Chatbots are on the rise, but customers still trust human agents more

Customers regularly contact businesses to purchase products and services, inquire about orders, make payments, and request returns. Until recently, the most common way customers contact businesses was by interacting with human agents via telephone or via the company’s website or mobile app.

With the advent of artificial intelligence (AI), we have seen a surge in chatbots, a new kind of interface. A chatbot is an intelligent software program that allows you to perform two-way conversations with your customers.

Spurring the possibilities of chatbots, closing 24 hours a day, businesses are increasingly routing customers to chatbots. As a result, the global chatbot market has increased from USD 370 million in 2017 to approximately USD 2.2 billion in 2024.

As these tools become embedded in customer service systems, it is important to understand customer preferences and behavior.

Do customers prefer chatbots and human agents?

Despite the chatbot’s business enthusiasm, customers aren’t that sure. A recent survey found that 71% of customers prefer to interact with human agents than chatbots, and 60% of customers also reported that chatbots often don’t understand their problems.

Underlying these preferences is broader skepticism about AI, with the majority of customers reporting low confidence in it.

Most businesses today use chatbots as their first line of customer support. The conversation will only be transferred to a human agent if the chatbot cannot provide the necessary information, and only if the customer asks to talk to someone.

Although efficient, this all-purpose approach can be suboptimal as it may prefer human agents and chatbots of other services for certain services.

For example, a recent survey found that 47% of Canadians are comfortable with their company using their purchase history for marketing, while only 9% have their company use their financial information.

New research provides insights

To better understand how customers actually interact with chatbots and human agents, I partnered with a large North American retailer to analyse the interactions of over 500,000 customer service between customers and agents or chatbots.

Three analyses were conducted in chat transcripts using machine learning methods.

Initially, we focused on why customers reach out to customer service in the first place. We found that most inquiries fall into six main categories: orders, coupons, products, delivery, account issues and payment. Customers rarely traverse chatbots for questions related to delivery or payment. The problem seems to prefer human agents when it contains more detailed or sensitive information.

The second analysis measured how closely the language used by customer service agents (both human and bot agents) matched the language of the customers they were interacting with. We found that human agents exhibited higher language similarity to customers than chatbots.

This result was unexpected. Given the refinement of AI today, we expected chatbots to closely mimic customer languages. Instead, the findings suggest that human agents can better track clients’ diverse and dynamically changing language use.

In the third analysis, we tested a paper that states that similarity produces preferences. This should increase customer engagement by a concept that suggests similarity between human agents with customers.

Customer engagement was measured in average seconds between consecutive messages from customers during a chat. The results show that customers responded more quickly and frequently when human agents showed higher language similarity. When customers felt “understandable,” they became more enthusiastic.

Recommendations for businesses

My findings have made three recommendations for businesses. First, businesses must identify the reasons behind each customer survey before assigning its customers to a chatbot or human agent. The reason for this is to determine whether the company matches its customers with a bot agent or human agent.

Second, both chatbots and human agents need to be trained to coordinate the language and communication style of their customers. For human agents, this kind of mirroring may come naturally, but for chatbots, it must be programmed.

My research shows that customers are more enthusiastic when they feel that the agents they chat with understand them and communicate in a similar way. Doing this will involve customers and lead to more effective and efficient interactions.

Third, companies should ask technology companies for evidence on how much more effectiveness and efficiency a chatbot increases compared to human agents. Specifically, how do their chatbots compare to human agents in terms of efficiency and customer satisfaction? Businesses should only consider using chatbots if the metric exceeds a certain threshold.

Customers want to feel understood and supported. For now, that often means talking to real people. Rather than viewing chatbots as wholesale replacements, businesses should respect customer preferences and treat the right tools as part of a hybrid approach that aligns the right tasks.

This article will be republished from the conversation under a Creative Commons license. Please read the original article.