Virtual Assistant technology has been a focus of market analysts for decades. Back in 1998, a paper published by this author, Personal Telephony-the Virtual Assistant Market, was already focusing on companies using speech recognition to create applications that allowed consumers to access information and entertainment. At the time, this included access to such things as sports scores, weather reports, stock quotes, and so forth. Some even enabled office productivity applications, such as updating calendars. While the potential for such technology seemed very bright, it never really got as much traction as it could have; until now, that is. 2017 is shaping up to be an inflection point for the VA market. Modern Virtual Assistants (VAs) are supplanting early self-service options, such as speech-enabled IVR, or these limited personalized apps, with a plethora of capabilities that speak to the change in consumer adoption and usage of technology. Back then, speech recognition was a pure novelty. Today, it’s taken for granted. Back then, related technologies, such as Big Data, machine learning (ML), natural language speech understanding (NLU), or Deep Neural Networks (DNN) had yet to enter the consumer realm.
Interactive Voice Response (IVR), Chat, Bots, mobile apps and social media channels are now being harnessed with rich self-service interfaces that enable customers to access information, make transactions, and engage with businesses in ways unheard of even a decade ago. VAs offer new approaches to service delivery that have particular growth opportunities for customer care organizations. From offloading tedious agent workload to engaging customers with a branded avatar, VAs are showing up as key interaction channels for some of the largest brands in the world.
The field, however, is scattered and crowded. Confusion reigns over the terminology to describe these applications, and the capabilities associated with each term. Companies are often unsure of the benefits of deploying them. As such, until recently, VAs were a niche application that often fell prey to lack of funding compared with other interaction channels or applications, or were deployed in different parts of the organization as silos, not tied directly to customer care.
The good news is Bots, Messaging Apps, VAs, and home personal assistants are now being viewed as necessary weapons in the race to provide streamlined customer care. (Note: for the purpose of this series, I’ll use the terminology interchangeably, except in actual customer deployment descriptions.) Study after study shows that consumers are increasingly requesting and using self-service options and prefer it over live assistance. In Frost & Sullivan’s most recent survey, End User Priorities for Customer Engagement, Global 2017 – Navigating Digital Transformation in Contact Centers (October 2017), two of the top three applications targeted for the largest investment over the next two years are virtual assistant voice (47% of respondents) and virtual assistant web chat (44% of respondents). In fact, 23% of our survey respondents answered that they have replaced over 20% of customer service representatives with technology-enabled solutions such as AI powered robots. The study showed that this trend will continue over the next two years, with participants saying they plan to replace another 34% of agents with automated solutions.
Similar results are seen in the data from Dimension Data’s 20th annual Global CX Benchmarking report, Digital Crisis or Redemption – the Uncomfortable Truth. Of the 1337 participants who responded to the question, “What services can your organization support, now and in the future?” 13.3% said that they currently support virtual agents, yet an additional 32.7% said that they have plans to support such automation within a year. If that is not enough of an indication of interest in automation, customer service–oriented or otherwise, consider that within three months of Facebook’s April 2016 announcement of its bot ecosystem for developers, over 11,000 developers had signed up. It’s clear that providing rich self-service options can play a significant part in improving the Customer Experience and creating brand recognition for a business.
The benefits of automating portions of customer service are many. VAs can impact companies’ bottom line by reducing costs and increasing revenue. They can offload tasks from live agents, reducing the need for live resources. They can proactively contact customers to reduce calls into the contact center while increasing customer engagement. They also can hand customers off to other channels, including live agents if necessary, passing content and context along with the customer, and they can work side by side with live agents for additional assistance.
Generally speaking, they fall into a few categories:
- Bots – an application that emulates a human using either a Web chat or speech interface. Bots can range from basic apps that answer simple things or seek to entertain, to fully conversational bots with intelligence embedded into the application and integration to back-end databases.
- Messaging Bots – A bot that is deployed in a messaging application, such as WeChat, or Facebook Messenger.
- Virtual Assistants – a speech or Web-enabled application that emulates a human in a customer service environment, such as a replacement for an agent in a contact center. These can be speech or text driven, or a mixture of both.
- Avatars – are bots or VAs with a persona and “face”. Many brands use a persona on Web chat or mobile Web apps to reinforce a brand image.
- Personal Assistants – consumer-grade virtual assistants that are general purpose, such as Apple’s Siri on the iPhone or Amazon’s Alexa in the home.
The key to providing superior customer care using VAs and bots is intelligence. Intelligent VAs integrate with back end enterprise systems, and employ varying technologies such as AI, ML, DNN and Big Data, to increase intelligence and efficiency and efficacy. The most mature of these can perform transactions, be proactive, personalize interactions, handle complex intents, and follow up on activities. Intelligent VAs supplement the ranks of live customer support reps, sales reps, or customer service technicians and are typically tuned to a specific domain, such as healthcare, travel, finance, and others.
While Botmania that is all around us, it’s not an excuse for creating yet another bad channel of interaction. That is so reminiscent of the 80’s when everyone placed an unsophisticated IVR as the gate keeper to the call center. Sometimes history is not worth repeating itself. To avoid Botmania blunders, here are a few things to consider.
Infuse, Don’t Confuse
Throughout the scores of vendor/customer conversations that occur, one question repeatedly comes up, and that is, should you, or how do you, tell the customer that they are interacting with an automated solution? There has been no consensus on when and how you do so, as design is related to the type of automation and application that is being automated, and the point at which automation is needed. For instance, a chatbot can be used on a retail Web site to pop up at the point of need for a customer; perhaps when the customer has come back to the same page multiple times or put something in and out of a shopping cart. That would be the perfect time to pop up help, along with identifying the kind of help.
It is less clear when a customer is calling into a company and a virtual assistant answers the phone. Keeping in mind that when a customer reaches out they really just want to get an answer or have some action taken, it’s better to err on the side of clarity than entertainment, something that the early IVR industry missed. It’s also imperative to set customer expectations on what they can expect from the interaction. While it might seem humorous when a caller starts to flirt with a bot, having them get frustrated when they don’t know they are interacting with a virtual agent, and finding out only when it can’t complete a request, is not funny. Similarly, be careful with Avatars. Creating a visual representation of a brand versus just voice or text can engage or turn off customers.
When to Escalate to Live Assistance
Despite the predilection for self-service, interacting with VAs and bots isn’t always preferable. After all, how many times have you personally encountered a speech-enabled IVR that just wasn’t getting you to where you needed to be and you yelled “AGENT”, repeatedly into the phone to be transferred? Best practice number one is if a customer requests live assistance, don’t try to corral them into staying.
Next, is to move to live agent assistance first when the bot just doesn’t understand the customer’s request. Certainly, responding one time asking for clarification might solve the issue, repeatedly responding with “Sorry, I didn’t understand what you said” prompts drives down customer satisfaction and defeats the purpose of automation. If a customer appears to be angry or frustrated, it’s time to move on. And, if there is a policy or regulatory reason that a live agent needs to be part of the call, don’t stall. Plus, there might be opportunistic reasons for handing off to a live person. If the customer is upset and threatening to jump ship to a competitor that might not be the best time to stay with automation. When converting that “save” situation into further sales, or personally handling high value clients, there is nothing quite like the human touch.
Pick Your Battles
To minimize the risk of a failed bot, pick your battles, or rather your core Key Performance Indicators (KPIs) that you want to improve. Are you trying to improve customer satisfaction (CSAT)? Are you using the bot to offload agents, reduce average handle time (AHT), increase sales, or reduce misrouted calls? Determining this up front is essential, as developing an intelligent VA or Bot requires the requisite heavy lifting of designing the consumer front-end, and integration with back-end systems from those that house company data, or from third-party suppliers.
Test, tune, train
Just as new agents need training, bots do too. So did IVR, by the way. But that was a matter of evaluating usage statistics and customer feedback. With VAs and bots there are a lot of solutions available that learn over time and get better and better with use. So test, tune and train, and don’t forget to incorporate the customer feedback that is vital to gauging acceptance and usage.
Launch as Part of an Omnichannel Strategy
The key to omnichannel is not just in the integration of disparate silos, but also in the integration of company silos. Don’t launch a VA or bot in a vacuum. Get cross functional awareness and buy in on the deployment of any part of a digital workforce so that others can reap the benefits of deployment. In particular, the multitude of conversations logged through usage can help gauge customer sentiment, feed product plans, head off customer issues before they become worse, and allow for a continuous feedback loop for improving the channel.
Stay tuned. Part 2 of this series focuses on some of the promising results so far to date.