AI applications are now blooming everywhere, and AI trials and consumer-oriented AI integration are actively being carried out everywhere. AI exhibits increasingly powerful insights, and can also provide valuable insights in extremely complex marketing scenarios, assist human decision-making, and complete customer-oriented tasks. So, what can AI do in marketing?
AI in marketing
One is the automation of digital marketing-when, where, and how to invest resources in email, social media, and other forms of marketing? To answer this question well, guessing alone will not work. Many companies are building AI systems to automate decision-making and make them more strategic.
The second is dynamic pricing-AI can realize automated pricing decisions based on product digitization, user dataization, combined with historical pricing and current market competition.
The third is content marketing-marketers always want to figure out why certain types of content are better than others. AI can disassemble content and perform effect evaluation, and predict its performance more accurately through image, text, and audio analysis (speaking speed, tone, expression, etc.). Therefore, AI can assist marketers in testing content strategies and selecting the best content marketing design in real time and dynamically.
The fourth is product development-as competition intensifies and product life cycles shorten, companies are facing unprecedented pressure to achieve success in innovation and new product development. AI can analyze big data from social media, e-commerce platforms, etc., to mine future market trends, and combine past sales performance to optimize product design, shorten development time, and improve the company’s product innovation capabilities and success probability.
The fifth is customer experience-more and more brands are trying to operate AI-enabled physical stores. For example, Alibaba has opened an AI fashion concept store in Hong Kong. Customers in the store can log in by scanning the code or swiping their face on the mobile terminal. The clothing in the store is equipped with electronic tags. When a customer picks a piece of clothing, the relevant information will be automatically displayed on the nearby smart mirror screen, and clothing matching suggestions will also be given. Employees can also use the AI system covering the entire store to obtain real-time information, so as to better improve inventory management, innovate customer experience and enhance customer stickiness.
AI solutions compared with real people
But is AI really more reliable than real people? Comparing AI solutions with real people in the same marketing context, is their performance satisfactory? Let’s take a look at the following 4 related experiments.
Experiment 1: AI coach or human coach?
This research involves a financial technology company with 3,500 employees, 19 million customers, and $3 billion in sales revenue. The company provides small-scale short-term low-interest loans to individuals. Once the customer has repaid the loan, the salesperson will call the customer to encourage him to take another loan. In order to improve the sales performance of agents, the company hired professional coaches to provide feedback and suggestions. Although this method of training is effective, it is expensive and time-consuming, so the company has created an AI coaching system as an alternative.
In the field experiment, 429 salespersons were randomly assigned to humans or AI coaches. In general, the performance of AI coaches has surpassed that of human coaches. All three types of salespersons with different experience have achieved performance improvement under the guidance of AI. Surprisingly, the group that has made the most progress is the mid-level salesperson, not the junior salesperson. After investigating salespeople, two main problems with AI coaches surfaced.
Compared with human marketers, AI solutions have their own advantages. Some advantages are based on powerful processing capabilities, while others are derived from consumer biases and preferences out of human instinct.
One is that AI solutions are not as adaptable and context-sensitive as human coaches in communicating suggestions. Although the three levels of salesmen feel that the AI coach’s feedback is more breadth and depth, the junior salesmen (with the least experience) feel very strongly about the AI coach’s “feedback overload” problem: the “feedback overload” that the AI coach makes them feel “Twice as much as a human coach.
The second is that human aversion to AI interaction is a key factor limiting the effectiveness of AI. Compared with human coaches, senior salespersons show stronger resentment towards AI coaches. Perhaps part of the reason is that they feel that their performance is already the best in the company and no further training is needed.
Subsequently, the company tried a cooperative coaching model: the AI first made recommendations, and then the human coaches relayed it to the salesperson. This hybrid approach has successfully alleviated the aversion and “feedback overload” problems. From the perspective of improving salesperson performance, this method is nearly 2.5 times more effective for junior salespersons and 3 times more effective for senior salespersons.
This experiment shows that humans are better at adapting and adjusting interpersonal communication skills than AI solutions. In addition, the hybrid approach helps to realize the full potential of AI.
Experiment 2: People’s antipathy towards AI
The company then tested the effectiveness of AI solutions in making sales calls, and wanted to see if AI could outperform human salespeople in winning new loans. At the same time, the experiment also tested the reaction of consumers when interacting with AI.
The experiment is divided into 6 groups, each group makes about 1,000 sales calls. The first group of testers are junior human salespersons; the second group is senior human salespersons; the other four groups of salespersons are all AI, the difference is that the third group did not disclose to consumers that they are AI before selling, and the fourth group consumes Before the start of the promotion, the participants were told that the communication object was AI. After the communication between the two parties, the fifth group revealed that it was AI before consumers decide whether to re-loan; the sixth group revealed that it was AI after the consumer made a decision.
In the experiment, the probability of the fourth group of consumers hanging up was as high as 56.3%. Except for the fifth group, which was 4.5%, the probability of being hung up in the other groups was 0%. In addition, the duration of the 4th group is much shorter, only about 10 seconds, while the other groups are 40-60 seconds.
From the perspective of sales and purchase rates, AI’s performance is the best when consumers do not know that they are talking to machines. The performance of Group 3 (23.7%) is almost comparable to Group 2 (25.1%), and far surpassing Group 1 (4.49%). However, the performance of the fourth group is much worse, only 4.48%, nearly 80% lower than the third group.
The results show that in complex consumer-oriented sales tasks, AI performs almost as well as experienced humans. But it is also obvious that many consumers are still wary and even uninterested when they learn that they are interacting with AI.
Experiment 3: Are humans biased against AI?
In applications where AI directly provides purchase advice to consumers, it is also necessary to take into account the deep-rooted perception biases of human beings. This experiment studies the comparison between AI systems and human decision-making in functional evaluation and experience evaluation. In this study, consumers have to try two pieces of chocolate cake and make an evaluation. They were told that the ingredients for one piece of cake were selected by human chocolate masters, and the other piece was selected by AI. But in fact, the two cakes are made in exactly the same way.
When consumers evaluate the taste and taste of cakes and other experience characteristics, they instinctively believe in humans, and they always give higher scores to the “Master Chocolate Cake”. In contrast, when evaluating the freshness and health of the cake and other functional characteristics, consumers gave the “AI cake” a higher score.
This shows that in the eyes of consumers, AI solutions are superior in the evaluation of functional features and design choices, but not in terms of experience features. People believe that AI is better at processing data, and therefore can make reliable recommendations for features that are easy to measure.
Another experiment on winter coats also showed this perception phenomenon. The tested consumers are more inclined to accept AI’s purchase suggestions on functional characteristics (warmth, windproof, and breathability), but when it comes to style characteristics such as coat color, beauty, and fabric feel, they are more willing to accept the choice of human salespersons. Purchase recommendations.
Experiment 4: Human intuition is stronger?
Inventory management is crucial to the auto parts industry. Managers must continuously make judgments among hundreds of SKUs (stock keeping units) to determine which ones will sell well or fast, and which ones need to be eliminated as soon as possible.
In this study, a large auto parts company established an AI system that determines whether its sales prospects are good or not based on the variable data of each part. These variables include: the number of registered cars using the part, the average number of failures, historical sales/price data, and so on. Based on this comprehensive analysis, the system makes sales forecasts, and gives out which SKUs will not sell well in the future and recommends that the company quickly eliminate these parts.
The AI solution was piloted for one year, covering 30731 SKUs, and its effect was compared with the performance of the company’s human managers. On average, AI outperforms human managers: the former improves gross profit margin by 5.77% higher than the latter. However, when making decisions on the parts of a new car (vehicle age less than 5 years), human performance is much better than AI, with a gap of 23%. This may be because the AI system has not collected enough data, and thus cannot analyze car performance and market behavior with sufficient information. This theory is also confirmed by the following results: in the mature car category (5-10 years), the performance of AI is 7.82% better than humans, and in the aging car category (over 10 years), the performance of AI is 17.89 better than humans. %. The reason is that these sub-categories have enough data for AI analysis.
This example shows that when humans rely on instinctive “intuition” to analyze situations that are difficult to quantify or rely on hard data, they can completely outperform AI. It also shows that AI can provide reliable and accurate market forecasts by interpreting a large number of complex data sets, and this ability is getting better and better.
AI capabilities continue to expand
Through the above four aspects of research, we can find that AI solutions have their own advantages compared with human marketers. Some advantages are based on powerful processing capabilities, while others are derived from consumer biases and preferences out of human instinct.
Human beings are more adaptable and better at controlling interpersonal skills and various situations. Compared with current AI solutions, humans are better at understanding and interpreting experiences and situations that are difficult to transform into tangible and quantifiable data. Consumers are already familiar with human marketers, so they will not have a strong aversion, and their aversion to AI limits the utility of AI to a certain extent.
However, in terms of processing and analyzing data and making more informed decisions and recommendations, AI’s advantages over humans are constantly expanding, and its analytical capabilities are expanding and upgrading in ways that humans can never reach. AI has complete stability and reliability in the decision-making process, because it can work around the clock, and the quality of analysis will not be affected by illness, emotion or distraction.
As AI solutions evolve, their role and effectiveness in various marketing scenarios will continue to improve. AI’s capabilities have evolved from managing mechanical and repetitive tasks to processing analytical tasks, and later may also cover tasks that require empathy.
In the future, the possible effects of AI in physical stores and online will be greatly enhanced, and it will become more common for merchants to interact with consumers through pure AI or human-AI hybrid marketing solutions. Of course, there is a prerequisite, that is, the plan designer can find effective ways to overcome consumers’ instinctive aversion to AI. Current research shows that as people become more accustomed to interacting with AI, this aversion may gradually fade.