What to Consider When Deploying Chatbots

By Mike Murphy, VP of Customer Success, Creative Virtual USA

Launching a new digital customer service and sales solution can be equal parts exciting and intimidating, especially if your company is new to alternative digital care and sales solutions. But deploying chatbots to lower customer effort and improve the speed at which they can get the help they need isn’t as difficult as you might think.

Here are a few things to keep in mind to ensure an easy, successful chatbot launch:

  1. Build from Customer Conversations, not FAQs

Your chatbot is designed to have conversations with your customers – not to dispense one-sided lectures. FAQs and knowledgebase articles are written to broadcast a detailed message to a large audience. Good articles address the full scope of an issue. Chatbot conversations should look more like your text messages than a Wikipedia article. Active dialogue between customer and chatbot results in a customer-specific journey, which lowers effort for the customer and improves the accuracy of solutions. While it’s fine to supplement chatbot responses with content from product pages and support databases, the majority of your chatbot’s responses should be based on real conversations your representatives have with actual customers. Start with live chat transcripts, transcribed voice calls, and email correspondence.

  1. Perfect is the Enemy of Good

Answer content typically comes from one of three sources:  Voice of the Business (knowledgebase articles, corporate policy), Voice of the Agent (chat transcripts, call recordings), or Voice of the Customer (customer feedback, exit surveys). There is a tendency to load up your new chatbot with lots of VOB before launch. This is a mistake. Get your top issues from VOA analysis, supplement with the minimum amount of VOB you can get away with, and release the tool to interact with customers. Let the “learning” part of “Machine Learning” be driven by VOC data that starts flowing in with the first conversation. The more conversations your virtual assistant has, the easier it will be to make quick changes that improve accuracy, reduce handle time for the customer, and improve customer satisfaction.

  1. Be agile

Your call center supervisors coach new agents daily as they first start taking calls to ensure quality and efficiency meets your standards. Follow the same approach with your chatbot. Once you launch, pay attention to what questions or inquiries it receives the most, and which conversations are receiving poor feedback. Have a subject-matter expert on your business work with an expert in digital customer service to coordinate on delivering quick updates and improvements – adding or tweaking answers for common questions and approving the machine-suggested recommendations.

When can you expect ROI after deploying chatbots?

Payback periods are heavily dependent on the pricing model you have chosen for the program. Traditional software costs, maintenance, and professional services fees may take months or years to recoup depending on the scope of the deployment and rate of customer adoption. However, Pay-For-Performance models tied to desired business outcomes (resolved contacts, converted sale, qualified lead, etc.) mean no up-front costs and returns begin with the very first chatbot conversation.

Be Aggressive

Once your solution is live and actively helping your company shift conversations to intelligent self-service, there is little risk with quickly opening access to the tool across more areas of your site and mobile apps. More conversations flowing through the chatbot allow for more data and Voice of the Customer intelligence to improve the solution. Plan to make your chatbot Easy to Access, Intelligent, and Conversational and watch as your customers adopt your new cost-effective channel as their preferred means of contact due to its ease and effectiveness.