Chatbot. No, it’s not a 21st century nickname for your friend who won’t stop talking. A chatbot, short for ‘chatterbot’, is the latest revived craze among tech experts and businesses looking to use cutting edge technology to get a leg up on the competition.

In basic terms, a chatbot is like a computer program that you can teach to respond to defined questions or commands. This whole question and answer approach lends itself well to customer service, and as Gartner’s 2016 Hype Cycle for CRM, Customer Service, and Customer Engagement highlights, chatbots are on the rise when it comes to customer service.

Companies that have started to experiment with customer service chatbots are excited about its potential, but while businesses are drooling over yet another automated point of contact to reach customers, the jury’s still out on how customers feel about this new way to talk to companies.

Says Jonny Everett, co-founder of live chat software The Chat Shop, “I think the key question businesses have to ask themselves is whether the use case they’re looking at applying chatbots to is better for them, or better for the customer. Companies are spending more time now backpedalling the ‘never speak to us’ customer service approach, and I think chatbots run the risk of deluding businesses into thinking that it’s a great solution for their customer, when they’re really looking at it as a great solution for them.”

Let’s take a look at what chatbots are, how they’re already being used, and what customers think about customer service in its current state, to see if chatbots, poised to be the next darlings of business to consumer communication, can satisfy a customer’s needs in the make-or-break world of customer service experience.

How do chatbots work?

For the non-technically minded, the term “chat” can be a little misleading when it comes to chatbots. You may conjure up the idea of live chat, which is currently a more common and familiar use of chat for customer service – it’s the messaging bubble that pops up on websites asking you if you need help. Most of these chats start with an automated message, but are usually picked up by a live agent to continue the conversation.

Here's an example of an automated message starting a live chat conversation

Here’s an example of an automated message starting a live chat conversation.

You might also be thinking Siri or Cortana, the virtual assistants from Apple and Microsoft that use voice recognition to carry out requests on your mobile device, sometimes with pretty unreliable results.

The chatbots we’re talking about here are a bit different.

More robotic than humanistic, chatbots can work in two different ways:

  1. Simple, canned responses based on a specific command;
  2. Artificially intelligent bots that can answer contextual questions by learning from previous interactions, carrying on a pseudo-conversation.

1. Rule-based bots

These bots are built with predetermined answers to pre-defined questions. Think of it as more of a logic tree, where the answers are hard coded based on the questions that you ask it.

Although the TechCrunch Bot on Telegram uses AI, it's limitations mean it can't answer much more than questions about the news.

Although the TechCrunch Bot on Telegram uses AI, its limitations mean it can’t answer much more than questions about the news.

The problem with rule-based bots is that they’re only as smart as you make them. They can only answer questions that they’ve been pre-programmed to. Anything else will come out with the equivalent of a “does not compute” answer.

2. Machine learning bots

Remember that Spike Jonze movie Her, when Joaquin Phoenix’s character fell in love with a machine-learning robot voiced by Scarlett Johansson? Although it’s an extreme case (and AI is nowhere near that good right now), machine learning bots work in the same way. Using artificial intelligence, they learn based on the way you interact with them. Here’s an example of a conversation I had with Poncho, Facebook Messenger’s machine-learning weather bot.

Poncho on Messenger

As you can see, Facebook tried to inflect personality into Poncho by bringing up his (its?) ex in conversation. Cute, but the conversation basically stopped at that. I’m sure if I had actually unloaded my woes on Poncho, it’d respond by giving me more weather forecasts. Poncho also had trouble understanding that I was in Barcelona– it’s not exactly nowheresville– and I had to clarify that it was in Spain.

The biggest problem with machine learning bots is simply that they’re not learning quick enough– the technology is far from advanced enough to get to Samantha levels of understanding. Microsoft’s Tay disaster shows that even if bots can learn, they might pick up some nasty habits along the way.

If you want to know more about chatbots, this is a great intro into the subject, including details on how you can build your own bot.

What are bots currently being used for?

Despite some of its technical failures, businesses have already hopped on board and started using bots for marketing, eCommerce, and yes, customer service.


Major brands have started using bots as part of their marketing strategy, especially in conjunction with consumer messaging apps. Facebook’s Poncho was more of an experiment than anything else, but other messaging apps have developed full-on bot shops, where brands have released interactive bots that actually work.

H&M has developed a bot in Kik Messenger’s Bot Shop that will give customers outfit suggestions based on a certain piece of clothing they like, while Sephora has developed a bot that gives makeup tips. Notably, the Weather Network also has a weather bot in Kik’s Bot Shop that proves much more useful than Poncho.

Kik's Bot Shop


Chatbots that use messaging apps are also a great way to experiment with bot-commerce. While Facebook may not have been successful with Poncho, it’s definitely capitalizing (no pun intended) when it comes to eCommerce bots. Partnering with companies like Shop Spring, shoppers can tell bots what they’re looking for and make a purchase straight from the app. Other messaging apps like WeChat, Line, and Kik have similar options for online shopping.

Shop Spring on Messenger


Customer service bots

Some companies are starting to use customer service chatbots to answer customer queries. A digital bank in India uses bots to open bank accounts for new customers, having trained it on millions of customer questions in order to be able to come up with the relevant and appropriate answers. The company behind the technology, Kasisto, has been able to create a bot that effectively answer questions and engages in “conversations” with its customers.

India Bank Bots


There are a ton of other ways that businesses are using chatbots. This is a great list, broken down by industry.

Domo arigato Mr. Roboto?

Cases like WeChat in China and LINE in Japan show that bots can work well when it comes to eCommerce and marketing, especially when it’s a simple black and white request. But they’ve yet to prove their worth for customer service. The Digibank example in India works well because the bot’s been programmed to answer thousands of different questions, but the second a conversation diverges from what the bot knows, things can get messy.

From what we know about customer service already, however, bots might be the answer to many gripes about customer service experiences. Let’s see how bots can stand up to some of these customer service stats…

FACT: Customers hate talking to people.

One third of customers say they’d ‘rather clean a toilet’ than speak with customer service.

How chatbots will help:

Despite the overall aim to make them more human-like, bots aren’t people, and chatbots mean that the dreaded phone conversation with a grumpy customer service rep won’t be necessary (unless it’s programmed to be grumpy).

FACT: Customer wants more self-service options.

The same study as above shows that 73 percent of consumers want the ability to solve product/service issues on their own.

How chatbots will help:

Traditional self-service includes things like a help section where customers can dig around to solve their own problems. Similarly, bots don’t involve relying on another human being to solve a problem, being more of a database query than a human interaction.

FACT: Customers want mobile support.

A study from Software Advice shows that 63 percent of customers use mobile devices on a weekly or monthly basis for customer support, although 90 percent of customers have poor experiences getting proper support on mobile.

How chatbots will help:

Chatbots lend themselves perfectly to mobile, especially when used on messaging apps like Messenger, Kik, and WeChat.

FACT: Customer don’t want to wait for an answer.

One study shows that 60 percent of customers will hang up after one minute of being on hold.

How chatbots will help:

With the power to scour databases at breakneck speeds, bots don’t require an extended waiting time to get a response (think seconds, not minutes). The problem is whether or not they can give you the right answer.

FACT: Customers want better service from people.

For those that must succumb to human intervention, 40 percent say they want better human service.

How chatbots will help:

They might not. If programmed correctly, a bot can be permanently friendly and answer the exact questions that you’re asking for. As they are right now though, people are likelier to get frustrated with an uninformed or underdeveloped chatbot than they are to find the right answer.

The verdict: They’ll help, eventually

Gartner’s right– if AI technology advances at a quick enough pace, customer service chatbots will hit a lot of the pain points that customers have when it comes to reaching out for customer service. The key takeaway here is, they’re nowhere near that level yet. Customer service chatbots are still largely in the developmental phase with lots of kinks to work out to ensure that they come up with the right answers.

Sabine Sipunova, founder of Sorry as a Service, sums it up nicely by saying that “chatbots could potentially be a part of customer service right now in companies where the offering is pretty standard and customers have similar issues. Anything more complicated, and chatbots would add an additional unnecessary friction, since they are still in their early stages. AI technologies are not as developed within customer service to ensure that this process is seamless.”

Until then, maybe the term chatbot is better reserved for your friend that says crazy things.

Create your own customer service chatbot

If you’re interested in creating your own chatbot for business, check out this great article in the Chatbot Magazine, which gives you info about how to create chatbots, and some tools you can use to start making your own.

If you want to learn more about how messenger apps can play a role in customer service, with or without chatbots, check out this article, ‘Why consumer messaging apps for business are the next big thing’.


Share This

Share this post with your friends!