top of page

The hidden, unsettling numbers behind chatbots

Updated: Jan 23, 2020

One of the latest big tech promises is the chatbot revolution. From customer service to sales, we’ve heard so much about the supposedly transformative power of this technology, it feels like an army of chatbots was programmed to spread this message. So much so that $4.5 billion are expected to be invested in chatbots this year. But what is the real success (or damage) rate of chatbots? After several years of interacting with chatbots, it’s time we ask a few difficult questions about where this technology currently stands.

I believe in asking KPI-related questions and taking market predictions with a grain of salt. In fact, one of my favorite questions to ask colleagues at conferences is “what percentage of your customer support communication is successfully handled by chatbots?”. This question has led me to a couple of conclusions; first, that I must be fun at parties and second, that even the world’s biggest brands enjoy a chatbot success rate of between 4-10%. These off-the-record statements come from executives at global brands in a variety of industries such as retail, travel, media, and more. If you’re curious to settle the dissonance between grandiose market predictions and practical results, here are a few interesting thoughts to consider.

The Helpful 5%

Chatbots are not useless. They are capable of sending automated messages, informing users that help is on the way, and in some cases provide initial assistance or even resolve standard issues. Problems usually occur when one of the following issues arises:

  • Chatbots are only able to address the specific question they are presented using a set of canned responses. In most cases, data coming from other sources, such as the company’s CRM, doesn’t come into play.

  • We expect bots to replace human support agents in more complex interactions, particularly those that require action – such as a refund or a coupon – instead of an answer.

  • Since B2C companies face a great deal of questions regarding the sensitive topic of billing, the majority of interactions cannot be automated.

  • There are multiple possible replies to a certain question, depending on the person’s position, pricing package and specific needs. Most chatbots simply aren’t sophisticated enough to understand which one fits most in each case.

That’s why 65% of businesses are worried about chatbots being unable to understand the issue at hand. Almost four years later, the right technology is still not in place and chatbot trolling in an international sport. Considering the fact that every communication with customers is an opportunity and that every bad conversation ends up on social media — this is a major problem.

Augment Agents, Don't Replace Them

We’ve mentioned before that chatbots seem to cover less than 10% of the customer service efforts, but what about the rest of the work? Well, in 2020 it is done manually by human representatives, the old fashion way. Based on the annual investment in support centers, agents can breathe a sigh of relief, knowing that bots aren’t taking over their jobs in the near future. Many customers focus on using terms that are designed to have their conversation transferred to a human agent and 75% of customers complain that it takes too long to make that transition, with the recommended chat time before this switch is made set at no longer than 2 minutes.

The most important lesson of all is that to succeed, we simply cannot separate agents from AI-driven solutions, because only the combination of technology and humans will triumph today’s customer service challenges. Assisting the human customer-facing employees will touch 95% of the problem and stop the unneeded divide between AI tools and customer support representatives, we must harness this powerful technology. Let’s stop worrying about the risk of bots replacing employees and start focusing on supporting representatives with super tools that have the power to transform the customer service industry.

224 views0 comments


bottom of page