Across the workplace, few aspects of business haven’t been affected by the growing use of machine learning (ML) and artificial intelligence (AI). The appeal to move more work to machines is understandable: offering faster processing of what are often mundane tasks with higher accuracy rates and no complaining makes sense! The increasing use of ML and AI in customer service has been no different.
Some argue this will mean the end of humans in customer service. Others believe there will always be a need for humans due to emotional moments in service. While I tend to believe the latter more than the former (and for more than just the emotional aspect), there’s no arguing that the role ML and AI play in customer service will continue to grow. I would offer, however, that as machines and technology move into customer service and take the work of humans, new and better opportunities will emerge. Consider these three examples.
The Knowledge Base
Knowledge bases started primarily as a means of organizing and sharing information. More than simply a document library, they are focused on providing step-by-step solutions to common or known issues. At their inception, they were initially for internal use but expanded to external (or customer) use as the Internet made this possible. Paired with strong searching and filtering capabilities, knowledge bases make it easy for those with issues to zero-in on the appropriate solution.
The rise of knowledge bases accessible to customers meant there was no longer a handful of specialized knowledge only in the minds of a few in the customer service center–the knowledge base was now the brains of the operation. As solutions were found, they could be documented and easily shared. With customers using online customer service websites to access a company’s documented knowledge, this reduced the need for “live” responses by agents over telephone, email, or chat.
A knowledge-driven customer service team, however, now needed new types of employees. These employees required the skills of writers and editors. During my time managing a service center in the early days of knowledge bases, we made a conscious investment in it, removing valuable resources from the telephone lines to begin growing and curating our knowledge base. These roles allowed employees with writing and editing skills to be better utilized, while also providing documented solutions agents could use while working with customers or customers could access online themselves.
Knowledge bases were an early entrant into the concept of customer self-service. Self-service through automation is another online customer service tool allowing customers to solve common issues. Rather than calling customer service to check the status of an order, change an address, register a warranty, request a replacement credit card, etc., now such tasks can be easily automated so the customer’s issue or request is captured and delivered via workflow to the department outside customer service that can best respond or complete the work. Customer service is no longer the middleman.
Like with knowledge bases, automated self-service means fewer agents are needed to staff the live service options. But with automation came the need to keep that engine running: performing analysis to determine what were common customer issues that could benefit from automation (and those that were no longer useful) as well as developing and testing the workflows powering the automation. Once again, new opportunities were created in the wake of what might seem like the elimination of customer-facing positions.
Chatbots have been a more recent addition to customer service. They work by analyzing historical chat interactions and offer a conversational interface between the customer and AI. When customers initiate a chat, the bot recognizes keywords and phrases to deliver simple solutions or to leverage available solutions in knowledge base articles or to direct the customer to automated self-service.
Chatbots have their limits, however. AI works by recognizing a pattern that matches an available solution; they can only serve to triage new or complex issues, with limited ability to reason and diagnose new, previously uncountered problems. This makes such situations better suited for humans–as well as more interesting than the common, easily-solved problems the chatbot is in place to address. Also, as the article above notes, they are not capable of responding well in circumstances high in emotion, lacking the ability to feel and appropriately respond to the customer’s pain. In those cases, the best practice is for AI to recognize emotional keywords and other context clues and immediately connect the customer to a human.
If the only constant is change, the rate of change occurring in customer service as ML and AI will continue and even increase speed as these technologies further mature. I have no doubt a time will come when ML and AI will be the ones identifying suitable topics and authoring knowledge base articles. Similarly, automated self-service will be created directly by the system itself as it identifies patterns. And perhaps it’s possible at some point for chatbots to effectively troubleshoot new, complex issues and to have emotionally-appropriate responses. Just as in my examples, despite these jobs going to the machines, I believe new opportunities will arise for humans as we continually seek to raise the bar on customer service.