At the recent InterCon in Las Vegas, the opening keynote session featured two talks focused on AI applications in business. The key takeaways were: identifying the business objective is key, access to good data is a major challenge, and AI has great potential for providing value to end customers.
The session began with a talk by Neil Daswani, director of Stanford University's Advanced Cybersecurity Program, titled "The Voice Of Customers Answered By AI." Daswani's talk focused on the use of AI in corporate Voice of the Customer (VOC) programs, highlighting the evolution of VoC and outlining future trends. Daswani was followed by Gary Neights, senior director of product management at Elemica, with a talk titled "Machine Based Learning: Mimicking The Working Of Human Brains." Neights described several challenges encountered and lessons learned when applying AI to Elemica's supply-chain management system.
Daswani opened his talk with an introduction to VOC programs. VOC, first described in the 1990s, is a marketing research process that identifies, organizes, and prioritizes customer feedback and desires. At first, VOC focused on garnering direct feedback from customer interviews and surveys; net promoter score (NPS) is common example. With the growth of the internet, companies also began incorporating data from the web, such as user comments or online reviews. Now, with the rise of machine learning and AI, companies are using these technologies to gain more insights about customers, and especially understanding and personalizing the full customer journey, from awareness to loyalty.
Daswani also called out several examples where companies have used AI to improve the customer experience. One example was the possibility of reducing the risk of a data breach. Daswani, a co-author of a book on data breaches, noted that almost 80% of customer service/experience professionals believe that AI can make interaction with customers more secure, since employee error is a major root cause of breaches. In addition, AI is immune to the temptation to "snoop" on the data of celebrities or politicians.
Daswani concluded his talk with a list of future trends in VOC. First, AI is helping increase predictive value of customer insights, providing early indicators of outcomes, both good and bad; for example, predicting customer churn. Another is the increased adoption of VOC. According to Daswani, to date fewer than 45% of companies have a VOC program, and Daswani predicts that adoption will accelerate and leverage AI to increase effectiveness. Finally, Daswani predicts that AI will enable VOC programs to have a much more sophisticated view of the customer than previously possible. In particular, customer profiles and "personas" will be more dynamic.
In his talk, Neights described some of the challenges of applying AI in supply-chain management. In many scenarios, the cost of mistakes can be quite high; for example, shipping hazardous materials to the wrong location. As an analogy, he noted that self-driving cars have always been "right on the horizon," but because of the many edge cases, a general AI solution never seems to materialize. Thus, it is key to identify where AI can be used, and where human intelligence is indispensable. Most importantly, AI output requires post-processing to detect high-cost errors.
Neights concluded with several lessons learned. The first is to identify the business objective, and determine if AI is the right tool for that task; it may not always be. The next is that access to data, in particular clean data, is a major challenge. Any data that is manually entered by humans, for example, street addresses on a purchase order, need to be normalized. Finally, customer expectations must be managed; many customers think that AI system continually improves automatically, whereas most systems need to be re-trained when the nature of input data changes.
