Where do chatbots fail?

In previous articles we have commented What is a chatbot, what are its advantages and disadvantages and How do they work. Today we want to focus and show you why it's difficult for chatbots to hold human-like conversations and why it's difficult to deal with natural language.

It goes without saying that chatbots are one of the most complex applications of Artificial Intelligence (AI) since the solutions they offer are mostly different from each other.

What do you expect from a Chatbot?

Normally, before purchasing a chatbot, a company expects 3 things: a simple interface, the ability to work at 99% and the response time of any task that they are expected to complete.

Not all chatbots meet the expectations generated, some have come to annoy users. Next, we will analyze the Why do some chatbots fail to provide a good perfect user experience, pleasant and efficient.

Why don't some chatbots meet your expectations?

Artificial Intelligence isn't that smart

La Gran Most chatbots that we can see they're not really intelligent. They are created with a logical tree decision, that is, where the answer given by the bot depends on the specific keyword indicated by the user's decision. These are the so-called RULED-BASED bots.

These types of bots will be as intelligent as the ability of the person responsible for their creation to anticipate all potential user cases.

Little knowledge about chatbots

When new technologies appear on the market, all companies want to implement or offer them as the first, but we have to be very careful because, sometimes, being the first can hurt us. First of all, we need to familiarize ourselves with the new tools and when they are fully mastered we can offer a good product and service.

Over the past year we've been able to see a plethora of bots that are solving irrelevant situations or that offer really poor experiences. This is why it is essential to master a tool or program before offering services in which, if not properly implemented, it can harm a company.

When did the first appear chatbots on Facebook, the programs did not meet 70% of user requests because the technology was not sufficiently developed, these requests could not have been managed without the help of people.

In 2016, Facebook had around 30,000 active chatbots, but since they didn't work properly, they frustrated users. This fact prompted Facebook to review these Chatbots.

Lack of transparency

The bots that are successful they are transparent from the beginning of the experience, they report that you are talking with a bot and not with a person. If we are honest from the start, it will make users more understanding of the mistakes that may be made.

Problem in interpreting the context

Talking to a person is not the same as talking to a bot, people understand irony, sarcasm, we can read between the lines... On the other hand, robots cannot do the same.

Another example of the limitations of chatbots with Artificial Intelligence (AI) and the interpretation of the context was when, also in 2016, Microsoft launched Tay, its AI conversation bot for the Twitter platform. This Chatbot created controversy for delivering offensive, racist and xenophobic messages and was discontinued 16 hours after its launch.

Bots can't do “everything”

It is believed that chatbots can answer all kinds of questions and that is not the case. If we want a chatbot to answer all kinds of queries and solve everything a user wants, we will compromise its quality. The best examples are Amazon Echo and Google Home smart speakers, we want your virtual assistants to respond to anything, but if you've had the opportunity to try them, you'll have been able to verify that it's not possible.

The user prefers people as attendees

When the bot fails, the user would prefer a human to take over in order to resolve all their issues. Although we have been able to verify how companies do not have a defined protocol when chatbots stop working or when they can't help the user. This is a very important point to keep in mind because users may feel more frustrated than when they started the conversation with the brand.

How can we fix it?

Although creating a chatbot is a big challenge, there are also tricks and techniques for creating successful chatbots and create a good user experience at the same time. Below we detail some methods that we can use to feed our bot and ensure that it improves over time.

Seduce users

The ultimate goal should be to offer a good user experience and to put a smile on the face of our consumer. To achieve this, the chatbot should do the following:

• Maintain a simplified language and context. There's no need to complicate things unnecessarily.
• The bot must be aligned with the value of our brand.
• Focus on developing the bot's personality.
• Have an alternative answer for when the bot goes blank or doesn't understand the user.

Train the chatbot constantly

If we want our bot to be a success, a constant training process is required. Bots must be efficiently programmed with the ability to learn from the conversations they have with their users so that they can improve over time.

At this point, the Machine Learning (Machine Learning - ML) which is the field of artificial intelligence that focuses on making machines learn and, the Natural Language Processing (NLP) which is a field of computer science, artificial intelligence and linguistics that studies the interactions between computers and human language. In the end, it's all about training and continuing to develop the bot day by day to make it better.

Use the people resource

Sometimes, no matter how intelligent the bot is, there will be aspects that it will not know how to answer and in this case there should be a person to try to solve this issue. For example, if the bot receives a question that it does not know how to answer, one of the answers could be: offering the user a direct customer service telephone number so that they can serve the user in a personalized way and the user feels that they have had a good experience.

Have a strategy

It's essential Set some objectives to the bot for the different actions that a user can perform. The chatbot must offer value and achieve a goal, regardless of the conversation carried out.

The bot can encourage users to have interesting conversations, if we have a well-trained bot and a well-defined strategy behind it, bots can become real drivers for the company. If we offer a good user experience, they can become true brand prescribers.

In the end, a chatbot needs to be simple and easy to understand. For a chatbot not to fail and the user not to feel frustrated, like everything in life, it requires a learning and training constant.

How can we help you?

If you need more information, do not hesitate to contact us.