Making AI Customer-Centric

Artificial Intelligence (AI) is a collection of technologies, including machine learning and natural language processing, that makes machines behave in a more human-like way. AI is becoming more widespread across a range of industries and disciplines, and customer experience is no exception.

While AI can support a number of different customer experience activities, Temkin Group took a look at what we call AI-Driven Interfaces (AIDI), which we define as:

Digital interactions with customers that are being directly manipulated by machine learning algorithms

Ingredients for Customer-Centric AI

While many companies are excited about the promise of AI, few know the best ways to deploy this technology. Temkin Group research has shown that there are five key ingredients needed to drive truly customer-centric AI solutions:

1.  Conversational Design. At the center of a customer-centric AI model is Conversational Design, which allows companies to deliver an engaging and emotionally gratifying experience for customers. This is based on Temkin Group’s Human Conversational Model.
2. Targeted Use Cases. Companies must avoid taking a scattershot approach to AI and instead select the appropriate situations for deploying it.
3. Optimized Data Aggregation. Rather than kicking off a never-ending data project, companies need to identify and pull together the data that’s relevant to AI.
4. Responsive AI Engine. AI requires a machine-learning platform to interpret, anticipate, and respond to customer needs.
5. Continuous Tuning. AIDI implementations are not meant to be perfect out of the box, which means companies need to commit the resources necessary for it to learn and evolve over time.

Applying AI in the Right Places

Identifying appropriate use cases for CX-focused AI is a challenge for many organizations. They often focus in the wrong areas, such as cost-reduction (e.g. cutting back contact center staff) or scenarios that are not appropriate for AI (i.e. highly emotional or overly complex interactions). Here are some criteria to keep in mind when identifying the most appropriate AIDI use cases:

• Focus on high volume interactions. Companies need to develop a method to determine the most appropriate use cases for this technology. Identifying high volume interactions is a good place to start. For example, one telecommunications company wanted to automate subscriber contacts, but it first needed to develop a comprehensive list of the reasons why subscribers contacted the company. Further, for each reason or “intent,” it needed to decide which ones AI could address. By analyzing chat transcripts, the company identified 150 different intents that drive customers to seek assistance. The telecom then prioritized the intents that drive the most call volume. Often the intents corresponding to the highest call volumes were easy to answer, making them ideal candidates for automation.
• Support contact center agents. AI plays an important role in helping agents provide quicker, more efficient responses to customer concerns. Companies can use AI to analyze the voices of both the customer and the phone representative during a call and then provide real-time guidance to the agent. For example, one regional health plan’s call center performed well when it came to providing accurate information to customers, but its satisfaction ratings were flat due to a perceived lack of agent empathy. To help demonstrate more empathy, the company uses AI to provide agents with visual cues via their computers that advise them on how to communicate better to build rapport. For instance, the system notifies agents if they are speaking too quickly or interrupting the customer. As a result, agents connect on a more emotional level with customers, listen more effectively, and feel more confident on calls.
• Respond more quickly and accurately. One of the oft-touted benefits of AI is its ability to respond 24 hours a day, seven days a week, leading to more timely responses. One bank had a service-level agreement (SLA) promising to respond to customer emails within 48 hours. However, meeting these obligations was a challenge as it meant going through 25,000 emails manually every day to understand each customer’s intent. Additionally, 60% of the initial routing had to then be re-routed due to employees misunderstanding the customers’ intent, and the bank only resolved 10% of issues in the first email interaction. The bank automated its email response and routing process so that customers received an instant notification informing them that the bank had read their email, thanking them for reaching out, and verifying the intent of the initial inbound email. By instituting this change, the bank routed 90% of the emails correctly on the first try, which increased the team’s efficiency and significantly improved the customer experience. 

AI is clearly here to stay, and it can provide incredible value in the customer experience space if it’s deployed properly. To learn more about the five ingredients for Customer-Centric AI, visit the Customer Experience Matters blog.