Insights

Overcoming Abandonment and Driving Adoption with Conversational AI

Vijayalakshmi Iyer Ph.D., Head of Data Science & AI, UST Product Engineering

Conversational AI uses dialog management systems to engage with users using voice and text to answer questions. Text-based dialog systems have been in use for some time now with varying degrees of success.

While some businesses rushed to deploy patchy conversational AI resulting in frustrated users, the recent success of ChatGPT may have heralded the arrival of maturity for the technology. However, the key to success lies in understanding the complex levels of conversation that need to be built into such dialogs, which seem effortless when humans speak with each other.

Despite the many misses and few hits, conversational AI is now gaining more ground in almost all walks of life, thanks to the rapid adoption of 4G and 5G-enabled mobile phones, wearables, and artificial reality and virtual reality devices.

While the adoption curve for conversational AI ascended rapidly in the earlier stages, it has dipped. But we have reason to believe that it will only pick up again rapidly as we make more inroads in technology, design and approach. Be it text or voice, for conversational AI to be successful, we need to minimize the abandonment rate and drive adoption. But what does this translate to in terms of technology?

DESIGNING CONVERSATIONS THAT HOLD HUMAN ATTENTION

As an AI expert would tell you, Conversational AI is far from simple. Just feeding the system terabytes of conversational data does not guarantee its success. Human interaction is a complex process that we’ve mastered over millennia. Training even the most advanced computer to understand the subtle nuances is easier said than done.

For Conversational AI, the complexity also stems from language understanding, conversational flow and integration. We tend to ask questions in two ways – one that may need a simple response (time of the day, weather, etc.) and another that may need reasoning (an issue with a product or service that isn’t listed). If not addressed in the right form or at an acceptable speed, users are prone to abandoning the conversation immediately and are reluctant to use it, casting a shadow on its projected outlook from a business perspective.

THREE FACTORS THAT INFLUENCE ENGAGING CONVERSATIONAL AI

Much like any human conversation, the key aspects that influence the adoption of Conversational AI systems are to select the right topic (selection of the right use case), keep the conversation personal (human-centric conversation design with minimal or nil human interaction), and track the conversations for its correctness. (a data loop to continually track and improve the performance of models and the systems).

• Right Use Case: Picking the right use case is key to the success of any Conversational AI implementation. Any dialog system can be informational and transactional. Carefully selecting the topics to be addressed based on the volume of intent and variety of all the intents becomes critical because of the non-deterministic nature of the solution. These two metrics directly impact achieving business objectives and reducing abandonment rates.
• Human-centered Conversational AI Design: This uses understanding human dialog patterns and acknowledging the limitations of technology to provide a seamless way for human-machine interaction. Strong copywriting is critical in making the conversational AI experience human-like and enables human-machine interaction at the right place. This leads to the user query getting addressed, better conversation completion, and low abandonment rates.
• Data/Feedback Loop: Capturing data to understand the behavior of the conversational AI at both the business level and the model level has to be planned earlier so that corrective measures are taken, and the customer experience is kept on track. Sentiment analysis is a critical metric for training and fine-tuning the system, which can lead to an overall positive customer experience (CX).

USING A HYBRID DESIGN APPROACH TO SWITCH SEAMLESSLY AND KEEP USERS ENGAGED

While Conversational AI may be a cost-effective and efficient way to engage users on business-as-usual transactions like keeping track of orders or providing information, businesses must build a timely switch from machine to human interactions, lest they lose the customer.

Adopting a hybrid approach to design a technical solution for Conversational AI allows users to have a machine-driven and human-driven conversation to complete the task.

For machine-driven/rule-based conversations, the personalized algorithmic recommendation becomes a key element of success by suggesting relevant topics based on users’ interests and context. And a robust natural language understanding model that can address single/code-mixed language to deal with the volume and variety of the topics and intents becomes critical. Robust NLU to comprehend multiple intents and perform quantitative reasoning improves the customer experience by providing more human-like interactions.

CHALLENGES IN DEPLOYING CONVERSATIONAL AI

According to a recent survey by Gartner, the top reasons for enterprises not using chatbots were the usual suspects: setup challenges, including training data and maintenance.

As technology evolves, the success of conversational AI systems is dependent not only on their technological prowess but also on the intuitive way it is implemented. Defining the business metric and its equivalent AI metric during the initial stages helps in giving direction to the design of the system.

Achieving business metrics (customer satisfaction, adoption rate, customer support efficiency, and customer support volume) and mapping its AI metric should be the key considerations when designing any conversation AI system. Getting this right from the word will lead to better adoption and low abandonment rates of conversational AI dialog systems.

THE FUTURE OF ADOPTION FOR CONVERSATIONAL AI IS LOUD AND CLEAR

It’s a known fact that Covid19 accelerated the adoption of Conversational AI. And looking ahead, as both enterprises and consumers prefer more digital-first services, the adoption rates will only rise exponentially.

This uptick will be governed by industries focusing on training agents, managing complex conversations, and hyper-personalization. The path forward is to move from conversant to engaging, irrespective of the industry, function, user environment, or device. And from being engaging to being sentient.