Turbocharging Customer Engagement with Conversational AITurbocharging Customer Engagement with Conversational AI

Artificial intelligence (AI) is going to redefine how customers can achieve success, and your CMO or CRO is likely focused, or should be focused, on customer success as a key to thriving in this digital era.

Findings from Gartner surveys indicate that enterprises are leveraging artificial intelligence (AI) technologies to greatly improve customer engagement. Eighty-four percent of respondent companies indicated they were investing in customer experience technologies in 2018 and a quarter plan to have virtual chatbot technologies deployed by 2020. Thirty percent expect to be using AI technologies to augment at least one of their primary sales processes within the next two years.

The benefits of using AI in sales and customer success programs are well supported by data and include better customer outcomes and improved customer satisfaction, as well as significant cost-savings. Providing highly effective and customized customer experiences provides an important competitive advantage, particularly as younger generations move more and more of their commercial activities online and through social media. AI provides foundational capabilities to make this a reality.

Eight years ago Gartner predicted that by 2020 85% of customer interactions would include some form of chatbot or virtual assistant technology. We’re not likely to hit that level of adoption, primarily because AI projects are challenging and finding capably skilled technologists to define and execute projects properly is difficult. Nonetheless, AI-driven customer experiences are achieving such dramatic results -- e.g., a 99% reduction in customer service times or a 10-point improvement in customer satisfaction -- that enterprises are highly motivated to move these projects forward despite potential obstacles. Most enterprises right now are trying to mitigate risks as they work rapidly to investigate and show tangible results from initial AI projects.

What Is Conversational AI?

AI as a field has been around for many decades. Machine learning (ML) is the subset of AI that has woken up in this decade to enable data-rich customer experiences. ML scientists build a mathematical model for predicting outcomes, then feed data to that model so that it improves overall predictive accuracy. This activity is known informally as “training the model.” ML has gotten very good in the last few years as the supporting science and computing power has increased.

Natural language processing (NLP) is the field of AI that allows computers to interact with human dialogue, whether spoken or written. NLP allows people to talk or write to devices and for devices to understand what that person has said or written. Like ML, NLP has gotten very good at answering straightforward questions within a constrained scope. (NLP still struggles in more open-ended domains where accuracy matters, such as in healthcare.) The subset of NLP that deals with the spoken voice is called speech recognition, which is approaching 95% accuracy.

In customer scenarios, NLP, speech recognition, and ML combine to provide services for automated call centers and digital assistants. The speech recognition module processes and contextualizes the spoken word and the ML module figures out how to handle and respond to the verbal input. These modules constantly learn and improve using new data fed into them.

Cloud computing and the availability of API libraries have provided enterprises an efficient way to build or integrate AI into their sales processes. In addition, the cloud enables multiple technology platforms and service providers to support a variety of economic models (build, buy, subscribe).

The cloud, NLP, speech recognition, and ML are the key technologies behind Gartner’s prediction that by 2020 AI will change the game for customer engagement.

Click below to continue to Page 2: Quantifiable Success Is Driving Adoption, and more

Quantifiable Success Is Driving Adoption

In a Forbes piece, contributor Shep Hyken, a customer service and experience expert, called a benchmark study suggesting that $62 billion annually was lost due to poor customer engagement “scary.” Other data suggests a third of American consumers will dump a company, after just one unsatisfactory customer service experience. That benchmark study noted above also found that 41% of companies failed to respond at all to a simple customer email query, and 99% of companies failed to follow up with the customer. For a routine pair of questions, the average response time was 15 hours. In this environment, an AI-powered response time of five minutes is a competitive game-changer when everyone else is taking 15 hours. Reducing response times leads to improved customer satisfaction and competitively better sales results.

A large number of customer contacts involve providing routine information, a simple listen-find-present activity (e.g., where is your closest location, when are you open, what’s my order status, is this item in stock) . For these inquiries, accuracy and speed are key. Contextual information, such as a customer’s identification, location, and open orders, adds to the quality and reduces time-to-resolution for the customer engagement.

Automating these routine tasks for call centers, chatbots, and email using AI improves customer outcomes. Capturing these interactions develops rich data for analyzing trends and emerging opportunities and problems to help better understand customer engagements.

Recent Zendesk research of 2,500 consumers in four countries shows that 67% of U.S. respondents have already used AI -- in the form of live chats -- social media, or texting to engage for customer service. Eighty-nine percent of U.S. consumers say that a quick response was a competitive differentiator when making a buying decision. Less than half will wait an hour for a response, and 10% will wait less than five minutes before moving on. Contrary to conventional wisdom, 98% of all respondents said they would prefer not to have to interact directly with a person.

Conversational AI might seem futuristic but in fact the technology is already here. IBM Watson, Google’s Contact Center AI, Microsoft Azure Cognitive Services, and Amazon Web Services all provide robust solution sets for customer self-service and automatic interactions. Numerous cloud providers offer pre-configured templates for TensorFlow, MXNet, Cffe, CNTK, and Torch machine learning environments.

Making It All Work

Unfortunately, putting all the needed pieces together is challenging, can be expensive to develop, and requires specialized talent for model building, data handling, machine learning, and predictive analytics. Maintaining all the moving parts and integrating with existing legacy or cloud apps is complex and potentially expensive. To simplify development and reduce costs, IT teams can adopt frameworks, but these don’t necessarily address all the technical and human complexities. As reported by Digital Journal, Gartner has noted that 85% of big data projects are failing.

In response, several vendors have created conversational AI platforms aimed at further abstracting and simplifying the process of designing, testing, and deploying AI-powered interactions. Such platforms are alternatives for companies looking to quickly deploy conversational AI that provides most of the benefits of built-from-scratch models, while avoiding the pitfalls described above. The approach is interesting as some platform solutions provide a library of apps and functions developers can drag and drop to create robust virtual agents that handle routine transactions themselves or assist human agents during customer interactions. The platform approach doesn’t require specialized AI personnel and the platform enables rapid deployment.

In summary, conversational AI offers significant benefits for customer engagement and is available now. Completely in-house solutions can be difficult, expensive, and have high rates of failure. Leveraging existing frameworks or platforms can help take the risk out of a project, simplify deployment – not to mention be less expensive.

Joseph Williams is a seasoned telecommunications veteran. At Microsoft, Joseph ran the telco partner engineering team for the 2010 launch of its Lync product. Prior to his Lync years, he was the CTO of Microsoft’s $24 billion enterprise sales division. Joseph has also had a long academic career where he specialized in telecommunications and networks. His ongoing research in the fields of decision analytics and artificial intelligence in enterprise UC stretches back to 2012. Joseph is currently the technology policy advisor to Washington Governor Jay Inslee.

Kevin Kieller is a globally recognized unified communications, collaboration and technology thought-leader, strategist, implementation leader and software developer. He and his team at EnableUC help organizations succeed with technology implementations by facilitating vision sessions, providing implementation coaching and using analytics to drive usage and adoption.

Don’t miss Kevin’s AI tutorial, “The Transformative Impacts of Artificial Intelligence on UC,” at Enterprise Connect 2019, taking place the week of March 18 in Orlando, Fla. Join him for this 8:00 a.m. Monday session! (And if you haven’t yet registered for Enterprise Connect, click here and use the code NJPOSTS to save an additional $200 at checkout.)