5 Steps to Delivering Delightful Conversational AI

You may have yet to notice, but we’re at the beginning phases of the conversational artificial intelligence (AI) revolution. This technology matured in the last year, and it’s now scalable and inexpensive enough to be widely deployed. At the same time, companies are more open to assigning their transactional contacts to virtual agents and benefiting from their elasticity, cost, and efficiency.

 

Gartner predicts that by next year, nearly one in six customer service interactions globally will be handled by AI, up 400% from 2017. That is before the battle to contain the Coronavirus spread disrupted agent availability around the world. But Gartner also expects that 40% of chatbot and virtual assistant applications launched in 2018 will be abandoned. That indicates a lot of lost time and investment.

 

Voice-based AI is already part of daily routines with Apple’s Siri, Amazon Alexa, and other digital assistants. But asking Siri a question and getting an answer isn’t a conversation. Real conversation is an art, and conversational AI is getting better, smarter, faster, and more intuitive with each day.

 

Still, the technology is in its early stages. To delight your call center customers, take these five steps to be assured of a successful experience with conversational AI:

 

1. Identify goals – deploying AI-powered virtual assistants can reduce costs for companies and improve customer service. But you’ll be more successful if you first identify your goals for the initiative so you can better determine whether today’s technology will do the job. Some of your goals can be wide and strategic, like reducing the dependency on an offshore partner or eliminating wait times. Other goals can be much more specific, like reducing customer effort by 15%, improving CSAT by 10%, or lowering cost by 20%. Once you define what drives the project, it will be easier to get others in the organization to buy in – and selecting the right call flows to automate will become much more manageable. A good goal statement will show the benefit for the organization, together with the challenge. For example, 30% of our callers abandoned calls due to long hold times. That leads to the potential for lost revenue and reduces our Net Promoter Score (NPS). If we can eliminate hold times, we can expect our NPS to go up five points, and ticket sales to increase by 7%.
2. Identify calls well-matched for conversational AI – as a rule of thumb, virtual agents are great with transactional calls, and human agents are better with more emotional ones. Virtual agents can handle tasks such as completing a product registration, changing an address, and paying a bill faster than a human, while making fewer mistakes. In general, the more complex and emotional a customer’s request, the more it requires a human agent for creative problem-solving and empathy. For routine inquiries, the virtual agent’s consistency creates more positive experiences for customers, with no wait times and faster resolution. The good news is that the bot will work alongside the human agents and will consistently document each step it’s taken. That information carries forward should the bot need to transfer the call to an agent. This way, customers don’t get locked in “bot land,” and agents don’t have to repeat any step the bot already took.
3. Gain support from IT - conversational AI isn’t something to drop into an existing call center. In many cases, virtual agents act as tier one support and live agents as tier two. Ensuring that your virtual agent can identify a caller based on their phone number will require access to your CRM. Confirming that it can change a shipping address will require access to your ordering system. Both will require help from IT. If the business side wants something done, but IT doesn’t have the time or the capability to help pull it off, your effort will suffer. Bring IT in early as part of the process, and make sure it fully understands the business case and the benefits of deploying an AI-powered virtual agent.
4. Design effective conversations – bot or not, your goal is to build the best conversation possible for each scenario and use it consistently. The base for a great conversation can be the collective wisdom of your agents. They can help design a winning script and highlight gaps. You can also use AI to analyze and classify all previous calls and find patterns and data to train your virtual agent. In addition to just being an objectively good conversation, a bot conversation needs to mimic the elements of a great person-to-person call. It must be speedy (not more than a second between turns); understand and derive meaning from complex sentences; be contextual based on your history and current intent; and be designed to achieve an end-goal. As a cherry on the top, bots that use expressive voices keep callers engaged and focused.
5. Measure success and adjust – just like human agents, bots should get better at what they do over time. The benefit of a virtual agent is that you only have to teach one (virtual) agent, one time, and this new skill will be available immediately for all callers to enjoy. If you set a business goal (e.g., CSAT or order completion rate), you can use it as a benchmark to assess the bot improvement. Make sure that your bot comes equipped with an out-of-the-box continuous learning platform, and that your users have an easy-to-access UI to review and analyze calls. Without continuous learning, your bot will quickly become stale as you callers use new product names and terms.

 

At its core, excellent customer service is about having useful and effective conversations with customers so that companies can meet their needs, whether spoken or not. A reliable and productive conversation leads to higher customer satisfaction, quicker response times, increased loyalty to brands, and reduced effort from the customer. Conversational AI will increasingly play a critical role in making that happen as more companies smartly deploy the technology that never quits.