Everyone's talking about artificial intelligence (AI) these days, and if you believe the hype, AI is rapidly replacing humans in the contact center. There are real applications and real deployments of AI supporting customer service operations today, but also a lot of stories that are still mostly smoke and mirrors. What's real and possible for you depends on a few factors. Let's separate fact from fiction and look at the reality of AI in the customer service realm today.
Any Google search will give you a lot of complex definitions of artificial intelligence. A simple definition is the ability of machines to imitate human behavior. In practical terms, this means a computer can apply algorithms to data to predict outcomes or recommend actions and then learn from those outcomes or actions to make better decisions the next time a similar situation comes along. To break through the marketing hype around AI solutions, there are two questions to ask:
- Is it rules-based? Now that computer processing has gotten faster, applying rules to solving common customer service cases is much more efficient and cost-effective, but it's not AI. AI goes beyond rules to apply algorithms that learn from outcomes.
- Is it static? AI is not; it's dynamic, improving its recommendations and predictions over time based on feedback from previous results.
The three most common applications of AI we see today impacting the customer service application space are natural language processing and speech recognition, image recognition and processing, and recommendations. When applied effectively, the benefits are broad, from accelerated case resolution and increased contact deflection to increased customer satisfaction, improved field service, and reduced agent turnover.
Real Applications Today
The most common applications of AI today, with companies actually using, not piloting, the technology and seeing measurable results are in knowledge libraries (for both agent support and self-service), chatbots, and recommendations.
Knowledge libraries
Curating and managing a knowledge library and ensuring the most appropriate content or solution reaches the agent or customers can be a costly and time-intensive task. Luckily, it's one that CRM vendors have enabled with AI to help even those with the leanest of budgets (and lack of data scientists to manage and train their own AI models). For example, one consumer goods company applied prepackaged AI functionality to its service knowledge base to automate the curation of content, bringing the most popular results to common queries to the top on an ongoing basis, flagging areas where new solutions and content needed to be developed based on customer feedback, and taking ongoing ratings of content accessed by customers and agents to continuously improve self-service results.
Chatbots
Although many companies have been using rules-based chatbots for some time, and most of us have experienced their limitations, bots that use natural language processing and AI to learn from previous interactions deliver much better success in resolving customer queries without agent intervention, both initially and over time. Although some types of customers won't choose to interact with chatbots no matter how intelligent they are, the growing population of consumers that are used to text and SMS have embraced bots as a way to get their questions answered or problems solved at their own pace, usually from their mobile devices. One travel services firm, for example, was able to resolve 50 percent of complex queries in fewer than five minutes without ever calling on a agent, with an 85 percent customer satisfaction rate.
Recommendations
AI-driven recommendations, whether they are delivered to the agent or to the customer via various channels, are another area where companies are achieving real returns today. They're often surfaced through a self-service knowledge base or a chatbot interface when they can resolve a particular case or recommend a solution. Recommendations can also be applied more broadly in service to optimize agents (going beyond rules-based routing to identify which agents are best at which kinds of queries as they emerge) and in areas like field service, automating much of the dispatcher's job by intelligently balancing expertise, geographic location, and other factors to assign the right job to the right field service technician.
Looking Ahead at the Possible
Most real AI applications in customer service today focus on text and speech interactions, but possibilities for value in the near-term go beyond text and speech to image recognition. We've found most customers are just exploring or piloting the potential of image recognition and image processing for customer service today. While the technology is there, there's still more work to be done in prepackaging image processing to make it accessible for companies without expertise. We expect to see vendors specializing in field service, in particular, focusing on prepackaged, embedded AI capabilities for common needs like identifying parts or recognizing common faults in the short term, driving down the risk and time to value of embedded image processing.
The Too-Good-to-Be-True
Those of us who have used Amazon's Alexa know she's not replacing human intelligence yet; and the same is true in embedded AI for customer service. Although we've come a long way, anyone who tells you that you can take a set-it-and-forget-it approach to AI in any area of customer service is overselling its current capabilities. We're not going to see artificial agents completely replace real humans in the near future. What we will see is the ability to accelerate case resolution, increase agent deflection, and increase customer satisfaction by using AI to automate the tier-1 and tier-2 cases that it can resolve on its own , freeing up humans for more engaging customer experiences.
Rebecca Wettemann is a vice president at Nucleus Research.