Building Trust in AI Chatbots: The Challenge for Customer Service

Building trust key as companies pivot to chatbots for customer service: experts

TORONTO -- Mohammed Halabi has spent countless hours negotiating with customer service agents over the past 20 years. Halabi is the director of MyBillsAreHigh.com, a company that finds savings on telecom and internet costs for both businesses and individuals. That means seeking out the best deals to fit clients' circumstances, plus taking the lead when problems arise requiring the attention of their provider. These days, there's just one problem. He can't seem to get anyone on the phone. "I've never seen customer service this bad," said Halabi. "I've seen such an influx of customers emailing me issues that I have never seen before in terms of customer service -- businesses and consumers. This is an issue across the board and it's a very frustrating experience." Halabi said hold times seem longer than ever -- over an hour on a good day -- which he believes is a reflection of telecoms shedding local support staff to cut costs. It also coincides with a "push to automation," he said, as providers encourage customers to skip the ordeal by trying to resolve issues solo, with the help of their AI-powered online tools. Halabi's experience in the telecom world echoes a broader trend in recent years. Various industries, from banks to airlines, increasingly rely on customer-facing digital chatbots to fulfil troubleshooting tasks as automation grows more prevalent. "You're having to crawl through hoops to try to get your money back and that's just ridiculous," Halabi said. "People just need people to speak to who are knowledgeable." Researchers say that as AI chatbots become more commonplace, companies must find ways to alleviate concerns about trust, while balancing them with old-fashioned human-to-human conversations. The worldwide chatbot market has grown from US$370 million in 2017 to about US$2.2 billion last year, according to figures cited in a study last month by Vivek Astvansh, an associate professor of quantitative marketing and analytics at McGill University. Astvansh said he feels "perplexed" as companies seemingly adopt chatbots as a universal one-size-fits-all approach to customer service. A case study conducted by Astvansh in partnership with an international footwear retailer showed customers' satisfaction with automated tools varies significantly based on their motivation for requesting support, he said. "When shoppers want sensitive or precise information, they seek human agents, not chatbots," said Astvansh. He noted that some people, for instance, abstain from giving a chatbot their credit card information. Many also prefer dealing with human agents who are able to use more discretion when offering solutions to problems. Whereas a call centre employee might bend the rules of a 30-day return policy by a day or two, an AI bot "will clearly turn this down because chatbots, like all computer programs, are rule-based," Astvansh said. When facing complex or emotional issues, around three-quarters of Canadians prefer to speak with someone over the phone or even in person, according to a report released earlier this year by software company ServiceNow. Its Consumer Voice Report, created in partnership with Opinium, showed that "empathy is fundamental" to the customer service process. But AI tools have seen a recent drop in consumer confidence, as Canadians "have begun to describe their interactions with AI as emotionally disconnected," the company said. Around 44 per cent of Canadians reported they were open to using tools such as chatbots in 2024 and one-in-10 this year said they trust voice assistants to handle basic support tasks. "This highlights a clear gap between the promise of AI and how many are experiencing and interacting with it today," the ServiceNow report said. "As AI adoption grows, so must its ability to earn customer trust." In the telecom sector, major players have touted their growing digital support options. Bell says its AI "virtual assistants" are "a far cry" from older models that struggled to operate outside of preprogrammed steps. That's helped reduce the rate of conversations abandoned to nine per cent across nearly four million interactions since 2024. It says its chatbots accurately understand 97 per cent of customer queries and can "carry on natural-sounding conversations" in English and French, freeing up call centre agents to handle more complex issues. Nazim Benhadid, chief technology officer for Telus, said in an interview last year that his company views the role of AI as being about "assistance." He said Telus' AI troubleshooting tool for customers and employees had significantly increased the number of support tickets closed automatically. For companies, Astvansh said other situations might justify the choice to replace human labour with digital tools. He said customers prefer dealing with a chatbot when seeking "off-the-shelf information" such as product pricing and deals. In addition to being the cheaper option, he said many firms likely see an advantage in the level of consistency that automation can bring compared to an employee. "Human beings are very subjective, idiosyncratic," he said. "Even if I match the same customer to the same agent, that same agent's performance level can vary from one day to the next. Mood swings happen. We are having a rough day, a good day. Chatbots offer that consistency in customer experience." But that same robotic nature can also spoil a shopper's opinion of the company when chatbots and voice assistants "come across as creepy" said Daniel Wigdor, CEO of AI venture studio AXL. Wigdor, a computer science professor at the University of Toronto, called it a major risk as companies seek to build trust for these systems. For many customers, he said the tipping point is when the AI seems to be trying to collect too much personal information during a conversation. "If right off the top, (the tool) is saying things like, 'Give me your first name, give me your last name, give me your phone number, your email address,' and all the person is trying to do is ask a question about a product and a website, then that trips over their creepiness alarm," he said. "They're likely to disengage with the tool." Wigdor said companies developing AI chatbots also need to be careful not to replicate the flaws inherent to traditional customer service interactions, such as human biases. "The thing to understand with these AI is you're training it from human data. So if you're training it with 10,000 recordings of previous calls and people are responding badly to folks with accents, or gender effects, your AI will do that," he said. "You have to be really careful with setting up your data well and properly and then moving past risks." As a consumer advocate, Halabi said he doesn't think chatbots can replicate a quality customer service experience. He said the last thing an angry customer wants to do after finding an error on their bill is argue with a robot. it can also be more stressful to navigate the automated process for those less technologically adept, such as elderly customers, he added. Halabi said customers simply "want to speak to humans." "People are not asking for all of this automation. It's not making our lives easier," he said. "I want to deal with somebody who can understand my emotion, speak similar to me, be in the same area as me, understand our circumstances. That provides for a better outcome. You need a connection, a human-to-human connection."

Building trust key as companies pivot to chatbots for customer service: experts

- has spent countless hours negotiating with customer service agents over the past 20 years. Halabi is the director of MyBillsAreHigh.com, a company that finds savings on telecom and internet costs for both businesses and individuals. That means seeking out the best deals to fit clients' circumstances, plus taking the lead when problems arise requiring the attention of their provider. These days, there's just one problem. He can't seem to get anyone on the phone. "I've never seen customer service this bad," said Halabi. "I've seen such an influx of customers emailing me issues that I have never seen before in terms of customer service -- businesses and consumers. This is an issue across the board and it's a very frustrating experience." Halabi said hold times seem longer than ever -- over an hour on a good day -- which he believes is a reflection of telecoms shedding local support staff to cut costs. It also coincides with a "push to automation," he said, as providers encourage customers to skip the ordeal by trying to resolve issues solo, with the help of their AI-powered online tools. Halabi's experience in the telecom world echoes a broader trend in recent years. Various industries, from banks to airlines, increasingly rely on customer-facing digital chatbots to fulfil troubleshooting tasks as automation grows more prevalent. "You're having to crawl through hoops to try to get your money back and that's just ridiculous," Halabi said. "People just need people to speak to who are knowledgeable." Researchers say that as AI chatbots become more commonplace, companies must find ways to alleviate concerns about trust, while balancing them with old-fashioned human-to-human conversations. The worldwide chatbot market has grown from in 2017 to about last year, according to figures cited in a study last month by Vivek Astvansh, an associate professor of quantitative marketing and analytics at . Astvansh said he feels "perplexed" as companies seemingly adopt chatbots as a universal one-size-fits-all approach to customer service. A case study conducted by Astvansh in partnership with an international footwear retailer showed customers' satisfaction with automated tools varies significantly based on their motivation for requesting support, he said. "When shoppers want sensitive or precise information, they seek human agents, not chatbots," said Astvansh. He noted that some people, for instance, abstain from giving a chatbot their credit card information. Many also prefer dealing with human agents who are able to use more discretion when offering solutions to problems. Whereas a call centre employee might bend the rules of a 30-day return policy by a day or two, an AI bot "will clearly turn this down because chatbots, like all computer programs, are rule-based," Astvansh said. When facing complex or emotional issues, around three-quarters of Canadians prefer to speak with someone over the phone or even in person, according to a report released earlier this year by software company . Its Consumer Voice Report, created in partnership with Opinium, showed that "empathy is fundamental" to the customer service process. But AI tools have seen a recent drop in consumer confidence, as Canadians "have begun to describe their interactions with AI as emotionally disconnected," the company said. Around 44 per cent of Canadians reported they were open to using tools such as chatbots in 2024 and one-in-10 this year said they trust voice assistants to handle basic support tasks. "This highlights a clear gap between the promise of AI and how many are experiencing and interacting with it today," the report said. "As AI adoption grows, so must its ability to earn customer trust." In the telecom sector, major players have touted their growing digital support options. Bell says its AI "virtual assistants" are "a far cry" from older models that struggled to operate outside of preprogrammed steps. That's helped reduce the rate of conversations abandoned to nine per cent across nearly four million interactions since 2024. It says its chatbots accurately understand 97 per cent of customer queries and can "carry on natural-sounding conversations" in English and French, freeing up call centre agents to handle more complex issues. , chief technology officer for Telus, said in an interview last year that his company views the role of AI as being about "assistance." He said Telus' AI troubleshooting tool for customers and employees had significantly increased the number of support tickets closed automatically. For companies, Astvansh said other situations might justify the choice to replace human labour with digital tools. He said customers prefer dealing with a chatbot when seeking "off-the-shelf information" such as product pricing and deals. In addition to being the cheaper option, he said many firms likely see an advantage in the level of consistency that automation can bring compared to an employee. "Human beings are very subjective, idiosyncratic," he said. "Even if I match the same customer to the same agent, that same agent's performance level can vary from one day to the next. Mood swings happen. We are having a rough day, a good day. Chatbots offer that consistency in customer experience." But that same robotic nature can also spoil a shopper's opinion of the company when chatbots and voice assistants "come across as creepy" said , CEO of AI venture studio AXL. Wigdor, a computer science professor at the , called it a major risk as companies seek to build trust for these systems. For many customers, he said the tipping point is when the AI seems to be trying to collect too much personal information during a conversation. "If right off the top, (the tool) is saying things like, 'Give me your first name, give me your last name, give me your phone number, your email address,' and all the person is trying to do is ask a question about a product and a website, then that trips over their creepiness alarm," he said. "They're likely to disengage with the tool." Wigdor said companies developing AI chatbots also need to be careful not to replicate the flaws inherent to traditional customer service interactions, such as human biases. "The thing to understand with these AI is you're training it from human data. So if you're training it with 10,000 recordings of previous calls and people are responding badly to folks with accents, or gender effects, your AI will do that," he said. "You have to be really careful with setting up your data well and properly and then moving past risks." As a consumer advocate, Halabi said he doesn't think chatbots can replicate a quality customer service experience. He said the last thing an angry customer wants to do after finding an error on their bill is argue with a robot. it can also be more stressful to navigate the automated process for those less technologically adept, such as elderly customers, he added. Halabi said customers simply "want to speak to humans." "People are not asking for all of this automation. It's not making our lives easier," he said. "I want to deal with somebody who can understand my emotion, speak similar to me, be in the same area as me, understand our circumstances. That provides for a better outcome. You need a connection, a human-to-human connection."