Customer Service & Conversational UX
Using customer behavior to simplify the complexity of oral communication and comprehension.
Introduction
A large service provider in the auto industry was looking to improve how they interact with their clientele while struggling with recent labor shortages. Priding themselves on high-touch and professional service, as the shortage began to affect the quality of the client service they could offer, they decided to turn to a new strategy.
Ultimately, the business strategy aimed to take smaller and less complicated customer service tasks off of the plate of the call center employees and allow agents to focus on more complex and nuanced customer interactions. Digital efforts focused on implementing a conversational AI solution that could route calls and schedule appointments for basic services. It turned out that there was already a conversational AI bot in this role; let's call this bot Alice. Unfortunately, Alice failed to contain calls, trapped customers in loops, and could not provide clientele with service in line with company standards.
The Challenge
Let's call this new bot Lexi. In the first release, the team designed and implemented Lexi to be able to answer and direct client calls to the correct departments. With Lexi's performance meeting expectations, the biggest hurdle was cultivating buy-in from the management across the brand's locations, as Alice had created a trust deficit.
At first, the business was hesitant to devote funds to research. Therefore, research was brought into the project just as development picked up for Lexi's second release, appointment scheduling. In this situation, my research findings would be implemented as a quick follow-up to the launch of Lexi's new scheduling capabilities.
Compared to routing calls, the design for scheduling was a much larger and complex effort. The conversational design was built by working with SMEs, analyzing calls, and analytics from Alice. Nonetheless, the designs were based on assumptions and hypotheses ready for validation. Diving into development at that point was quite a risk as the conversation design needed to carry the customer through to the end of the conversation to secure the appointment and was expected to exceed three minutes. Implementing untested designs was a considerable risk, and Lexi still lacked internal buy-in, adding more pressure.
Research Objectives
Coming in to support the Lexi team with research was an exciting opportunity for me to build my skills and experience with AI and conversational UX. I quickly worked with the team to create an understanding of Lexi's end-to-end flow to schedule an appointment while fulfilling business and secret shopper requirements. From there, I worked with the product owner and the lead designer to solidify the research objectives into three main questions.
-
How usable is Lexi when guests call to make an appointment?
-
How well does Lexi understand guests?
-
What is the overall perception of Lexi?
Focusing on these three questions provided just enough feedback to understand if Lexi's feel, sound, and usability would meet the needs of customers and the business.
Insights
Top takeaways included:
-
Impressive customer service touches within the service experience, which anticipated caller needs and reduced effort.
-
Productive and efficient interactions with Lexi delighted callers. The calls inspired customer confidence, regardless of whether they resulted in booking an appointment or being quickly escalated to an agent for support.
-
Finding a balance between increasing containment and providing best-in-class customer service. Enhancing recognition and expanding Lexi's scope of service will increase containment. But an agent will always be the best way to serve the company's high-end clientele and inquiry types.
Overall, the sessions provided valuable feedback to our research questions:
-
How usable is Lexi?
-
Guests struggled to identify open appointment slots when booking with Lexi. 🚧
-
-
How well does Lexi understand?
-
Improving how well Lexi understood some clients would help avoid call escalations. 🚧
-
-
Overall perception?
-
Lexi was well-received as helpful and considerate. Often converting bot-resistant participants into fans. 🤩
-
Personalization and transparency themes emerged to enhance the customer's experience.
-
Research Strategy
I conducted this study using one-to-one remote user sessions. Since development was underway, we utilized this to our advantage, providing participants with a realistic experience by creating a fully functioning prototype for testing.
Though our sessions with participants were remote, the team gathered in person with the product owner to create an immersive research experience for our product stakeholders, who had never engaged in UX research before. Being together allowed for collaborative analysis as we collaborated to deconstruct each session and identify the main takeaways. Having stakeholders present also allowed design and research to collaborate and make rapid, iterative design changes to demonstrate immediate research value.
Our final debrief session solidified the value of research with stakeholders. Identifying research as a modern practice for cultivating informed business and product strategy, design, and development. 😍
Design
Following research's conclusion, improving the usability of scheduling was prioritized first. With design focusing on new work, I leaned into this design work using my research findings.
When negotiating appointment day and time with customers, Lexi was designed to provide two available times on an available day. Testing found that unless Lexi listed the customer's desired time as one of the two times available, they would request to move on to the next day instead of requesting the time they wanted. Customers felt that the two times Lexi verbally provided were the only two times available. For the business, this could result in the pushout of appointments, leaving unfilled appointments and causing revenue loss.
Based on the findings, I proposed two new concepts for Lexi to present availability. To test these concepts and learn more about callers' perception of availability, I designed a lean 3-day study at two of the company's locations. Taking full advantage of the opportunity, we utilized different roles from each location's staff to test these concepts, demonstrate value, and drive a grassroots increase of support for Lexi.
I facilitated the testing of three concepts using a Wizard of OZ approach, with each participant placing just as many "calls to" Lexi.
-
(Control) Lexi presented two times with slight adjustments to the language around the times offered.
-
Three-time slots are presented to the participant
-
The earliest time is provided, stating later availability upon request.
The results were energizing. Most employee participants left with excitement. Lexi's abilities exceeded expectations and eliminated hesitancy. 🙌 Usability increased when Lexi provided offerings using concept C. Participants felt more options were available and felt more comfortable requesting a time. 🎉
Lexi's first converted call demonstrated the usability concern identified in research precisely as anticipated.
Implementation
It was time to implement the new design concept into the existing conversation flow. This stage of the conversation design was by far the largest and most complex feature. I approached the redesign in layers. I wanted to insert the new design and look for opportunities to simplify the design itself. Having joined the team later, I noticed it wasn't easy to follow the existing design without having the context of being involved in the original design creation. In the future, we expected Lexi to continue to scale and for another team to take over her maintenance. Therefore, finding opportunities to clarify functionality was important.
Working in Lucidchart, I used caller behavior patterns to help group and isolate Lexi's corresponding behavior. These design decisions helped me simplify the overall flow presentation and created space for more information around specific functionality.
Outcomes & Conclusion
I worked in collaboration with the lead designer and product strategist to outline user stories, ensure all scenarios were covered, and verify the enhancements were communicated back to stakeholders for approval and development
.
As I write this case study, development is in full swing, and the release of these enhancements is expected in the near future.
Conversational UX is an exciting frontier full of opportunities for experimentation. Working with team Lexi, I utilized lean research practices to drive valuable findings for the business and create an immediate impact on the design to avoid revenue loss.