Avoid Creating Virtual Idiots for Customer Disservice

Virtual assistants are a white-hot topic nowadays. The vendor marketing drumbeat is loud, and they promise nothing short of eradicating world hunger with their chatbots! However, if you look beyond the hype, success stories are few and far between.

In fact. 54 percent of online U.S. consumers think that interactions with customer service chatbots will negatively impact the quality of their lives, per Forrester Research's latest data, which predicts a chatbot backlash this year.

I've heard of one customer calling the virtual assistant of a business he was contacting a "virtual idiot." How do you prevent your virtual assistant from suffering such ignominy? Here are 11 mistakes to avoid:

  1. Boiling the ocean: While a chatbot can become smarter over time, trying to do too much with it at the outset often leads to failure. It is best to limit its scope to a narrow set of use-cases and intents to get a quick win and then gain momentum.
  2. The onus boomerang: When customers have specific questions, many chatbots just push back web pages or FAQs or documents instead of answering the questions. It is like giving the entire haystack rather than finding and handing over the one needle that the customer truly wants.
  3. Not understanding intent: It is important to first understand the intent of the customer for fast time to answer. Except for the lonely soul or two out there, consumers are not looking to socialize with chatbots; they want quick answers. A good practice is to use human chat conversations, label the intent, and use machine learning to match the customer utterances to intents.
  4. Stopping with intent: Understanding intent is a good first step. Next, the chatbot needs to be able to converse and guide the customer to an answer, especially for the more complex queries. Look for a chatbot that is backed by artificial intelligence to provide such guided, conversational assistance. One of our major telco clients uses our agent-facing bot, backed by reasoning and knowledge, to guide 10,000 agents in the contact center and associates in 600 retail stores to answers. The company has since seen a 37 percent improvement in first-contact resolution, 30-point improvement in Net Promoter Score, and 50 percent improvement in agent speed to competency.
  5. The bot relay: Some businesses are looking at creating concierge bots and a set of specialist bots, where the concierge passes the baton to specialist bots if it is unable to answer a question. Bot-switching can be as painful as channel switching and can lead to poor customer experiences, especially if the specialist bots cannot resolve customer queries. The answer is to implement a smart chatbot powered by a robust knowledge base and reasoning that can escalate to human chat agents with full context. Another approach is to make the bot switch invisible to the customer. In either case, the customer experience should be at the front and center of the approach.
  6. The disconnected chatbot: Chatbots should be able to escalate to human agents, based on customer sentiment, customer value, customer situation, its own inability to resolve the issue, or a combination of these factors. And, it should do so with all the context intact so that the conversation with the human agent moves the conversation forward instead of starting over. Some queries might need long-lived, multi-step resolutions. In such cases, the chatbot should be able to pause a conversation and pick it up where it left off without asking the customer to repeat information or steps that had already been completed. You need a unified, omnichannel customer engagement system, backed by knowledge management and AI, to ensure these capabilities. One of our premier clients uses our chatbot to answer DIY tax-payer questions. Where needed, the customer can escalate to a human advisor with all of the context intact. The advisor then chats and cobrowses with the consumer to answer questions and help fill out forms in real time, a novel experience for the consumer and a win-win for both the consumer and the tax preparation giant.
  7. The stagnant bot: While it is OK to start small, companies should make their chatbots smarter over time in both breadth and depth of knowledge. Moreover, the knowledge base and AI reasoning paths need to be updated and optimized with analytics on an ongoing basis.
  8. The hidden bot: Chatbots can go beyond reactive customer service to proactive engagement. It is important that you make it visible on your website. Feature the chatbot at least on the top 10 most-visited web pages, in addition to the support section.
  9. The curse of the cursor: When consumers look to get customer support, the last thing they want to do is stare at the wait cursor, whether it is the wheel or the hour glass. If your chatbot is not fast, customers will defect. Make sure it can scale.
  10. Misalignment with your brand: Your chatbot needs to be aligned with the brand in both style and substance. If it is fronted by an avatar, make sure it is aligned with the personality of your brand and the target customer. The bot's conversational tone should also mirror your brand style. If it is a high-touch brand, you might want to escalate more quickly to a human if the chatbot is unable to answer the customer's question.
  11. Not being able to try before buying: With many vendors promising omniscience from their chatbots, you are faced with the unenviable task of picking just one. Technology capabilities are important, and so is best-practice domain expertise. How long has the vendor been in the space? Does it offer risk-free pilots with best-practice guidance, all free of charge? Get answers to these questions.

By avoiding these mistakes, you can make your virtual assistant a virtuoso assistant!


Anand Subramaniam is senior vice president of marketing at eGain.