In preparation for the AI Summit San Francisco, we sat down with Jason Maynard, VP & GM of Guide & Data Products at Zendesk. Zendesk was founded upon a simple idea: make customer service software that’s easy to use and accessible to everyone. Today, they offer a growing family of products that work together to improve customer relationships, and can be embedded and extended through an open development platform.
Jason heads up Zendesk’s analytics and machine learning teams that build products that improve customer relationships using data. Previously, Jason worked at Deloitte, figuring out how much the next blockbuster would gross and built prototype LEDs at California Polytechnic University, San Luis Obispo. In our interview, we discuss the shifting expectations fuelling transformative customer service technologies, the importance of making machine learning software accessible to small and medium businesses, and the augmentation of personal customer service with AI.
Changing expectations of customer service
The rise of consumer-level automation, convenient smartphone technology, and other technological trends over the last decade have increased the expectations of consumers. The Zendesk Customer Service Benchmark offers extremely useful insights into the rapidly changing nature of customer service across different industries. It not only demonstrates that real-time channels (such as chat platforms and Twitter) yield higher customer satisfaction ratings, but that mobile devices are being used more than ever by consumers looking for support. Significantly, it notes that speed of the first reply is one of the biggest influencers on customer satisfaction.
“Customers are adapting to a faster pace of service,” explains Jason. “We can now order food, have laundry picked up, or get a ride within minutes simply by clicking a few buttons. These changing expectations are now applied to interactions with businesses of all kinds. More than 40% of customers expect their questions to be resolved within the hour.”
Fulfilling customer demands
“Based on our benchmark data, a large portion of businesses aren’t fulfilling these expectations,” Jason notes. “Additionally, customers expect to self-serve if it means they’ll be able to get to the right answer quickly. They only want to reach a human for certain nuanced issues or when they get stuck.”
AI-assisted technologies are able to meet these new demands in unprecedented ways. Indeed, 45% of retailers plan to add AI to implement customer service. “Currently, many of the most relevant applications of machine learning improve the customer experience by helping businesses resolve their customers issues faster, better, and more proactively by automating the resolution of common customer issue,” Jason argues. “Humans are good at many amazing things, but getting things done quickly isn’t one of them. One of the most effective levers a business has is to improve their customers’ experience by reducing the time and effort required to give them support. This is where AI can make a huge difference.”
Making machine learning accessible to all
Jason claims that Zendesk are focusing their efforts on meeting customer demands by making traditionally complex software accessible and easy to use. This is where machine learning comes in. “One of the great things about applying machine learning to customer service problems is that it allows customers to automate repetitive data synthesis tasks. As a result, an agent can be more productive and customers can get the answers they need—quicker and easier.”
“The biggest challenges around integrating machine learning with our experiences have been where agents need to interact with machine learning recommendations or scores. To a large degree, the results from deep learning models aren’t interpretable. It is frustrating for customers when they disagree with an output, but can’t understand why a model produced that result. Understanding the UX and using a model that has the right amount of interpretability is extremely important.”
Zendesk’s approach to the future of customer support
“At Zendesk, we see many applications of machine learning as an augmentation of the role of a support agent. AI technologies should work in partnership with support teams to help predict insights, provide recommendations, and automate simpler tasks—but step aside as soon as a human touch is needed. Backed by machine learning, the support agent’s role is evolving to be more strategic and productive. We foresee that the role of service organizations will spend more time designing the customer experiences and training algorithms to automate specific tasks along that journey in the future.”
Zendesk’s roots are in small and medium-sized businesses, and many of their products are designed for problem-solving at scale for these organisations. “Much of our machine learning R&D is focused on enabling machine learning applications for small businesses that either don’t have the capabilities or the data volumes to use machine learning. We think that machine learning enabled products provide opportunities for businesses to scale effectively.”
“An example of this is Answer Bot, an add-on to our Zendesk Guide knowledge base product that automates sending relevant articles back to a customer based on the questions they ask. In many cases, this response is all they need to resolve their issue and they get it instantly. As a result, businesses have more time to spend on inquiries where a human touch is important and customers get the answers they need instantly, rather than having to wait an hour or two for response.”