Chatbots have made many a transaction more efficient. They’re great when customers need a quick, simple answer like, “What’s my balance?” On the other hand, if you’ve ever had a frustrating “conversation” with a bot – online or on the phone – you know that bots have limitations on their own. That begs the question: How can businesses create better bots and make the handoff to live agents more seamless, and less painful, for the customer?
Experts in bot development agree that efforts to create a satisfying experience must begin with the customer, which is why Mitel partnered with Google to power intelligent customer experiences. While doing so may seem obvious, how often do we actually build systems that fit our processes?
What we should always do is start with the user. Talk to your agents to identify the questions customers ask. What are their most common issues? Then, analyze contact center interactions that customers have rated the highest. In other words, start by doing your homework.
The truth is, it’s easy to lose sight of the customer as exciting technologies like artificial intelligence (AI), conversational AI, machine learning (ML) and natural language understanding (NLU) emerge. But as promising as these technologies are, building a solution begins with a roadmap of the customer journey. Contact center managers need to set guidelines for what bots can do and when they need to pass conversations over to a human.
Issues that are time-critical or require sensitive information should be handed off immediately. For instance, when someone says, “I’ve just lost my train pass and I need to board the next train. What do I do?” they don’t want to chat with a software program. That’s why AI developers are enabling bots to notice context using words that express emotion, urgency and frustration. In building any bot-human interaction, designers must set markers like these to direct the bot on when to handoff a call.
Central to a successful experience is giving customers control over when to switch to live agents, ideally at any point in the contact process. Many businesses let customers know from the start when they’re interacting with a virtual agent and explain how the bot can assist. However, providing a quick out, like a “chat with agent” button, is a good practice. Before the handoff, though, be sure to capture and transfer any information customers have already shared. It’s maddening to have to start all over again when the live agent takes over.
Successful bots are trained by successful humans. No matter how sophisticated the software, its performance depends on what information it’s fed. (No doubt you’ve heard the expression “garbage-in, garbage out”– or GIGO.) As they stumble, bots need to be retrained. Identify where bots lose the customer, such as when the call ends, or the customer requests a live agent. Was the bot able to understand the customer? Could it answer the specific question being asked? If not, did it quickly transfer the customer to a live agent?
To start, analyze the interactions that have been successfully handled by human agents. Then use your best agents to help train your bots. Nowadays, bots can even learn from customer conversations with human agents by “listening” to conversations in real-time and learn how human agents respond to what customers say on the phone or type in the chat box.
As Mikhail Naumov, co-founder and chief strategy officer at DigitalGenius, explains in this Tech Target post, bots with AI and machine learning can learn over time when humans provide ongoing training. “The more often a human agent helps train the AI by affirming or correcting the outcome, the more AI – and its algorithms – will learn and make confident decisions on its own.”
Will bots entirely replace humans? Definitely not. But even so, the near future promises much more sophisticated bots thanks to NLU and conversational AI. Imagine bots that can understand the intent behind a customer’s request, solve the immediate problem, then answer follow-up questions and learn how to improve its future performance, saving time and money while bettering the customer experience.
“We need to find a human-centric solution that’s ‘conversationally intelligent,’” explains Kathryn White, Client Innovation Services Lead at Accenture. “A chatbot that not only understands how humans speak, but also senses and interprets tone and context.”
Within the next year, White predicts, bots will be able to take relevant actions in the proper context without the need for human intervention. The challenge that remains is integration across diverse AI interfaces and ecosystems, but once this is resolved, she explains, we’ll see bots that can draw on previous interactions, in essence “remembering” context and responding to the customer based on this knowledge.
As Mitel’s Director of Solutions Ryan Smith writes about in his six-part blog series on artificial intelligence and collaboration, “There’s no doubt artificial intelligence is transforming the very concept of work and, when combined with effective collaboration, the possibilities are endless.”