI for data collection? When mechanical AI is used to collect and integrate multiple sources of real-time, in-context consumer data, the risk of privacy infringement and the consequences of a data breach are much higher. It is more difficult to keep such “all-in-one” data anonymous, and consumers are more sensitive to data sharing and data breaches. We have seen the public outcry when Facebook permitted the unauthorized licensing of 30 million user accounts to Cambridge Analytica, with its brand reputation plummeting dramatically in two weeks (Rust et al. 2020). In the AI age, both competition and customers become more transparent, due to AI’s data collection capability. How should privacy and data security be handled? The existing studies suggest that both data and consumer characteristics need to be considered. It has been shown that some types of data are more sensitive, such as personal data, financial data, health data, or medical records. Agarwal et al. (2020) urge healthcare organizations to find ways of protecting patient privacy as a competitive advantage, because healthcare consumers are especially weak in protecting their own data. It has also been shown that some consumers are more sensitive to data sharing. Thomaz et al. (2020) identify two types of consumers who vary in willingness to share information. They advise that marketers need to understand which type of consumers they are dealing with when providing personalization. Some methods of protecting marketing data have been proposed (Schneider et al. 2017, 2018). Future research will need to explore more delicate approaches to handle the ubiquitous data collected by AI.

How to balance data- and theory-driven market analysis? Traditional market research has the strength of hypothesis testing, whereas the emerging machine learning approach to analyzing unstructured big data has the merit of data exploration. How to balance the two approaches of inquiry and benefit from them? The Silicon Valley viewpoint is that prediction, driven by data, is everything. Balducci and Marinova (2018) show a data-driven approach to marketing management and research and Ma and Sun (2020) argue for machine learning methods for marketing research, while Lehmann (2020) proposes a method to blend theory and data in the evolving world of marketing research. These practices and studies suggest that AI will expand the role of exploration in the development of theory; thus, investigating how the two approaches can be balanced, integrated, or blended would allow marketing to leverage the benefit of data-driven market analysis.

What algorithms and models are needed for feeling AI? Much of the progress in technical algorithms has been made in the area of deep learning neural networks, applied to thinking AI problems, such as personalized advertising. The current neural network-based machine learning is mainly for prediction, rather than for understanding. We need algorithms to understand emotions and react to emotions appropriately (Rust and Huang 2020). There are different ways to understand emotions, such as understanding emotions in written language (text-based sentiment analysis), in oral conversation (natural language processing), or in facial expression (facial recognition). Although there is some work already in using thinking AI to analyze emotional data (e.g., affective analytics), with respect to facial recognition and natural language processing, models in this area are still fairly rudimentary.

How best to visualize segmentation? The emerging AI approach to segmentation can be based on an unlimited number of variables. This creates a challenge as to how best to visualize for marketers to make sense of these multi-dimensional segmentation results for targeting. Thinking AI can easily decipher such complexity, but human marketers may not. Earlier studies have attempted to visually display three-dimensional segmentation and positioning from a larger set of variables (e.g., Donthu and Rust 1989; Rust et al. 1992). There are some existing tools, but they are not designed for STP decisions. We need better and friendlier human-computer interfaces to summarize and display those extremely high dimensional results to inform human marketers.

What happens when the customer is AI? Alexa and Siri are not the only AI that can potentially buy things. Increasingly, customers use AI as their agents for information collection, price negotiation, or purchase. When it becomes more real that marketers will be catering to non-human customers (Rust 1997, 2020) or a combination of humans and AI as customers, how AI customers would behave, and how marketers should serve them are both pertinent issues. Are customer emotions still pertinent? Will rational decision making become the only route to decisions? How will AI customers change the scope of consumer research?

How should marketers and AI collaborate, to resonate with customers? Positioning requires creativity and empathy about the preferred way that a customer would like to see a product. AI can optimize a positioning recommendation, but may not be mature enough to resonate with customers, given that we do not have true feeling AI yet. Thus, for the time being, it is important to explore the best approaches to marketer-AI collaboration for a positioning that resonates with customers. This is a question that involves exploring the role of AI intelligences in creativity; for example, to what extent should marketers let machines be creative on their own (e.g., the Lexus commercial example), or use them as creative support? What will be consumers’ attitudes toward machine-generated creativity? What will decide the acceptance or rejection of machine creativity?

How is AI best used in developing new products to meet customer needs and wants? Thinking AI can be innovative and creative, but it may not be straightforward as to whether the new products will meet customer needs and wants, because many of those needs and wants are implicit, rather than explicit. This is the challenge of converting product attributes into consumer benefits. Rogers’ (1962) diffusion of innovation theory suggests that both consumer and product innovativeness matter for new product acceptance. AI, with its capability to find patterns and regularities in data that otherwise would be hard to see, can be used to uncover consumers’ implicit needs and wants, and match those needs and wants to products at different stages of lifecycle. For example, for the earlier stages of product lifecycle, AI may play a bigger role, because consumers’ needs and wants tend to be implicit, whereas for the latter stages, human marketers may play a bigger role, because those needs and wants become more explicit.

How to manage AI-based price negotiation? Negotiating price to achieve price personalization can be expected to be more common and increasingly at the individual level (e.g., price bidding). The price discrimination literature (e.g., Montes et al. 2019) tends to view pricing as a process of marketers setting the price and consumers reacting to the price. In AI-based price negotiation, the process is dynamic and real-time. We have seen the popularity for technology giants to use AI for price negotiation, such as Google’s search keywords bidding and Facebook’s ad bidding for reaching the desirable customers. What will be the new mechanisms and methods for pricing when AI is used to negotiate prices more widely?

How to manage customer disengagement due to place automation? The goal of the place decision is to provide the customer a convenience benefit. When the entire place process is automated, which is increasingly common these days, there is no human-to-human contact, and thus it is likely for customers to become less engaged with the brand. How to avoid customer disengagement while striving to provide convenience? Are there ways to mitigate the part that is lost?

How to use AI to build strong relational bonds? When two-way communications and interactions are made easier and richer by AI, how are trust and commitment affected by interacting with AI? What will be gained, and what will be lost? Customers may have changed by interacting with AI. For example, when AI does more thinking, will customers think less when they may feel more? To what extent does the advance of AI, from mechanical, to thinking, to feeling, lead to customers feeling more, at the expense of thinking, such as accepting fake news on social media at face value?