Chatbot Challenges

Chatbots on mobile

By ian tomlin | Jun 09, 2021

Chatbots on mobile Cover Image

In this article, we discuss how a chatbot strategy can be deployed on smart devices.

Did you know that the latest data suggests 1.4 billion people use messaging apps, there are today over 30,000 Facebook bots and that 63 percent of people would consider messaging an online chatbot to communicate with a business or brand to find emergency answers?

Customer service experience has gone mobile. That means your brand credentials, its ability to tap into new business opportunities and its chances to impress customers and new prospects – all rely on the quality of response your bot gives to enquirers. By 2019, more than 25 percent of the world’s population (roughly 1.75 billion people) will be using mobile messaging apps and, according to [[24]7](, online chat and messaging apps are the preferred way for 29 percent of people to contact retailers when making a purchase decision. That means people are equally likely to contact a retailer by phone or use a chatbot and more likely to use a chatbot than to contact a retailer via email (27 percent).

Is your mobile bot up to the job?

Most bots start out as pretty dumb machines. It’s not their fault. As with human customer service agents, you will need to educate them on your products and services, the common requirements of customers and prospects, and the typical sorts of questions they should expect. Like humans, bots need to receive training to improve. Artificial Intelligence has transformed the bot training industry, but–like Sherlock Holmes–your AI system requires ‘data, data, data.’

In this article, Diana Lee explains how she and her team studied 200,000 chat conversations to determine the best way to train their mobile bots–but the team had to do it manually. If you want to avoid having to throw that level of man and woman power at the problem of optimising your bots, is there another way? The answer is yes.

How mobile bot training works

Modern AI-driven alternative training solutions combine a knowledge of linguistics (this work has been about Natural Language Processing backed up by linguistic research methodologies) with the ability of machines to progressively learn by themselves and recommend better outcomes. Systems like INTNT.ENGINE are Software-as-a-Service (SaaS) platforms that ‘plug-in’ to your technology ecosystem to continuously train your bots. Using patent-pending NLP research understanding, INTNT.ENGINE is able to understand aspects of conversation such as intent, anger, humour and annoyance. Blending these characteristics of human conversation with the words spoken helps INTNT.ENGINE to more accurately predict the response the enquirer is looking for.

These systems look to resolve two key weaknesses in your mobile bots, namely:

  1. False positives – when the bot thinks it understands the question being proffered but gives the wrong answer.
  2. Missed intents – when your bot doesn’t get what the requestor is talking about.

There are various suppliers that claim to automate the resolution of false positives, but far fewer are able to address the second category of missed intents with any degree of reliability. This is because the means to interpret missed intents requires a fundamental understanding of ‘sentiment’ and to do that requires a deep appreciation of Natural Language Processing concepts. Fortunately, the clever people at A*Star–the Agency for Science, Technology and Research in Singapore–have come up with a new algorithm that means your bots can be taught the finer details of how language works. This algorithm is at the heart of INTNT.ENGINE, the SaaS platform created by INTNT.AI.

Find out how to train your mobile chat or voice bot by requesting the INTNT.ENGINE demo video, or book an informal virtual coffee chat with one of our customer success team.