A customer service agent performs a lot of tasks to provide a great customer experience – asking questions, collecting information, entering customer data, searching for information in systems or conferring with colleagues, reading and summarizing, answering requests, taking actions, detecting outliers, and communicating with their supervisors. It’s no wonder artificial intelligence hasn’t enabled robots to do the job of a customer service agent: their work is a large collection of tasks that require skill and expertise to do well.
While computers can do calculations very quickly, the kind of reasoning that humans do is difficult and varied. Just like our brains have different centers of neurons for different kinds of thinking and high-functioning teams are composed of people with different skills, we find that a combination of different intelligent components working together is the best approach to tackling the customer service challenge.
The most powerful solution of all combines the talents of human intelligence and machine intelligence. AI can do many computations at the same time. But humans understand social situations, emotions, exception handling, and combining information from disparate sources better. This means that many different kinds of computation – both human and machine – can be done and then the results can be compared and combined to make a “wisdom of the crowds” decision where many minds perform better than a single one.
Once we’ve accepted the system works best with a human in the loop, our mission is clear. We focus on building a collection of ML-powered models that work interactively with humans to tackle customer-service problems both individually and collectively. This collection of models must be designed to operate as a system, optimized to improve the overall customer experience, and then properly leveraged by a human agent. As they work as a system, the savings and improvements from each model add together. On each turn of the conversation, the various AI models can improve understanding of the customer’s request, pull up relevant information, predict the best actions to take, and help craft a response. Each of these assists might save a few seconds, but they all sum together for the agent and contribute to real time savings in every conversation. With many conversations per day per agent and many agents the savings add up for customer experience teams. Furthermore, a well-designed system enables agents to handle more conversations at once, which leads to even further productivity gains.
All this is enabled by recognizing that a collection of minds working as a team works better than just one mind alone. The right system of tools can amplify the volume and quality of work agents do every day to help millions of customers. By building a large suite of individual AI models that target specific tasks, and ensuring they work as a system and in concert with humans to truly help, we can provide a much better customer experience than either AI or agent could provide alone.
Aaron Isaksen, PhD is Senior Director of Machine Learning Engineering and Technical Site Lead for the ASAPP NYC office. His career spans natural language understanding, embedded artificial intelligence, interactive UX, and extensive startup leadership. He has engineering degrees from UC Berkeley, MIT, and NYU and his doctoral research focused on using AI to optimize for human experience in interactive software systems.