Finding the Right Balance: Pratibha Sharma on Implementing AI while Preserving Value of Human Interactions
According to the recent report, published by ExplodingTopics, the role of AI technology continues to grow, as more companies move from exploring and experimenting to practical implementations. Globally, 40% currently use AI in their operation, with the most common applications of artificial intelligence and machine learning being customer service, cybersecurity, and digital assistants. However, for making AI part of interaction with customers to be beneficial for the company, several common pitfalls need to be avoided. Pratibha Sharma, a software development and engineering manager, currently working at Airbnb, shares his view on the risks and benefits of AI proliferation.
The Importance of Human Input
The main benefit of implementing AI-based chatbots as an element of customer service is the automation that allows processing of user requests faster and frees human resources from answering routing and frequently answered questions. Chatbots can provide the answers 24/7, so the customers do not have to wait, no matter what time they send their inquiries.
However, there is an important principle more and more companies have started to understand: to prevent customer dissatisfaction, you need to find a balance between human input and automation. “You should carefully and thoroughly analyze the interactions with customers to find out which of them can be delegated to chatbots and which still require human agents with their understanding, emotional intelligence, and problem-solving skills,” explains Pratibha Sharma. For instance, if the issue is rare or requires a personalized approach, humans will resolve it more efficiently. Moreover, the mere fact that they are talking to AI while having a pressing issue can be perceived as discouraging by many customers.
A practical example of finding the right balance when implementing chatbots in customer service can be found in Pratibha Sharma’s work at Amazon, where she took the position of Software Development Manager in 2019. For the next couple of years, she and her team had been continuously developing and improving a chatbot that helped with resolving customer issues. She led the implementation of the customer-facing automation products from proof-of-concept to the final version, performing multiple iteration, analyzing customer feedback and finding an optimal way to implement automated solutions. Eventually, it took over a significant part of customer contact volume, lowering operational costs for the company, reducing customer effort in resolving issues at the same time. She comments that one of the key aspects of achieving this success was finding a realistic balance between automation and human input. “While routine tasks could be automated easily, thorough testing allowed us to determine the edge cases and non-standard issues that are better resolved through human participation,” she adds. “Often, when the interaction between customer and the chatbot is not successful, human support should step in, and it is necessary to quickly recognize such cases to prevent customer dissatisfaction.”
Ensuring Cross-Team Collaboration
Another common mistake is developing various AI implementations in the company in isolation, focusing on narrow goals too much and siloing the teams. AI-based tools are often flexible and can be adjusted for performing different tasks. In addition, established interactions between teams and departments create an opportunity for a more integral view of the company’s operation, helping to discover the most efficient AI application.
Pratibha Sharma mentions that at Amazon she took part in organizing strategic partnerships between Product, Data Science, and Machine learning teams, which eventually resulted in the successful launch of the chatbot in multiple countries with over 2 million weekly visits. Speaking of various countries, considering local specifics, such as cultural and language differences, is also important, as customers’ expectations can vary by country. The efforts of Pratibha Sharma, including designing and executing A/B tests and improving machine learning models played a crucial role in creating a product that caters to the needs and preferences of customers all over the world. Even the difference between American and British English can lead to misunderstandings in some cases, and with the chatbot that is expected to operate in multiple countries the task becomes even harder, as machine translations still require quality control, for example, to avoid misinterpretation of idioms or emotional expressions. Thorough testing allowed Pratibha’s team to catch such issues early, eventually creating the service that is available for users in the United States, United Kingdom, Germany, Japan, and India, successfully providing Amazon clients with the support they need.
Following an Integrated Approach
One more common pitfall of AI implementation worth mentioning is focusing exclusively on the technical aspects of the project. Whether the AI application in question involves customer interaction or is more focused on internal business processes, such as inventory management or accounting, its development will require a cross-functional team, which makes communication and emotional intelligence skills no less important than technical competence and decision-making.
“For productive teamwork, especially in rapidly evolving domains such as AI and ML, technical skills should be supported by efficient communication, adaptability, and the ability to understand other people’s perspectives”, explains Pratibha Sharma. She adds that she has multiple examples of this principle at work while she was a member of a jury board at Globee Awards, a global business award that honors organizations from around the word for their achievements in business and technology. Globee Awards consists of several programs, dedicated to various industries and domains, such as Customer Excellence, Cybersecurity, Leadership Awards, with each of them including over a dozen of nominations. This year, the Artificial Intelligence Awards program was introduced to celebrate achievements and innovations in the rapidly developing AI industry. Depending on the nomination, companies of various scales can be seen among the award winners, from tech giants such as Amazon or IBM to smaller local companies who manage to disrupt the market. Pratibha Sharma notes that teams able to enhance their technical knowledge with soft skills and human interactions are often seen among the award winners. She finds it important to support such teams as a member of the award’s jury board, as well as nurture this approach in the team she leads herself.
Naturally, while customer service remains in the top position among the most frequent AI applications, the opportunities opened by the technology are not limited to that. Another application of growing significance is using AI for cybersecurity and fraud prevention. Advanced methods of user behavior analysis often allow companies to act proactively, catching fraud attempts before they cause harm and financial loss. Staying on the leading edge of technology, Pratibha Sharma is currently a leader of the team doing just that. Since 2023, she has taken a position as Software Engineer Manager at Airbnb, building a trusted platform for online fraud detection, mitigation, and enforcement. She concludes that the principles laid out above are applicable to various domains of implementing AI into business processes and help to employ the technology efficiently, without losing the company’s identity and human values.
