Â©Sergey Tarasov – stock.adobe.com
Customer service centers, whether a shipping group dealing with consumers or an IT organization working with corporate employees, has a lot of information to manage. Technology has been improving over the last decades, and the move of artificial intelligence (AI) into the real world hold promise to help.
Network management was one of the first internal groups to begin to leverage AI to better manage the vast amount of data in a more timely and accurate fashion. One reason was the limited ecosystem of networking and the more technical nature of the users. Machine learning is still fairly technical, especially machine learning (ML), and the early audience had to be able to handle the detail. Now natural language processing (NLP) and generation (NLG), are starting to mesh with machine learning to provide an interface than can help the non-technical audience.
The ML side has its own complexity. While there is much overlap in customer service systems, each company has its own way of doing things. Understanding how an individual organization works, what terms it uses, and what systems are involved is a signification challenge. That means a company wanting to provide a system can pre-train systems with large data sets, but also that companies who have their own data can train systems on corporate data.
The combination of natural language and machine learning techniques are components needed in order to leverage the power of AI to enhance customer service. It’s not one or the other, both are useful tools.
Aisera is a young company working in customer service management (CSM), and it is addressing that challenge. IT helped desks are often understaffed and overworked. That means any system that can effectively handle the most basic questions means both IT and their customer are happier to get faster resolution.
For instance, a business customer is in a conference room, preparing for a meeting. She can call or text the support system and say “I need access to the wireless network.” The system can understand her location, understand that “wireless” is the same as “Wi-Fi”, see what is available, and quickly send back connection information. The sales executive is quickly ready to present to the sales prospects, while IT isn’t called off of more complex issues.
“The ability for AI to both help in a natural language interface to people, while also performing intelligent process automation (Conversational RPA) in the background is key to modern customer service management,” said Muddu Sudhakar, Ph.D., Co-Founder and CEO, Aisera. “Whether fully autonomous for end-users or enhancing service agent tasks with AI-driven RPA workflows, both natural language and machine learning are necessary for a successful CSM system.”
As customer service systems have to work with people, the next aspect of a strong system is in the handling of customer intent. The same word can be positive or negative depending on context. It can also mean multiple things. Syntax is also fuzzier than most people think, being modified by overlapping semantics. One of my favorite pairs of sentences is the following:
· Time flies like an arrow.
· Fruit flies like a banana.
Systems need to have the intelligence to adjust meaning based on subtle differences.
While systems have more quickly gained knowledge of NLP, and it has become a “must have” in the market, NLG has lagged because of the complexity of generating sentences that both make sense and have the right sentiments. One of the first areas where NLG is being applied is in the arena of Robotic Process Automation (RPA). Business communications is more limited, more structured, than is that on open, social media platforms. That allows for more controlled access to language.
Aisera is using NLG for email responses, beginning to move away from the template systems that have been used for decades. NLG is still, today, a “want to have”, and it will take at least 18-24 more months for it to be developed, gain trust, and join NLP in wide use.
Natural language is important but only in the interface. Behind the scenes, ML is being used in order to analyze the knowledge based and, in many situations, the physical systems, in order to detect the problem and provide a solution. For instance, “the Wi-Fi isn’t working” doesn’t automatically mean the Wi-Fi network isn’t working. There could be technical issues on the laptop, in the router, or deeper in the system. The customer’s complaint must be analyzed in a wider context to provide a solution. Machine learning can provide that expertise for many cases, and the system can then escalate fewer problems to the personnel who can solve the problem – alongside the full context of the system’s analysis.
The Customer Service Sector
A key to the near term success of AI in customer service applications, is the two-fold approach found in many software systems. Companies such as Aisera have their own user interface to provide customer service. At the same time, customer service isn’t in a void. That information must integrate with other customer facing applications – otherwise business will continue to fail to have a complete picture of customers. Integration with existing CMS, EMS, and even larger CSM systems is a necessity.
The market is key, and I see this area being part of the full customer experience just mentioned. Given that, I will expect to see the exit strategy for most of the new firms to be acquisition. This is a component to a bigger picture. The companies that survive independently will have a very open architecture that allows easy communications with other vendors. Either way, customer service management is an area that is seeing significant change thanks to AI.