Will chatbots take over customer service – and replace more jobs done by people?

AI creations set to become ever-more important in our dealings with companies

The use of chatbots looks set to continue, with the technology improving and the market expanding. Photo: Alamy stock

Chatbots are becoming increasingly common, with more than three quarters of consumers saying in a 2022 survey that they had interacted with one in the previous 12 months.

But the poll also highlighted the reality that chatbots do not always result in happy outcomes, with four fifths of respondents frustrated by them.

Research looking at the experiences of UAE-based consumers who had interacted with chatbots used by delivery apps also indicates that the artificial intelligence sometimes fail to please.

In a paper published in December in Balkan Social Science Review, Dr Hassan Mustafa, an associate professor at the Rochester Institute of Technology Dubai, found that users’ experiences were “mixed”, with chatbots often failing to deal with “complex or nuanced” queries.

Dr Mustafa said the availability of chatbots 24 hours a day, seven days a week, and their ability to handle high volumes of queries, made them “indispensable in many sectors”.

But he said that while many users appreciated the speed and efficiency of chatbots in resolving straightforward issues, others found them frustrating, especially when responses were inaccurate or failed to address complex queries.

Speed isn't everything

“According to our research, users value chatbots for quick resolutions, but criticise their lack of empathy and contextual understanding,” Dr Mustafa told The National. “There is a perception among some users that companies use chatbots primarily to cut costs rather than to improve service quality, which can lead to dissatisfaction.”

Adnan Bashir, a technology commentator and global lead for external communications at software company Hansen Technologies, said it was hard to see the ever-growing use of chatbots as “anything other than cost-cutting by enterprises”.

“Especially when we see major, critical, increasingly profitable social utilities such as banks, credit unions, energy companies and telecom service providers utilise them on a large scale,” Mr Bashir said.

He said given that surveys had indicated that trust in businesses was low, outsourcing the first layer of contact between customers and a company to chatbots was “ill-advised”.

“People want to be seen. People want to be heard. People want to feel that they are being taken care of,” Mr Bashir said.

Improving technology

While not always popular, chatbots are becoming more capable thanks to breakthroughs in natural language processing (as shown by the rise of AI technologies such as ChatGPT) and machine learning.

In a recent study, researchers at the University of Leeds in the UK found that “retrieval-augmented generation” (Rag) chatbots “show significant improvements” at dealing with queries from students. They were found to be more effective than “traditional intent-based” chatbots, which formulate responses according to the perceived intent or purpose of requests by users.

One intent-based chatbot familiar to many is Amazon's Alexa which is programmed with “standard question types and answers", said Eric Atwell, a professor of artificial intelligence for language at the University of Leeds and one of the authors of the study.

“This works fine for standard questions, but may not handle more complex or one-off questions,” Prof Atwell said. “Large-language model chatbots such as ChatGPT are trained with huge amounts of text, all of Wikipedia plus many discussion forums to answer almost any question.

“But they generate answers from snippets of training text and may put together snippets which do not add up to a sensible or correct overall answer. Recent improvements come from combining these approaches in new ways.”

The Rag chatbots were better because of their ability to combine the capabilities of retrieval-based chatbots, which rely on pre-written responses, with the generative power of AI. Dr Mustafa said chatbots have become more effective at understanding user queries, even if these are phrased informally or with incomplete sentences.

Chatbots have also improved in their ability to understand what we are thinking, with features such as sentiment analysis allowing them to adapt their responses based on perceived emotions, making interactions more personalised.

Widespread use

Tying in with these advances, chatbots are being used for an ever-wider array of functions, such as with helping to place orders, deal with complaints and provide updates.

“Beyond that, they are being employed in marketing to suggest products, in healthcare for appointment scheduling and in e-commerce for resolving enquiries,” Dr Mustafa said.

The significance of chatbots extends beyond their impact on customer experience, as they are performing many roles that a person would previously have been employed to do. Just as jobs in remote customer service are at risk because of AI, so are they being lost in other fields such as data entry and administration, graphic design and retail.

Mr Bashir said he did not think chatbots would “take away everyone’s jobs”, although some positions could go as basic customer queries could be automated.

“Higher-level tasks that require escalation to a manager or director are not going away,” he said. “As is the case with generative AI in any kind of writing or creative endeavour, there will be a need for the human touch, and for a level of analytical and nuanced thinking to solve pressing customer issues.”

He said that for companies to strengthen public trust, they should realise that customer service “must be truly customer-centric” and retain a human element.

But as chatbot technology advances, will this happen? Looking ahead, Dr Mustafa said he thought that chatbots would become harder to distinguish from people as their ability to understand context, emotion and nuances advanced.

“Features like voice-based interactions, real-time personalisation and memory retention of past interactions could further blur the line,” he said. “However, ethical considerations, such as transparency about chatbot use, will remain crucial as these technologies evolve.”

Ultimately, he thought that they would be replaced by “virtual agents”, which are better able to interact with customers on a conversational level.