December 14, 2012
Siri, Apple’s voice-activated virtual agent (VA), has raised the profile of this technology category. Siri provides the right engagement paradigm: ask a question and get an answer – a right answer. Because of Siri, companies focused on increasing customer satisfaction scores to move the needle on customer loyalty often ask “Why can't we offer Siri-like experiences on our web or mobile sites to help customers ask questions in their own words?”
Let's look at the facts: customers today are trained to go online to get answers to their questions by navigating a company’s FAQ list, or by typing in keywords to surface the right piece of content. In fact, 66% of customers use this channel. But at 51%, the satisfaction ratings for this channel are the lowest of all the communication channels that Forrester tracks. It's because keeping content in line with customer demand and making it easily accessible to customers is very hard to do.
Enter the world of virtual agents. These solutions use natural language processing and artificial intelligence to greet a customer and to serve up the right answer to their question, which is asked using a customer's natural speech pattern, not just keywords. These answers could be a straightforward response, knowledge base content, data (like "Did my check clear?"), or the result of a transaction (like "Book me on the next flight home out of San Francisco”). With the right tuning and integration to back-end systems, virtual agents are able to personalize and contextualize customer interactions. In fact, Gartner Research predicts that by 2015, 50% of online customer self-service search activities will be done by virtual agents.
But virtual agent technology is not completely rosy. It takes continuous monitoring of the success of conversations and tuning algorithms for continued success, which can be a heavyweight task and is a job that should be pushed on the vendor to do.
Here are some best practices if you are considering virtual agents:
- Align your virtual agent with your brand. Virtual agents use graphical representations of characters as their interface – like Anna at Ikea or Jenn at Alaska Airlines. Make sure that you choose an avatar that is well aligned with your brand proposition.
- Understand the conversations that customers want to have with you. Make sure that you have answers to commonly asked questions and that you understand the scope of dialogues that customers want to have with you.
- Build out a corpus of industry-specific terms. Many virtual assistant vendors will have large corpuses of industry-specific terms that you can leverage to improve the success of conversations.
- Integrate with back-end systems. You may need to integrate with knowledge management systems or with transactional systems like billing solutions to answer common questions.
- Monitor the success of virtual agent conversations. You need to measure the success of each virtual agent conversation and use these results to continuously tune your algorithms to ensure success.
- Consider the user experience if the interaction fails. Virtual agents' interactions can't be a dead end to a conversation. Make sure that your customers are able to escalate to live-assist channels like chat or a phone call to get a quick answer to their question. Make sure that a record of the interaction is passed to the agent so that the customer does not have to restart her conversation.
- Leverage the expertise of vendors in this space. Code Baby, Creative Virtual, IntelliResponse, NextIT, NoHold,, and VirtuOz all have marquee deployments and years of experience.