Chatbots — A fancy communication tool? Or an integral member of your customer service team?

Original article was published on Artificial Intelligence on Medium

Chatbots — A fancy communication tool? Or an integral member of your customer service team?

Why chatbots fail — and how to leverage them even on a modest budget.


2017 was the year of the chatbot. So was 2018.

But fast forward to 2020, many chatbots have emerged — but only a few remain relevant today. The Next Web proclaimed the death of chatbots. Others are questioning the wisdom of investing heavily in chatbots, when their purposes seldom fulfill customer satisfaction.

The allure of chatbots is obvious. Cheap, instantaneous, around-the-clock customer service. Automated replies to frequently asked questions, freeing up their human counterparts to tackle the complicated issues. While still retaining the ‘human touch’ of speaking to a customer service officer. What’s not to like?

Chatbots — like any tool or software ­– are only as effective as the ones who wield them.

One common pitfall is that chatbots are dependent on language recognition patterns — which are highly dependent on the following technologies to make their conversations more human-like:

  • machine learning (ML),
  • deep learning (DL), and
  • Natural Language Processing (NLP)

Given that the above technologies are expensive, and combined with the fact that most companies are in a rush to launch their next chatbot, it becomes obvious why most chatbots swiftly become obsolete. These chatbots don’t respond well to customers’ input, show the clunky workings of a badly designed user experience, and worse still, drive customers up the wall with their stone-faced replies.

A chatbot’s effectiveness depends on how well the company understands their customers’ pain points.

And in turn, how focused the chatbot’s functions should be. If not done well, the chatbot stands a high risk of being implemented poorly — and becoming irrelevant.

Rather than having an ambitious plan of having a chatbot that can replace a human employee, the key is to make the chatbot a specialised tool for providing quick solutions to simple problems commonly encountered by customers.

By constraining the roles of the chatbots, it’s also easier to pare down on the third-party services and APIs the chatbot needs, which cuts down on unnecessary costs for maintaining the chatbot.

In the meantime, as advances to conversational AI continues slow (but steady), chatbots can still contribute to businesses in the following ways:

  • 24-hour services for simple and easy-to-automate queries, eg. checking order status
  • 24-hour service for simple and easy-to-automate functions, eg. updating scheduled appointments
  • Answers to Frequently Asked Questions (FAQs)

Even as the chatbots’ functions are narrowed to a specialised range, they can still provide value by helping to streamline customer support and improve user retention, especially when a customer service officer is not instantly available.

Internally, conversational AI can also be used to co-automate internal business tasks to allow staff to be more productive such as scheduling meetings, assisting sales processes and handling legal documentation.

And before you forget, take a closer look at what your chatbot is saying.

Even as chatbots seem like purely technical tools, your chatbot’s main medium of communication is still words.

(This applies even if you are using ML, DL and NLP to help your chatbot respond as humanly as possible.)

Does it have personality? Is it chatty — but not too chatty? Or does it sound like an instruction manual broken up into speech bubbles? While designing a chatbot’s conversational flow, it is important to consider the context. A bubbly, slangy tone might be great for a lifestyle brand, but maybe not so much for instance, a FinTech startup (though of course, it also depends on your brand image).

On a practical level, it also makes better sense to create structures — such as prompts and buttons — to direct the conversation so that users can ask the right questions, and the chatbot can answer them. This would also minimise instances where the AI encounters difficulties in understanding nuances in human language — and create hiccups in the user experience.

Despite claims of the chatbot’s demise, chatbots still have an integral role to play.

Both as a means to automate conversion of leads and prospects, and to help answer common queries with a human touch. This is especially so in a pandemic, where call centres are emptying for lockdown or social distancing, and people need more answers than ever before.

In future, with further advances in ML, chatbots may one day help us conduct market research and predict future trends. Or tell us the answers even before we ask them. But for now, chatbots will do what they can.