A virtual assistant playbook

In our many years of deploying virtual assistant technology across industries, Nuance has learned several lessons on what works best and what simply doesn’t work. From leveraging customer data to cross-channel integration, here’s what your business needs to know in order to successfully implement an intelligent virtual assistant.
Businesses should follow these best practices in order to successfully deploy a virtual assistant.

There is no substitute for experience when it comes to learning best practices, especially in an area of innovation like intelligent virtual assistants. For new technologies, trial and error is often the only way to figure out what works and what doesn’t.

A partner with experience deploying enterprise virtual assistants can help shortcut the process by applying lessons learned across various industries. Recently named the top-rated enterprise intelligent assistant vendor by Opus Research, Nuance has had the privilege of guiding many Fortune 500 clients to success. Here are five things we’ve learned along the way:

Know the customer

An intelligent virtual assistant is most effective when it customizes the interaction based on prior touch points in the customer journey. The key is bringing customer touch point and profile data together across channels – retail, website, app, messaging, social, and phone. With the underlying data, business rules and predictive models can determine when to proactively engage, which agent should be offered (either virtual assistant or live chat agent), and personalize the conversation to each customer. To be successful, businesses must be able to identify the customer as they move from channel to channel, which will depend on a combination of web tags, interaction logs, integration to back-end systems and biometrics-based authentication.

Deploy across channels

A highly functional virtual assistant requires investment in natural language understanding, dialog design, and integration to transactional systems. But once the design legwork is complete, this intelligence can (and should) be leveraged across channels.  Designing and optimizing virtual assistants built in different frameworks – one per channel –- is not an option. It simply costs too much and it’s nearly impossible to provide a consistent experience given the varying capabilities of each channel-specific bot. Deploying a virtual assistant across channels increases efficiency, consistency, and customer satisfaction.

Integrate the experience

Some client interactions warrant a human touch, so virtual assistants must seamlessly transfer to live agents when needed. Seamless integration means that the same chat window is used across virtual assistant and live chat agent conversations, and that the full context of the conversation is passed to the live agent, which lowers average handle times. Integration to live chat is a great option because it keeps the customer in the digital channel, where costs are lower and features like co-browse are available to improve the experience. Additionally, relaying context is equally important if the customer transitions from digital to voice. Knowledge of interactions in the digital channel can be used to anticipate intent and streamline the experience of dialing into the call center. Future investments in WebRTC will further integrate the experience by keeping the caller in the digital channel as they interact with a live agent over voice.

Measure KPIs and ROI

It may seem like a no-brainer, but measuring the right KPIs is key to the long-term success of a virtual assistant. Whether it’s driving conversion, average order size, transaction completion, or deflection from higher cost channels, virtual agent interactions must be measured in terms of how they affect these desired outcomes. To determine success, businesses should track what happened after the interaction with the virtual assistant, and conduct A|B testing to measure “what if” scenarios. With this kind of measurement rigor in place, it becomes possible to demonstrate the positive return on investment that can sustain a virtual assistant over time.

Optimize and expand

In the end, the most important thing to do is learn from actual interactions with your customers. What works and what doesn’t is often different from business to business, so optimizing performance after launch is critical. Conversation data should drive machine learning loops that enhance the virtual assistant’s natural language understanding. Meanwhile, cross-channel analytics are useful for identifying opportunities to improve effectiveness. Finding the right solution, in turn, often requires A|B testing capabilities to play out various hypotheses.  At other times, the solution may be to expand the areas in which the virtual assistant is trained. Either way, it’s an iterative process, and continued investment in optimizing the intelligent assistant is what separates success from failure.

With these best practices in mind, your businesses will be operating a successful virtual assistant in no time.

Let’s work together
Engage us

Consumers wants conversational virtual assistants

Widespread adoption of virtual assistants has changed the way consumers interact with their personal devices. Today, consumers expect the same natural, conversational interactions when they engage with businesses to obtain service.

Learn more

Tags: , , , , , ,

Let’s work together
Engage us
Pablo Supkay

About Pablo Supkay

Pablo Supkay is Senior Manager, DCS Product Management, responsible for Nuance’s virtual assistant, Nina. Pablo has over 20 years of experience in UX design, customer engagement, targeting and personalization. Prior to Nuance, he held product management positions at Corbis Corporation and Microsoft, where he worked on a variety of initiatives to optimize customer experience across the customer journey. Pablo holds an MBA from Rice University and a Plan II Honors BA from the University of Texas at Austin.