Conversational UI, the software-enabled agents that mimic conversation with a human, have grown more widespread over the past five years. Worldwide consumer retail spend via conversational UI is predicted to reach $142 billion by 2024. Their benefits are extensive, solving user problems ranging from booking hotel rooms to alerting first responders in the event of a medical emergency, while saving on operational costs, freeing staff to focus on more important tasks, and streamlining users’ brand experiences.
However, many organizations have had difficulty applying the technology successfully, and don’t know where to begin when building out its capabilities. We’ll highlight some of these roadblocks and discuss our conversational UI framework for overcoming them.
Technological Challenges to Developing Conversational UI
Conversational experiences use natural language processing/understanding (NLP/NLU), and sometimes artificial intelligence, to mimic conversations with real people in both written and voice formats. Because of this, a major challenge in building conversational UI lies in the technology’s ability to identify and understand the underlying intent in a human phrase by extracting key elements. Consider the following statement:
“Could you please [order] [coffee] from the nearby [coffee shop] to my [Concentrix Catalyst Office]”
“Order” and “coffee” show the intent of the request, and “coffee shop” and “Concentrix Catalyst office” show the entities involved. This leads us to the second challenge: the technology’s ability to recognize an almost infinite number of ways to ask for the same intent. Understanding such intent depends heavily on Natural Language Processing (NLP), and requires relevant context in order to provide accurate results or services. Simple phrases such as “I need a drink” could mean many things without context.
Structural Challenges to Developing Conversational UI
Conversational UI has matured to the point that the technology can be broken into different categories. They include:
- Chatbots: Chatbots operate on a single-turn exchange basis, such as looking up business hours or submitting an online order.
- Conversational assistants: Conversational assistants engage users in a conversation to understand the nature of a problem, such as retrieving account information.
- Personal or virtual agents: Personal agents retain information associated with a user to provide contextualized answers, leveraging machine learning to learn more about the user over time, and can help with tasks such as managing users’ calendars.
As you can see, conversational UI has become increasingly more complex than just asking a yes or no question, and adopting it requires more strategic thinking and human-centered engineering. With all the benefits of conversational UI, organizations still struggle to enable the experience. We’ve seen firsthand a few areas where organizations often need help:
- Strategy: Most enterprises do not have a holistic strategy with human-centered design and data-driven insights to systematically plan and organize conversational experience. In addition, customers leave information in multiple channels such as text, voice interaction, and operation records, but organizations haven’t established integration to deliver synergy between these. Lastly, most enterprises highly rely on vendors’ capabilities instead of accumulating intelligent assistance competence, such as AI/ML.
- Engineering: Organizations tend to lack the appropriate human-centered engineering methodology in implementing conversational experience, and they lack the technical craftmanship to centralize intent, prepare datasets, and train and manage models.
- Operations: Virtual assistant operation processes are not well designed, with a focus on basic work with high repetition. Organizations lack holistic analytic functions to monitor virtual assistance performance, as well as maintenance and optimization of algorithms and models after the production rollout.
How to Solve the Problem of Applying a Conversational UI Framework
Concentrix Catalyst has developed a differentiated approach for applying conversational UI, manifesting in three stages:
- Experience enablement: Organizations need to understand the consumer problem that needs to be addressed, as well as their competitive offering and value to their audience. What would the addition of conversational UI add to the customer experience?
- Key capabilities: In order to design, build, and run conversational UI for our clients, we have developed six key capabilities to let this happen. These include:
- Strategy: We help organizations put together a consumer need analysis, a list of business goals and objectives, and a vision and roadmap.
- Foundation: This includes conversational platform adoption guidelines, human-centered design, security policies, and a fail-fast mindset.
- Process, tools, and governance: We help organizations establish a Community of Practice (CoP), and we enable them to select tools and platforms.
- Engineering practices: We provide guidance on standards and best practices, reference architecture, security practices, and integration and testing.
- ChatOps: This includes DevOps for virtual assistants with continuous exploration, continuous integration, continuous delivery, TestOps, and release automation.
- Delivery excellence: In this final phase, we learn, refine, and optimize our strategy, and implement solution accelerators and patterns, in order to increase speed to market.
To take advantage of the benefits that conversational UI has to offer, organizations should carefully assess why and how they should implement it, keeping customer experience at the center, and come up with a clear strategy for maturation. Consumer expectations for fast, personalized customer experiences will continue to evolve, so applying a conversational UI framework should be at the top of companies’ lists for their CX initiatives.
EVP of Digital Engineering & Cloud Engineering