Exploring the Growing Role of Conversational AI to Mastering Chatbots

How to take advantage of the latest advancements in conversational AI and its growing role in the modern enterprise

Discover how to tailor the latest conversational AI capabilities to the needs of your enterprise. Conversational AI has advanced rapidly since its introduction and has revolutionized CX, reducing operational loads and costs while boosting customer satisfaction.

Today, with ever more advanced tools on the market, the ability to give customers real-time, humanized and empathetic service without a human agent is a key point of differentiation in CX. It is also a standard that customers have come to expect.

This article covers the latest advances in conversational AI and presents examples of how this cutting-edge technology is transforming experiences for customers.

Read this article to discover:

• The importance of conversational AI in a total transformation project.
• How customer data and insights can personalize conversations and support hand off to live agents.
• Why a cross-department collaboration is required to launch a chat-driven self-service function.
• The progression from rules-based chatbots to intelligent virtual agents
• Practical advice on how to work with multiple stakeholders during deployment
• Real-world examples of how the State of Illinois, Cash App, William Hill and
• ASOS leveraged conversational AI to transform their total experience

Content Summary

Foreword
The evolution of NLP technology
The rise of the intelligent virtual agent
How to master chatbots
Conclusion
Key takeaways

Foreword

The natural language processing (NLP) and artificial intelligence (AI) technology that underpins chatbots has advanced rapidly in recent decades. In the consumer space, this initially saw widespread deployment of rules-based bots followed by innovations in AI-powered shopping assistants, concierge bots and sentiment analysis capabilities.

The Covid-19 pandemic demonstrated to organizations in the public and private sectors that chatbots can improve customer and employee experiences while driving cost reductions and efficiencies. They can also gather data that can be utilized by agents and journey designers and they can, of course, continually refine their conversational abilities.

Such capabilities mean conversational AI tools have the potential to become virtual agents that can transform the total experience (TX) for customers, employees, agents and other stakeholders.

This article guides CX practitioners through their own virtual agent implementation by allowing them to benchmark against best-in-class deployments in financial services, education, retail and ecommerce. It also demonstrates how organizations in these sectors have utilized virtual agents to deliver opex reductions and customer satisfaction gains.

Drawing on real-life use cases, it explores conversational AI as a driver of TX transformation, looks at the development and application of sentiment analysis capabilities and explains what it takes to master chatbots at present.

The evolution of NLP technology

Natural language processing (NLP) draws on computational linguistics and computer science to facilitate human to machine communication. Along with artificial intelligence (AI), it emerged in the 1950s and soon caught the attention of scientists, linguists and mathematicians around the world. By 1964, the first NLP-powered chatbot, Eliza, was in development. Capable of limited conversation, Eliza used pattern matching and substitution methodology to converse.

Fast forward to 2011 and the world had its first voice-activated assistant, Siri. By the mid-2010s, rules-based chatbots were an integral part of the customer-service function, due to both their popularity with customers and their ability to increase operational efficiency by handling large volumes of repetitive tasks. Whether applied to chatbots or voice assistants, however, NLP has achieved much more since.

Kanishk Mehta, product leader of the conversational AI practice at Quantiphi says: “Prior to 2018, interactive voice response (IVR) was the most common form of telephony-based customer service. IVRs had long, complicated decision trees. Talking to chatbots instead of hitting buttons, however, was considered bleedingedge tech.”

When faced with global lockdowns during the Covid-19 pandemic, many enterprises and government departments saw virtual agents as a key tool to meet the demand for remote and instant customer, citizen and employee experiences.

The Illinois Department of Employment Security (IDES) saw a surge in inbound unemployment claims from those seeking help to access Federal and State pandemic assistance. To automate and streamline the demand it experienced, the decision was taken to utilize chat and telephony virtual agents to answer more than 35 common queries on state and federal unemployment benefits.

IDES deployed Quantiphi’s Rapid Response Virtual Agent on its website, which in its first two weeks effectively assisted citizens with 3.2 million inquiries. Within four weeks it had been trained in more than 100 further customer intents.

The virtual agent was also integrated with a content management system for easy update and modification of intent responses by IDES staff. It is estimated the state will save US$100mn a year through the efficiencies delivered by these tools.

Deployment case study: Illinois Department of Employment Security

• US$100mn Anticipated annual saving for the state
• 140,000 Average volume of phone and web inquiries handled daily
• 40,000 Number of after-hours calls answered on average each night

Development of the technologies that underpin virtual agents continues.

At the end of 2022, OpenAI launched ChatGPT. Defined as a generative language model it is an AI chat tool trained and designed to hold natural conversations.

Its release follows that of InstructGP, which is trained to follow an instruction in a prompt and provide a detailed response. Both use reinforcement learning from human feedback (RLHF), although ChatGPT has a slightly different data collection setup.

This game-changing technology has the potential to transform how NLP is used in CX. Some CX practitioners report they are exploring how such tools can be deployed to digitalize and automate their service operations and customer-facing front end. More advanced use cases are expected to emerge over the first half of 2023.

Chatbot trends at ASOS.com

The end of initial lockdowns did not dampen the demand for chatbots and virtual agents. They have become a central tool for almost all customer facing organizations.

Ecommerce marketplace ASOS.com is well versed in practical applications of NLP and AI-powered assistants, with multiple market-leading deployments pre-pandemic.

In 2017 it was an early pioneer of the “gift-guiding” chatbot, which assisted Christmas shoppers in navigating its vast catalog through a series of questions. The following year its proactive fashion bot Enki debuted to help personalize browsing based on a customer’s previous interactions and purchases. Powered by Google, Enki was also available on Google Assistant making ASOS one of the first fashion retailers in the UK to sell to customers through a voice assistant.

Projects like this have acted to reinforce the brand’s reputation for innovative, hyper-personalized experiences, but when similar technology was applied to the ASOS customer service function in 2020 it significantly moved the needle on customer satisfaction.

ASOS chose to streamline its service channels by introducing a single point of entry via live chat. The aim was to overhaul CX and establish a strategic step toward service automation. This led ASOS to deploy a virtual assistant to work alongside front-line advisors and tackle the customer care workload.

Joseph Vassie, head of insight and analytics at ASOS, says: “ASOS used to speak to customers through social media, telephone calls, emails and live chat. Some of these channels were more effective than others and customers were used to using the channels that were most convenient for them.”

Deeper analysis, however, confirmed that live chat was “evidently the space in which ASOS could create the best customer service experiences”.

In the first 24 months after deployment, ASOS added 50 points to its NPS, saw attrition “reduced significantly” and saw improvements in resolution rates and waiting times.

The next part of this report looks at the continued need to advance the customer and employee experience and the role chatbots can play in this as they too advance from rules-based bots to virtual agents.

The rise of the intelligent virtual agent

As NLP-powered tools become more advanced, they are embedded not just into workflows but the workforce itself, driving the emergence of virtual agents and hybrid teams.

Such benefits, however, were not widely recognized until the Covid-19 pandemic when business models and brand experiences changed overnight. The situation saw organizations laser-focused on retaining customers and employees while driving efficiencies. With their ability to handle large volumes of repetitive tasks while collecting data, the situation saw conversational AI come into its own.

As demonstrated by IDES and ASOS, today these tools are more than a means by which to support customers. They can be leveraged to transform all experiences, including those of employees and other stakeholders.

Mehta says chatbots are advancing to become virtual agents for improving customer experience, agent experience, and employee experience and provide assistive AI that can improve the health of the contact centers.

Quantiphi has worked on more than 250 enterprise-grade conversational AI solutions and has automated more than five million conversations. With this experience, it is now introducing an AI-first TX transformation platform, Qollective.CX. The latest release can be integrated with contact center software for combined text and voice driven virtual agent support.

Sentiment analysis capabilities

Intelligent virtual agents offer empathetic, humanized support to customers through the application of sentiment analysis technology. In situations involving potentially large sums of money, bots with sentiment analysis capabilities can effectively diffuse negative situations and direct customers with urgent issues and queries.

At mobile payment service Cash App, its customer support bot is augmented with sentiment analysis capabilities to understand a customer’s mood by analyzing sentence structures and verbal cues. This allows enquiries to be prioritized in a way that provides a convenient and empathetic response for customers at any time of day.

Joshua Tye, CX Network board member and senior customer operations lead for Cash App, says: “Our support bot utilizes a human-centered design process when customers are reaching out, particularly outside of normal business hours. The bot will field the enquiry and [using sentiment analysis] will assess the pain points and sentiments behind that enquiry, understanding the emotional state the customer arrives in.”

The support bot analyzes the level of emotional volatility for each customer to determine who is most in need of support. It then prioritizes the enquiries to ensure customers receive support as quickly as possible.

A similar approach has been taken by international gambling company William Hill. It sees huge spikes in urgent customer demands during events such as the FIFA World Cup or the UK’s Grand National horse race. These demands cannot be reasonably staffed by human agents.

It embarked on its first chatbot deployment in 2020 as part of a channel expansion strategy, starting with an FAQ bot that could assist with basic queries including deposit information. In less than a year it had developed an advanced chat estate that can leverage customer data to personalize the conversation and liaise with other bots to solve queries.

Capabilities include checking payment methods; knowing when and how to obtain anti-money laundering paperwork for larger deposits; and using pattern recognition to monitor word groupings and identify behavior that could signal a customer requires immediate support.

William Hill head of self-service Chris Coyle says: “It can be a very serious situation if somebody has done something they should not have done in terms of losing money or spending money they don’t have. We need to be very articulate in terms of how we spot that.

“We have seen chatbots in the past with dead ends and a lot of chatbots in other sectors and markets that make finding a person for support extremely difficult. It is understandable that the volume of customers can impact this, but our bot is very much geared to get a person to that customer straight away,” he adds.

The chat channel has been so successful, the business is turning away from email and voice communications to become fully message based. It has made significant financial savings and is now deploying the technology across its operations in Spain, Denmark, Italy and Sweden.

As demonstrated by these use cases, each industry that deploys such capabilities is looking to meet a different set of needs. This drives the need for industry specialized virtual agents or platforms.

The intelligent virtual teacher

In the age of advanced chatbots, the best bots can even become school teachers.

One higher education provider in North America required support in fielding large numbers of student requests. As an educational institution, it wanted to instill a data-first academic culture leveraging AI services to support its students and as such, wanted AI tutors that could coach students through their coursework and degree programs, convey knowledge and interact via natural language dialogue in speech and text to increase engagement on any device.

It turned to Quantiphi to design a large-scale AI-based tutor that could assist students across multiple learning contexts and domains, evaluate coursework and provide real-time feedback, creating a personalized learning experience.

Quantiphi’s solution included a tool to assist staff in tagging the content of the courses and creating a knowledge graph to empower the dialogs, questions and chat flows of the AI tutor.

The next section of this report looks at how to initiate a chatbot project and asks what it takes to master chatbots.

How to master chatbots

The market may be seeing ever-more advanced conversational solutions but William Hill’s Coyle says the first step to mastering chatbots is to “start small and fairly basic”.

From there he has the following advice:

• When building bots, utilize the knowledge of customer-facing staff. People in the quality team, frontline agents teams, Backoffice functions and support departments all understand the customers, processes and customer journey.
• Focus on complete journeys. If a question on account resets can be answered, for example, include that journey in the bot estate rather than presenting a customer who is locked out of their account with multiple, potentially inaccurate choices based on an interpretation of their request.
• Always ensure a customer can easily reach the right human agent.

Mehta says that to master chatbots, the technology should have multilingual and multi-modal support capabilities and a high first-call customer resolution rate.

Chatbots and virtual agents that are connected to a backend CRM, ERP or finance system offer a superior total experience. In the context of a contact center, this will improve average hold times, speed of query resolution, call containment rates and quality of service, driving gains in customer satisfaction and agent productivity.

Furthermore, the customer should be able to seamlessly move from speaking with a virtual agent on a web interface to conversing on the phone with a live agent without loss of context.

Mehta says: “These virtual agents have the ability to perform a contextual handover to a live agent and guide the agent to serve the customer better. As they learn over time, they dramatically improve call containment rates.”

TX transformation projects with conversational AI at their core demand multiple partners and vendors, as well as multi-stakeholder collaboration.

Typically, customers will coordinate with contact center software providers, the systems integration partners of the provider, an IT services vendor, an AI services vendor, a hyperscaler or cloud services provider and a consulting services provider. In addition, they must manage the input of all internal stakeholders in the organization. Such collaboration is extensive, but its benefits reach far and wide.

Reflecting on the results ASOS saw, Vassie says: “Attrition in ASOS customer care has reduced significantly in the 24 months since our service bot deployment. The move toward digital service has created promotion opportunities for many people within customer care and employee engagement has risen reliably over this period due to the improved culture, great business results and improvements to the customer experience.”

By engaging the right departments and connecting the technology accordingly, it is possible to build a Chat driven service function that delivers competitive results in any sector and industry.

What to expect from TX transformation vendors

A cohesive AI-led and modular total experience (TX) transformation product can bring all stakeholders together on one platform to create a cohesive ecosystem that is industry focused and accelerates time to market.

Professional services to build, deploy and manage AI and TX solutions.

Advisory services to guide the customer journey, track ROI and ensure the TX transformation program is a success.

Conclusion

A growing number of enterprises are realizing the benefits of conversational AI and advanced virtual agents as they deploy them in growing numbers to optimize CX.

Virtual agents are already part of a wider ecosystem of tools and workers and their rise supports hybrid workforces, allowing customers to be directed seamlessly from AI to specialized human assistance without adding friction to their journey.

Mehta says a successful virtual agent should deliver multilingual and multi-modal support capabilities, high query resolution rates and intelligent responses. The case studies presented in this report from IDES, the education provider, William Hill and Cash App detail real-life examples of how virtual agents are performing in live customer scenarios.

The story does not end here, however. These case studies demonstrate how conversational AI is now a core driver of TX transformation as virtual agents progress from a solitary solution to part of a wider ecosystem.

Many organizations will see virtual agents take on a new role where they support contact center agents through assistive AI during call handover, analyze and present actionable insights for assisting decision-makers and use their learnings to improve their own performance. This will improve overall contact center health, boosting productivity and reducing operational costs.

“Start small and fairly basic.” – Chris Coyle, Head of self-service for William Hill

Key takeaways

• NLP technology has advanced rapidly since its introduction. When applied to CX it has the potential to reduce operational loads and costs, while conversing with customers in real time to effectively direct and solve their queries.
• The widespread adoption of this technology is driving further advancement meaning rules-based chatbots are no longer enough.
• Virtual agents gather and utilize data from across the business to present a 360-degree picture of the customer and empower agents to give that customer the best possible experience.
• Conversational AI technology can be connected to a backend CRM, ERP or finance system to initiate a TX transformation that encompasses Customer Experience (CX), Agent Experience (AX) and Employee Experience (EX).