The Future of Contextual Customer Conversations through Artificial Intelligence



We've all experienced it: making a call to a customer support hotline to have a real-time conversation with a specialist, only to be put on hold. The repetitive elevator music being played in the background makes the minutes feel like hours, leaving us frustrated, and with no immediate solution to mitigate the problem yowe're facing.

Thankfully, this elevator music will soon become a swan song of the past. Gartner predicts that by 2020, 85 percent of customers will manage relationships with enterprises with no human interaction whatsoever and 25 percent of customer service offerings will deploy chatbots to deal with customer inquiries, compared to the 2 percent that integrated this feature last year.

With that being said, studies have shown that 83 percent of customers prefer dealing with human beings over digital channels when it comes to fulfilling their support needs. So, how can vendors ensure that their customers receive quick and efficient service, while still providing the personal experience they desire? Though chatbots can't literally walk the walk of a real person, through the integration of artificial intelligence (AI) and machine learning, they're better able to talk the talk.

Ironically, chatbots often receive criticism for their perfect use of language. Although they're able to provide real-time responses, their tone often comes across as canned and impersonal, which does not bode well with customers who value the charisma involved in a human interaction. Chatbots are often the first, and sometimes the only, touchpoint that a consumer interacts with during their support experience. As such, their language must reflect the friendly, customer-centric tone that every company aims to deliver.

Although many chatbots are currently limited to pre-scripted messages, with the rapid pace in which machine learning is advancing, it won't be long before they're capable of expressive conversations that mimic live human interaction. Natural language processing might be in its early stages, but it will eventually allow chatbots to learn the nuances of tone, sentiment, and common jargon. Adding a bit more color to a chatbot's vocabulary and delivery will emulate the fluidity of a human conversation. Furthermore, the ability to identify reactions required to effectively respond to different emotional states as indicated by a customer's tone will allow chatbots to go beyond artificial intelligence and become truly emotionally intelligent.

Avoiding the Awkward Silence

At their current stage of acceptance, consumers feel that chatbots are helpful for quick answers to simple yes or no questions, but issues specific to unique business processes still require the help of a human support specialist. However, where chatbots can be unable to address a specific issue, they still play a role in connecting users with the appropriate specialist and help to equip these specialists with the necessary information the second they begin the evaluation process.

Upon the initiation of a chat, this digital tool can read between the lines of the terms and phrases used in the request, identify the unique factors playing into the problem at hand, and characterize the business processes that should be considered. This information allows chatbots to connect with an expert knowledgeable in their specific need, ensuring that the issue won't have to be re-explained once the connection has been bridged.

With the help of machine learning, the forward-looking ability to pinpoint the consumer's tone will also help seamlessly transition assistance from chatbot to support specialist. Information on customer sentiment can be picked up by the chatbot and then flagged to the specialist before initiating in conversation. At the end of the day, customer support needs to be both informationally and emotionally rich, so AI's and machine learning's ability to guide a personalized response will be just as valuable to the customer as reaching the resolution itself.

Putting the Conversation into Context

The integration of chatbots into vendors' websites is the first step, but embedding this tool into the solution itself allows for a whole new level of context. This built-in support approach allows the AI running on the back end of the chatbot to cull data based on the customer's usage and typical processes, using personalized information to guide the conversation as opposed to pre-programmed language and generalized numbers. This ensures that the solution provided is not only solution specific, but also specific to the behaviors of the customer.

With the rapid pace of innovation, digitally transforming companies appreciate fast and effective customer support to keep their companies and systems ahead of the curve. While they cautiously welcome the convenience provided by chatbots in certain scenarios, they are also still very aware of their limitations. This hesitancy is a natural part of the evolution of customer support, and I have no doubt that in 15 years, chatbots will work seamlessly with both customers and support specialists to streamline the support experience on both ends. Until then, the conversation on chatbots and customer support is to be continued, in real time, of course.


Jens Trotzky is head of artificial intelligence technology for SAP Support. In his role, Trotzky leads the integration of technologies like artificial intelligence and machine learning into SAP Support.