Ever since chatbots were conceived, much of the business world has been waiting for a chatbot and AI revolution. Due to important advancements in machine learning and AI technology, it has only been in the past decade or so that e-commerce industries have embraced chatbots wholeheartedly.

Chatbots in e-commerce companies are designed to help customers with online purchases by providing enhanced quality of assistance in the form of relevant product recommendations, available offers, etc. They also give business organisations new opportunities to improve their customer engagement process and operational efficiency by reducing the imposing costs of customer service. In recent times, the time-saving and efficient nature of chatbots have made them so popular that Gartner predicted that by 2020 more than 85% of customer interactions will be managed by bots.

However, the main reason behind the purportedly seamless performance of chatbots is human intervention which plays a crucial role in configuring, training and optimising the system without which bots risk failure. Also, the concept of cost optimisation by reducing headcount and increasing chatbot employment has failed to elicit the desired outcome for most companies as investments in chatbots are not being able to generate the required meaningful ROI. As a result, many companies have not been able to implement them even after investing in them. According to Gartner, though 90% of enterprises have invested in chatbots, only 4% are actually using them efficiently.

There are in fact multiple such reasons which have led to chatbots falling short of the hype and expectations which had resulted in their wide-scale initial adoption. And limitations in the current training and technology has a major role to play in it.

How Chatbots are trained

when not to use chatbots

Chatbot training often requires highly qualified experts who have to use a lot of data to train these ‘intelligence’ systems. Despite the efforts put in during the initial training, even the most advanced chatbots are limited by pre-training models as they are not capable of cognitive perception because they lack a true understanding of human context.

Neither can they read between the lines nor can they understand sarcasm or jokes.

So, chatbot experts have to come up with all possible combinations of data and have to feed it into the software. But compiling an all-inclusive database for the bots is not just time-consuming and arduous, it is impossible to contrive all the plausible customer inquiries.

Also, if there is the slightest variation in phrases or syntax, the bot will not be able to understand it. For example, a single e-commerce platform catering to a plethora of industries will have hundreds of thousands of customers who will describe the same problem in many different ways.

So, chatbot training requires the tagging, segregation, and formatting of a lot of training data and even then, they cannot recognize human context in many situations. This causes chatbots to be fairly limited even in simple scenarios.  For e-commerce companies, even a 5% failure rate can lead to a torrent of bad reviews across all social media but the truth is that the maximum level of efficiency that even the most advanced chatbot can attain is only 85%. And their rate of failure and the potential for causing great harm increases manifold in more complex situations.

When a customer has been wronged

In the e-commerce industry, from the time of placing an order until its delivery, anything can go wrong. Orders often don’t get delivered at the promised date and many times, they end up getting delivered in a damaged state.

These can be highly frustrating situations for consumers and they usually contact customer care immediately to pour out their discontent. They expect the customer support agents to not only help them conclusively but also lend a sympathetic ear to their distressed rantings. A recent study by Xu et al found that 40% of users connect to customer service for emotional reasons rather than for seeking specific information.

 And it is very important for the customer support agent to be able to engage emotionally with the users at these times. But bots are incapable of showing empathy which leads to unpleasant user experience during times of distress. Without the ability to relate to customers’ emotions, bots bungle up the situation even more.

In fact, a report by PointSource states that 80% of consumers would favor human assistance when resolving problems after a purchase, as they cannot trust chatbots to render the same quality of support as humans.

This is because chatbot technology hasn’t yet progressed enough to address uncertainty and neither can they process real time information to render better services. And if this problem persists, an irritated and harassed customer may renounce that particular retailer altogether. As per Convey, 70% of consumers won’t be inclined to shop from the same retailer if they undergo a substandard customer experience. In order to prevent this from happening, businesses must be prudent enough to understand when human facilitation is necessary for remarkable customer support experience.

When customers require personalized advice

According to Statista, in 2018, the retail e-commerce sales accounted for 10% of all retail sales in the US and this is expected to reach 12.4% in 2020.

At present, a major market share of these online retail sales is captured by industries like Beauty, Furniture, Interior Designing, etc which require a highly personalized customer support system to cater to their customers precisely. In fact, A.T. Kearney E-Commerce Study shows that 60% of all regular online shoppers purchase beauty and personal care products online whereas more than 14% of furniture sold in 2017 occurred online, according to Furniture Today.

Now suppose, a consumer wants to know whether a sofa would look nice in their living room or someone is asking the customer support to suggest an ideal extension for her particular hair color. These distinctive and individualistic queries can only be resolved by human experts who should possess both a superior understanding of all the attributes of the product as well as a significant perception of customer demands.

Naturally, in such situations, chatbots prove to be a major disappointment because most often chatbots fail to deliver the required efficient and customized user experience as they lack the ability to sense and feel. That is why, many customer care experts feel that when it comes to complex or specialized matter, consumers more often than not seek human assistance.

This is because consumers are emotional beings and bots, on the other hand, can neither perceive nor reciprocate emotions. As mentioned earlier, their scope of performance is quite limited and they can work effectively only after receiving extensive training. All these reasons are impetus enough for most industries to think before deploying chatbots as their main customer support as it may not be ultimately beneficial to them.

Baby boomers and seniors prefer phone and email support

Baby boomers and seniors spend around $7 trillion per year on goods and services. So, targeting them for marketing campaigns and business plans is a lucrative idea. But, a lot of the people in this segment are not very tech savvy, nor are they willing to hand over their tasks to robots. The reason is that they consider themselves much better equipped than a bot to work confidently. In fact, research shows that 50% of boomers cannot trust a robot to complete any task as well as they can.

For this reason, when compelled to use technology and artificial intelligence, many of them become quite apprehensive. As a matter of fact, of all the consumers who buy products with the help of chatbots, only 13% are baby boomers and seniors. On the other hand, more than 90% of baby boomers and seniors consider phone support to be their topmost prioritized channel for customer service, according to a Dimensions Data benchmarking study.

So, in order to generate more revenue from this demographic group, businesses must provide them with humane personalized help at every possible step because at a time when they are unable to even navigate through the basic technicality of a normal e-commerce website, feeding them more technical jargon may drive them away permanently.

Unfortunately, chatbots do just that as they do not possess any better understanding. They cannot provide the outstanding and exclusive quality of customer support these people need to help them make purchases online. So it’s always desirable for businesses targeting boomers and seniors to include human customer support rather than only chatbots.

Chatbot trust is an important factor in e-commerce

Statistics show that in the US, the average order value (AOV) per customer is quite high. iPhone users, for example, have an AOV of $228 whereas Windows PC users have an AOV of around $187.

So, obviously, when so much money is involved, consumers are bound to feel vulnerable about the security of their transactions. And if the customer is ready to spend a good amount of money on a particular order, they might want their decision to be reinforced by customer service agents before finalizing their decision.

This is something that only a human can do and not a robot. Research shows that 73% of online shoppers believe that if they can connect to a sales representative directly for a little reassurance, they would buy things online more often.

Likewise in certain serious situations where case-specific information is essential, say health insurance or doctor’s appointment, customers may demand human help as bots may not be able to anticipate unusual instances and act accordingly. A study by Juan Ageitos states that only 5% people would trust a bot to diagnose an illness whereas 30% of consumers won’t trust chatbots at all to assist them with anything.

This shows that though bots have come very far, they still have a long way to go when it comes to gaining trust from humans. So, it’s highly advisable for businesses to be very careful in using chatbots as their sole medium of customer support as they might end up losing their customers to competitors.

You may also like when not to use chatbots.

Chatbots cannot work for enterprise sales

As per Forrester Research Inc., in 2018, B2B e-commerce in the US alone was expected to soar up to $9 trillion in total sales from all possible sources.

The market potential for B2B sales in e-commerce is so high that every industry wants a share of this pie. But, B2B Sales is a completely different ball game as compared to B2C Sales. The basic ideology of customer support is perhaps similar for both B2B and B2C industries, however, there are several major differences in supporting a business customer and a consumer. Firstly, the issues cropping up in B2B sales are much more complex and rigorous. Secondly, though companies in B2B sales may have fewer customers, their AOV is very high amounting to an average of $491.

Thirdly, in B2B, to provide exceptional customer support, personal contact and clear knowledge of the issues faced by the company are of utmost importance. In fact, 80% B2B sales organizations find personalized interactions to be the most effective way of communication with buyers.

This is because there may be multiple points of contact in B2B Sales. Research shows that 70% of B2B buying decisions include at least two decision makers.

As a result, while requesting customer support, more than one person can initiate similar sessions or one person can start a session while a different person can conclude it. To be able to deal with such situations competently, the service provider needs to be endowed with a very comprehensive and thorough knowledge of the customer.

Unfortunately, chatbots can’t ensure qualitative customer support for these kinds of businesses and can’t be good substitutes for human-to-human interaction. Chatbots can only be programmed to answer templated questions and they cannot improvise. So for industries providing B2B customer support, human-to-human interaction is still an indispensable element which can never be replaced by a bot.

Specific niche industries

Niche Industries essentially denote a distinct segment of the entire market where the customer demand is extant but unmet. Tapping this market through e-commerce can be great for business.

Today, a significant part of the e-commerce industry is comprised of niche businesses like beauty, dropship, finance, travel, etc. These enterprises not only offer a deep catalog of products with varying attributes, they also have very complex organizational policies.

And, their customer base is also very specialized and esoteric.

So, to flourish expeditiously, these businesses require adroit live customer service agents to provide personal and dedicated support to each and every customer. That is why, in such situations, the customers get exasperated when confronted with bots as bots cannot provide support at the same level.

According to DigitasLBi, only 20% people are open to receive recommendations from chatbots for hotels or accommodations whereas not more than 18% people would trust chatbot’s advice for travel planning. This figure further drops down to 12% in case of buying products from a pharmacy and to 9% only for fashion/style counsel.

Implementing chatbots within these enterprises will restrict the business from providing an exemplary customer experience because even after extensive training, chatbots may fail to reply to their FAQs intuitively.


When Chatbot’s Don’t Fail

Even though artificial intelligence has made quite a bit of progress in the last few decades, still at the end of the day, they require human assistance to work seamlessly. Hence, in certain scenarios, human intervention will always triumph over an artificially created robot.