On a sun-filled San Francisco afternoon, Peter Szulczewski is climbing the stairs to the top of a Sansome Street skyscraper, past floors filled with Wish data scientists and engineers, pool tables and DJ equipment. Large windows give way to a stunning view of the city. But most of Szulczewski’s customers don’t work in offices like this or live in Northern California coastal enclaves. In fact, most of them don’t have much money at all. Wish’s customers are typically working-class Americans from places like the Florida Panhandle or East Texas, Dollar Store shoppers who find Amazon Prime’s $120 annual membership too rich for their blood……..
Businesses today want to boast two medallions – artificial intelligence and machine-learning. As a result, they try putting AI into every potentially-automated process. But is this necessary?
While AI is the answer to several business issues; it is still evolutionary. It has a lot of room for improvement. Can AI be trusted with customer interactions? Yes and no.
If you have been thinking about implementing an AI-based conversational chatbot on an automated interactive user experience, you may need to think again. Consider the following arguments:
1. Do you want to eliminate all human interaction from automated business processes?
Conversational chatbots do not have an unlimited decision-making capacity; in fact, they are very focused. A conversational chatbot can only make conversation based on its training, but using a trained conversational chatbot still results in impediments when interacting with humans. When a chatbot that was designed to provide incremental interaction efficiency with customers cannot fully deal with them, it becomes a liability you don’t want.
One negative bot experience can result in serious business damage. Of customers whose initial bot experience was negative, 73% of them won’t interact with it again. Chatbots run at 85% efficiency – less than the 90% customer satisfaction rates provided globally by call centre executives.
Chatbots should automate redundant processes. How? Every company has a set of queries frequently asked by customers. Using these rules, chatbots help customers; however, a human is still needed to solve the communication puzzle. This is imperative because any query can result in an anomaly that the chatbot hasn’t been trained for.
2. Does Your chatbot have sufficient decision-making capabilities?
Many people believe the most complex part of running a business is the sales or finance aspect; but it’s the decision-making. Humans are intuitive and pragmatic enough to make the occasional sound decision. Even when facing a problem for the first time, a human can make sense out of and deal with it. Therefore, it is logical to provide human employees with a certain degree of freedom when interacting with customers.
For instance, allow your customer representatives to give freebies or special offers when dealing with customers who’ve experienced issues while working with your business. You can do this because you trust their decision-making skills. Can you say the same for a chatbot? Many times, customer executives act outside their job description to leverage their business relationship when helping a customer. This does not immediately affect the business’ bottom line but it does help establish a relationship with the customer. Can a chatbot do the same?
Some say the very essence of neural networks in machine-learning (ML) is to break processes and learn the art of decision-making. That said, any machine-learning engineer would state that ML continues to grow and has miles to go before it will reach the general AI stage. Can you trust this kind of evolutionary technology with decision-making? A chatbot is unable to go beyond its programming to assist customers. A chatbot cannot be trained on complex scenarios where it would need to think “outside the box” to do something. Such actions are subjective, so it is counterintuitive to expect them from your conversational chatbot. It cannot handle complex conversations.
3. Does the idea of posing your chatbot as a human seem fascinating to you?
For many entrepreneurs, incorporating AI into their business processes is something they want, rather than need, to do. This is because they want to see themselves as pioneers in the field. They want conversational chatbots in their processes that are as seamless as human customer executives; but here’s a staggering fact for you – 75% of people want to know whether they are talking to a chatbot or a real person.
The idea of a chatbot being as conversationally fluent as a human may work in science fiction, but real human customers have different expectations. So, if your desire for a conversational chatbot is based on your belief that it will amaze your customers, you should reconsider your decision.
4. Do you have the data, process redundancies and infrastructure to deploy a conversational chatbot?
Human interactions bring empathy to a conversation. This applies to both the executives, who can interact with greater empathy and the customers, who will be more considerate towards an executive. Unfortunately, empathy disappears when chatbots completely dominate the conversation. In fact, 61% of people find it more frustrating when a chatbot, rather than a human, is unable to solve their problem.
To make the conversational chatbot more effective, you need to have the right start. First, determine whether the process has enough redundancy for automation. Second, confirm the appropriate data for chatbot training purposes, because the data used in building is not sufficient preparation for actual human interactions. This is important because customers are less tolerant when a chatbot fails to resolve their queries.
Finally, review the infrastructure to ensure the chatbot has the technological space it needs, i.e. can it integrate with the ERP and CRM systems while still providing secure processes? Businesses need to answer these questions before choosing to deploy a conversational chatbot, and they need to consider their intent behind deploying one.