Artificial intelligence is showing promising results, but customer service employees will immediately feel its effects.
In July 2023, Euronews headlined that AI was coming to take the jobs of customer service employees. Artificial Intelligence broke through into the mainstream over the course of 2023. Ever since, companies were at the edge of their seat to spot the massive potential of the technology to cut costs and streamline their operations. Employees meanwhile are seeing a horror story unfold in real time. Especially those in customer service who are driving into uncharted territory.
Generative AI impacts customer service
In July 2023, the Boston Consulting Group (BCG) released its view on the impact of generative AI on customer service. BCG took OpenAI’s ChatGPT as the starting point for its brief assessment, highlighting that companies have been exploring how artificial intelligence could enhance their customer care departments. As the general public has been playing around with text-based solutions like ChatGPT, the technology is becoming more commonly accepted, BCG noted.
This development leads researchers at BCG to believe that as the technology matures and becomes integrated within customer service operations, productivity could increase between 30 to 50 percent or even more. Adding that in previous surveys, 95 percent consumers expect to be helped by AI solutions in the next 3 years. This doesn’t mean that customer service employees will become completely redundant. The technology is still in its infancy, trained on large, uncurated, datasets. Hence, the AI solutions presently available, require human oversight.
Nonetheless, the stage has been set for large language models (LLMs) to drastically reshape customer service operations. The BCG highlights that LLMs can go beyond current technologies such as interactive voice response (IVR), against assist and chatbots. The datasets upon which generative AI solutions are built are enormous, allowing for enhanced problem recognition and classification, in turn prompt accurate text and speed answers. Algorithms in turn keep building and expanding the LLMs knowledge centers.
The Boston Consulting Group identified five stages of AI-enabled customer service maturity. As of now, customer service centers operate on the first two steps. Customers have access to online portal, apps and can interact with customer service representatives. On top this sits the self service layers, where customers can interact with chatbots and robot process automations.
The next generation of AI-assistant will be able to mimic human support through more complex customer journeys. During the fourth stage, AI will become indistinguishable from humans and able to tailor customized support to customers. In the last stage, generative artificial intelligence will be able to predict problems and solve them for customers. The AI will be able to fully understand customer needs, interacting with other internal systems to provide support.
As promising as the technology looks during these stages, organizations will have to overcome the inaccuracy gap, the BCG notes. Generative AI might provide factual inaccuracies which pose a significant risk during customer service interactions. Solutions proposed by the AI might be incorrect, but due to the strong language models, it is able to deliver them convincingly.
The ChatGPT Customer Service Agent
The Boston Consulting Group takes ChatGPT as one of the primary drivers and precursors for a rapidly changing customer service industry. It’s important to contextualize the role of ChatGPT within the customer care narrative. While its technical prowess is impressive, there are limitations to the technology, as brought forward by major customer service software company Zendesk. Right from the get-go Zendesk clarifies that ChatGPT was never developed as a customer service tool and in its current form, isn’t able to interact with customers. Hence, Zendesk doesn’t recommend implementing ChatGPT in customer care departments at this stage.
Zendesk has an invested interest in this development, but its observations aren’t that far-fetched. Those who’ve played around with ChatGPT will know the tool can deliver false information, presenting it as fact. The same concern was brought forward by the BCG in its assessment. Furthermore, ChatGPT is a language model and not a reasoning model. The way it might jump to conclusions might be irrational. This becomes especially apparent in niche topics, where the answers returned aren’t applicable to reality. Zendesk also points out that ChatGPT can be a major security risk, especially when feeding it customer data.
If OpenAI is able to extend ChatGPT into a full-fledged customer service solution, if it wishes to do so, companies can integrate the solution into their customer service workflows. Zendesk highlights faster support delivery to customers through ticket summarization. ChatGPT would be able to distill lengthy support requests, highlight key errors, helping support agents to find a solution faster. Zendesk notes that this would reduce the average resolution time. By supporting customer service agents, the productivity per employee will increase, Zendesk argues.
This effect was observed during a research study conducted by Erik Brynjolfsson, Danielle Li and Lindsey Raymond from Cornell University, published in April 2023. The researchers analyzed 5,000 customer service interactions at a Fortune 500 company who integrated conversational AI into its customer care operations. They found that issue solving rates per agent per hour, increased by 14 percent. These effects were greatest at novice and low-skilled workers, the team found.
Meanwhile, on the other end of the spectrum, call center agents are aiding these AI systems to enhance their capabilities. In July 2023, The New York Times made a special report on customer service agents at At&T’s call center in Ocean Springs, Miss., who saw their way of work change drastically in recent years. AI tools were making transcripts of calls, which were parsed to their managers, who could analyze their performance. The tool suggested the optimal answer raised by the customer. The customers in turn, were interacting with automated systems more frequently, the news outlet noted.
The technologies rolled out through the AT&T call center weren’t as well received by the employees themselves. Ms. Sherred explained to the New York Times she felt irritation and fear about the tools aiming to improve her productivity, wondering whether she was merely training her replacement. Fearing whether it soon would replace her, leaving her jobless and unable to support her family. Zendesk however remains very optimistic.
The integration of solutions like ChatGPT will result in tailored solutions to customers. Customers who are used to generic replies, that feel robotic, and require multiple interactions with agents to get to the appropriate solution. The immediate effect will be solutions reaching the customer faster, and less time spent getting down to the core of the issue. It must be said that information provided by customers can oftentimes be fragmented or illogical. Perhaps, ChatGPT could support retrieving relevant information, before parsing it through to the customer care employee.
Convincingly incorrect
The way at which ChatGPT delivers information convincingly to users raises debates across the customer service industry. In December 2022, Director of Machine Learning at Intercom Fergal Reid sat down with the Co-Founder and Chief Strategy Officer of Intercom Des Traynor about the enormous hype around ChatGPT about how the technology would reshape customer support. Intercom delivers customer support solutions that help agents interact with customers and automate customer onboarding. The software company is eying the developments in the conversational AI space with great interest, as it’s in the middle of customer support departments.
At the time of the interview, the debate was still going on as to how ChatGPT would interact with customers. While Zendesk stayed clear of highlighting a real life scenario in its initial assessment, Reid warns about the tools factualness. In this particular example, Reid asked ChatGPT how to reset his Intercom two-factor authentication. ChatGPT forwarded Reid to an article referencing Intercom’s help article, but with made up information. All dressed up in a very convincing package. Reid pointed out that ChatGPT was very versed in mimicking human speech, making it look accurate and convincing users the answer it delivers is correct.
Traynor shares the vulnerability in chatbot technologies powered by conversational AI like ChatGPT. If run freely, there’s a high probability these AI systems will deliver wrong answers. This becomes especially true if its databases are fed with sensitive information, such as medical records. Traynor notes these systems shouldn’t make diagnoses, as it can formulate them in a very convincing and natural sounding manner. As it stands now, the accuracy of such systems is unknown.
Job automation
In September 2023, Senior partner Kweilin Ellingrud and partner Saurabh Sanghvi at McKinsey spoke about the impact of generative AI on jobs and workflows in years to come. The interview comes after a McKinsey report about demand for jobs and skills in the future job market. The researchers found that by 2030, up to 30 percent of current hours worked in the US, could be automated. A driving force behind the radically changing job market is the rise of generative AI.
McKinsey distinguishes the pillars behind the changing job landscape into three different categories: automation enabled through generative artificial intelligence, federal infrastructure investments aiding the transition to net-zero and digitalization of the workplace and consumer market. These developments have accelerated over the course of the pandemic, impacting around 75 percent of the American workforce, with the team at McKinsey expecting the trends to persist up to 2030.
Job segments that experienced decreased opportunities are expected to shrink further in coming years. These jobs primarily revolve around customer-facing roles, the team found. This has been a result of the growing e-commerce market, which has reduced customer interactions and office support jobs, as remote work has been welcomed by many employers and their workers.
Kweilin Ellingrud explained that there are growth segments in the job market, most notably transportation jobs, delivery and healthcare jobs. However, Ellingrud believes that major shifts will occur in the food service, office support, manufacturing and customer service indstrusties, where 80 percent of workers are expected to transition to other jobs through either reskilling, upskilling or supporting coworkers to find new employment opportunities.
Customer service employees and the other aforementioned industries don’t stand alone. Ellingrud elaborates by highlighting that generative AI alone could automate 10 percent of tasks in the United States economy across the entire workforce. However, the rate at which the technology will impact workers is much more concentrated in the lower-paid jobs section of the job market, primarily those that earn less than $38,000. Those workers in particular, Ellingrud warns, have a 14 time higher chance of being automated than jobs above $58,000.
One such low-wage worker was Larry Collins, who worked as a bridge toll collector. His job was automated over the pandemic. A decision made by the state to protect employees and drivers. The pandemic, as pointed out by McKinsey, has accelerated the automation of such jobs. Robots were dispatched to clean floors at airports, measure people’s temperatures, with certain hospitals in the United States implementing salad making robots, replacing dining-hall staff.
Senior vice president of cloud and data platform at IBM, Rob Thomas, explained that the pandemic had merely accelerated a development that was bound to happen. McKinsey commented that generation AI will accelerate automation across the labor market. As the technology evolves, it will be able to write code, design products, analyze legal documents and much more. Humans will remain in the loop in the short term however, McKinsey notes.
Organizational uncertainty
The rapid rise of generative AI seeps through each layer of the organization. This translates to companies jumping on the opportunity, but also being uncertain of the outcomes. While many are certain customer service will be directly impacted, and rightfully so, how generative AI will reshape the organization as a whole, remains a great unknown. A January 2024 Deloitte survey found that 79 percent of business leaders believe generative AI will drive substantial organizational changes.
Surprisingly, there seems to be high confidence among many C-suite executives that they have the internal know-how about the organizational benefits of the novel technology within their organization. 44 percent of respondents said they had high or very high expertise in Gen AI, with 9 percent stating to have a very high expertise in the technology. 73 percent have said to have already integrated generative AI into their research and development programs.
However, despite all the optimism around the technology, a large subsection of the respondents experiences a high level of uncertainty. 38 percent of leaders who have some expertise with Gen AI, indicate uncertainty as the most prevalent emotion, with only 9 percent having trust toward adoption. That doesn’t mean that internally there’s no interest in Gen AI. The same group indicates that across their organization there’s high interest in the product, with 38 percent saying that employees are very curious about the transformative capabilities of artificial intelligence.
This heightened tension around the technology, leaves many business leaders worried about the ramifications across the workforce, with 51 percent stating that the rise of generative AI, will lead to more economic inequality. This mimics the analysis from the BCG, who theorizes that AI in due to time will be able to fully assist customers along their journey. Being able to tap into various systems across the organization to deliver tailored solutions to customers.
Artificial Intelligence Strategic Growth Offering lead and principal at Deloitte Consulting, Deborshi Dutt, commented that Gen AI was at an inflection point where organizations were in the beginning stages of determining how the technology could impact their operations. Instead of being on the fence about AI, Dutt urges companies to explore how the product can drive business growth. This requires them to build trust around the opportunities the technology brings and reskill employees to help them adjust working alongside AI.
Cost savings
Uncertainty aside, for organizations there’s much more to be gained in terms of cost savings through the adoption of AI. In an August 2022 report from Gartner, analysts predict that conversational artificial intelligence could lead to $80 billion in cost savings worldwide by 2026. Spending for the technology meanwhile was expected to reach $2 billion over the course of 2022. The American technological research and consulting firm predicts that one in ten agent interactions will be automated by 2026. A strong increase from the 1.6 percent of interactions that are automated today.
VP analyst at Gartner, Daniel O’Connell, commented in the press release, that many organizations struggle to maintain steady customer care operations, despite the current industry size of approximately 17 million agents. On top of labor shortages, companies are looking to cut costs. As the labor pool is very volatile, costs are increasing. Especially those that rely on external call center agencies experience high overhead. Artificial intelligence solutions offer a way out for these organizations.
Allowing organizations to provide quick support to customers, whilst simultaneously cutting costs, O’Connell adds that these technologies can already be useful automating certain portions of the customer care operations. Instead of fully deflecting or containing calls, artificial intelligence technologies can support customer service agents through data collection such as the customer’s name, policy numbers and reason for calling. Implementing these solutions alone can already reduce interaction time by one third, O’Connell notes.
The high savings potential however, will only be achieved through large upfront investments, Gartner warns. The technology itself is still maturing, with many vendors operating in the space trying to claim their share of the market. Investors meanwhile are flocking to conversational AI start-ups who receive hundreds of millions of dollars in cash injections. In 2023, ASAPP alone received $380 million, followed by Observe.ai, who gathered $214 million in funding and Ada with $191 million. Together with Cresa and Woeboth, these five start-ups alone received $1.09 billion in funding.
These factors combined will result in a slower adoption rate over the course of 2024. O’Connell commented that highly technical knowledge is required to implement conversational AI solutions. Before companies can think of rolling out AI across their organization, they first have to have systems in place such as data analytics, knowledge graphs and software to interpret natural language.
Once conversational AI has been implemented O’Connell notes, it has to be continuously monitored, updated and maintained, further driving up overhead costs. Implementation of such complex systems might also take several years, with costs per virtual agent ranging between $1,000 to $1,500, with some going up to $2,000. This means that only companies with large budgets can transition from call center employees to virtual agents and have the resources to keep the systems afloat for extended periods of time.
Customer service employees at risk
Automation in the customer service trade could already be observed for several years as more and more customer interactions move online. Consumer behavior was also shifting in a favorable direction with the rise of conversational AI solutions. A 2023 survey found that 44 percent of consumers were interested in receiving support from chatbots when searching for product information before a purchase. Followed by 35 percent who would like to have customer service support. 34 percent meanwhile would like to receive updates on products on their watch list, with 31 percent interested in receiving customized offerings when shopping online.
However, at this stage, conversational AI is still in its infancy and too expensive for many companies to integrate properly. The technology also borders on the controversial, as it is very apt in delivering false information as correct. If left unchecked, it can do more harm than good in the customer experience. That doesn’t mean that technologies like ChatGPT won’t drastically change the customer case industry. In years to come, as the technology evolves, it will be a strong first line support agent.
Venture capitalists and other investors are showing great interest in the technology, with vast amounts flowing to start-ups who are trying to become the preferred supplier of major corporations. In the short term, this will mean that many players will come and go, with a handful of strong suppliers remaining. On the agent side meanwhile, automation is undeniable. Many positions will be terminated, with customer service employees seeing their roles shift to overseeing chatbot interactions and taking over when the chatbot gets stuck.
The timeline is still unclear. Coming years will be a wild west for many organizations. Some will successfully implement conversational AI solutions, while others struggle to get it up and running, having to fall back on their trusted agents. But once they’ve learned from their mistakes, they too will follow along, greatly reducing their customer support departments.