The Reasons for Chatbot Failures

phone chatbot failiure

Chatbots are lead generators, help desks, customer services, reinforcements for brand recognition and so on… but what if customers intentionally avoid talking to your bot? What are the reasons for chatbot failures? What are the biggest impediments for them to interact with your precisely designed “colleague”? How to overcome the bias against chatbots that results from bad experience?

In hopes of answering these questions, we conducted a survey at Botium to see exactly what “chatbot-refusers” think and what is their first choice of information source. We tried to assess their general experience and how they feel towards different types of chatbots. I discuss the results of the survey and provide insight into people’s likes and dislikes when interacting with chatbots. These results will hopefully help you optimize your business’ customer experience with conversational AI!

We’ve passed the phase where chatbots are an “emerging” technology. They have been around for a long time and in that time they’ve rarely met customer expectations or provided a positive experience. The real game changer is not to have a chatbot anymore, but to empower it with real potential.

Without proper development they may confuse, rather than help users and will negatively affect the general impression of your company and the service you are providing. Given how often we’re looking for instant results, it’s easy to understand why some businesses deploy a half-baked chatbot and end up training it on their customers, resulting in poor quality. This mindset feeds the number of people disillusioned with conversational AI.

Let’s see their point of view

According to our study 28% of people stated that they do not interact with a chatbot in any circumstances. Refusing new technological solutions could be caused by generational differences, but in our case the “refusal group” is enjoying their twenties and based on the answers given to our open ended question, their negative attitude clearly results from bad experience.

The TOP 5 worst chatbot experiences

Dead end road

The fancy “Let’s direct everyone to our conversational AI” solution coupled with an awfully functioning chatbot is considered the most annoying customer service ever. The bot fails and human takeover is also not an option.

Harassing chatbots

The chatbot that never stops popping up. To make things worse it shows up accompanied by an annoying sound when you just click anywhere on the website. It gives you the feeling of a fly landing on you that you just simply can’t shoo away. In this case, customers just leave the page without finding what they were looking for because the perceived irritation is greater than the expected benefit.

Replay

The third most frequently mentioned problem was a failed chatbot, followed by a human takeover where the user had to repeat their whole issue again.

Limited number of options

Button-based chatbots are cool, but they radically limit the directions of where the conversation could go. Most customers find it frustrating that the conversation is highly restricted and there’s no option to find the topic they are looking for or they end up in an endless loop.

Low understanding

Surprisingly, the often mentioned problem is that chatbots don’t understand the complexity of the human language, ended up in fifth place.

In the above listed cases — even though chatbots performed below expectations — , we were not opposed to an instinctive rejection and despite their disappointment, people still tried to achieve their goal by talking to a chatbot. In this article, we want to focus on a different segment, the so called “chatbot-refusers’, meaning people who don’t even try to interact with chatbots by default. So let’s jump into the main point of finding out the reasons behind:

Most common reasons for failed conversations with Chatbots

Once we take a closer look at why users aren’t happy with chatbots, we begin to see that the problem truly can be solvable. The biggest reason that customers often feel that chatbots will fail to deliver is simply the complexity of the task at hand. The problem is that once we move away from simple command-response interactions we end up in the complex terrain of the human language and (as we saw above) in case we choose the button-based structure, the conversation will be highly restricted.

Let’s remind ourselves that at the end of the day, the customer is the one who determines if the experience was good or not. It doesn’t matter if we are pushing these tools as being the next best thing in providing quick help, if they are malfunctioning, or even worse, not used. At the end, the customer experience will tell us if the solution is even worth investing in. Let’s not forget that according to our survey only 28% of people avoid chatbots.

If we further examine the discussed group, we can get an insight on what other sources of information they prefer. Not surprisingly, search engines are still the most used option, when it comes to finding quick answers on the internet.

Information sources used by ‘chatbot-refusers’

They are followed by the website of the company about which users are trying to find out more. Among “chatbot-refusers” the FAQ is still a popular channel, however most of the time they are just content mapping of chatbots (vice versa).

What’s the perceived benefit then?

In order to acquire specific information from a chatbot, we have to reach a certain conversation depth, which can be time consuming. On the contrary, search engines and FAQ pages offer you the option to easily search for keywords and to directly jump to your interest, speeding up the process.

How to overcome the danger of chatbot failure?

Start investing in chatbot development! Our experience shows that companies want to keep up with the latest trends with the lowest costs possible. There is a famous saying: “who buys cheap, buys twice.” This is explicitly true for chatbot development. The money companies save on quality assurance will repeatedly backfire on them, often in a form that can no longer be price tagged, such as frustrated customers, market share loss, brand value and sales declination etc.

To make things worse, this negative attitude was created by the whole conversational AI industry, which is extremely hard to overcome. So even if you have a super smart chatbot, users will still be affected by their bias and could avoid the interaction. Therefore, it would be a common interest for companies to launch properly designed, trained and tested chatbots.

Summary

Unfortunately, a large number of users seem to not trust chatbots and feel that chatbots don’t fully understand their issue. As mentioned, chatbots have been around for some time and the serving infrastructure has been also built around them. There are chatbot testing tools developed by highly trained, experienced professionals like Botium that helps to polish your chatbot. This quality assurance does not only help your own business, but to improve the general negative experience accumulated over several years, created by poorly designed bots. As the industry offers now so many opportunities not only to develop, but also to improve chatbots, there is no longer an acceptable excuse for a malfunctioning bot!


One response to “The Reasons for Chatbot Failures”

  1. […] Chatbots fail. It’s a known issue, but there are not enough actions to avoid it. If we take a look at the global market, only 3% of the big companies that own a chatbot (and could easily afford QA) have automated testing in place for their conversational AI. The other 97% are wasting enormous manpower on manual testing, or don’t even pay attention to quality assurance. However, with this attitude, the primary purpose of chatbot development becomes meaningless. The chatbot can not complete the task in hand and returns the request to the human agent as it would have been without the chatbot anyway. The only difference between the two scenarios is the millions of dollars spent on chatbot development. […]

Leave a Reply

Your email address will not be published. Required fields are marked *