“I alerted three employees who showed complete indifference towards me…so began a yearlong saga of pass the buck, don’t ask me, and ‘I’m sorry sir, your claim can go nowhere.’”
United Breaks Guitars, Dave Carroll (2009)
This paper begins 20 years ago, long before the rise of AI, when Myspace hadn’t yet folded to Facebook, and Amazon had only recently become profitable. Here, on the ninth floor of the Starbucks corporate office, I engaged an analyst in a spirited though misguided debate over why people were so loyal to the brand. I’d spent two years in the call center, listening to customers from Los Angeles to New York say the same thing: “If I got this service at McDonalds, I wouldn’t care; but at Starbucks I expect more.” Prior to that, I’d spent another two years in the stores “surprising and delighting” customers in accordance with the company’s mission. I actually believed that we were changing the world one cup of coffee at a time, and my conversations with customers across North America proved as much. Or so I thought.
“They’re statistically irrelevant,” the analyst said. “All the people you’ve talked to don’t matter.” An awkward silence followed and the conversation soon ended, but it marked the moment that I started to question the nature of my reality. At that time Starbucks served around twenty-five million customers a week, while the number of calls handled by the contact center was somewhere in the low tens of thousands. On a good day there might be a few pleasant conversations, on a bad day it was hard to comprehend how the company stayed in business. The reality, however, was that all of these complaints amounted to less than one-percent of Starbucks’ daily volume. The overwhelming majority of customers were satisfied. The complainers, even those who parroted the customer connection Howard Schultz believed was essential, weren’t representative. People went to Starbucks because it was there. The success of the company wasn’t based on surprising and delighting customers, it was premised on consistency and market saturation. As one VP of store development put it, “If we don’t keep opening stores, someone else will.” The connection to the brand was, to a certain extent, fungible, and with it, the need to personally apologize for failed expectations. And so it was a year later that Starbucks announced it was outsourcing its call center to an outfit in Nevada. From that point forward, if you called customer support, you’d not be talking to the company.
The cost-cutting target on customer service is nothing new. Call centers had been outsourced for decades prior to Starbucks’ decision. However, the rise of artificial intelligence has presented companies with a new opportunity to replace agents, provide more consistent and available service, and keep headcount flat while continuing to grow. This paper explores the growing tension between reducing costs and providing a human connection. It reviews how companies value their AI investments, the importance of good partners, and whether they should even listen to their customers at all.
Background & literature review
The cloud of hyperbole and fear surrounding artificial intelligence makes discerning fact from fiction extraordinarily difficult. When Amazon announced it was laying off thirty-thousand employees, The Wall Street Journal was quick to attribute the job cuts to AI (Ellis et al 2025). However, CNBC argued that CEOs are under increasing pressure to show gains from artificial intelligence, even if reductions in force aren’t the result of AI. Intelligent automation, they say, provides a convenient excuse for layoffs that would have happened anyway (Morabito, 2025). Furthermore, while Amazon CEO Andy Jassy extolled the potential for artificial intelligence to reshape the customer experience (Coleman, 2023), it’s doubtful the company has figured out how AI should be used, much less deployed a solution at scale.
Executives may want to implement AI but determining how and where and which vendor to choose is difficult. Furthermore, the market is saturated with AI products that offer little actual value. According to an MIT report on AI adoption, one executive described the bulk of AI products as science projects, adding that of dozens of demos, only one or two were genuinely useful (Challapally et al, 2025, p. 7). Similarly, in a report on the use of artificial intelligence in contact centers, McKinsey found that CEOs are flooded with vendors predicting an AI end to human agents. However McKinsey cautions against such exuberance, citing technical challenges, software compatibility, and the continued need for human escalation as reasons to be skeptical of such predictions (Blackader et al, 2025). Nonetheless, interest in AI adoption is real and driving phenomenal growth in investment, infrastructure, and AI-related services.
Business case
It’s difficult to argue that AI doesn’t present tremendous advantages to contact center operations. Microsoft, for example, announced that AI had saved it over $500 million in call center costs in 2024 alone (Babu, 2025). Payment processor Klarna (2024a) claims their chatbots do the work of 700 agents while achieving satisfaction scores that are on par with humans. The company reported that it had saved over $40 million as a result of AI. And, IBM (n.d.) partnered with Camping World to implement self-help and agent-assisted AI solutions that have improved wait times, decreased dropped call rates, and improved agent efficiency by 33%.
AI has also been used to achieve better customer results by creating better human engagements. For example, McKinsey (McK, n.d.) partnered with Deutsche Telekom to implement an AI coach that analyzes agents’ interactions with customers and provides tailored feedback in real time. Deutsche Telekom reported a 2% decrease in transfers, a 10% increase in first call resolution (FCR) and a 14 point increase in net promoter score as a result. While it’s easy to downplay the softer benefits of AI, they carry real impact to the bottom line. For example, Qualtrics found that 53% of bad experiences resulted in customers cutting spend, while customers were twice as likely to recommend if their issue was resolved on first contact (Zdatny & Brown, 2024). Desmarais (2025) echoes these findings, writing that 93% of customers expect their issue to be resolved on the first call, adding that for every 1% gained in FCR companies reduce their operating costs by 1%.
AI can be used to automate agent tasks and transform call centers into data centers. Google and Deloitte partnered with Canadian insurance company Definity to modernize their contact center operations. According to Google (GOOG, n.d.), GenAI has been used to automatically document calls and reduce total handle time to 3 ½ minutes per call. Prior to AI automation, agents had been spending 3-5 minutes on documentation alone.
That documentation is itself an asset to businesses. Morrell (2025b) writes that companies should start viewing call centers as business intelligence hubs, not cost centers, noting that contact center leadership have consistently polled more favorably toward analytics since 2018. Respondents said contact center data helped them spot opportunities for self-service, process failures, and assist with the customer journey. Moreover, data has helped consultants and business operators alike. Mckinsey (McK, n.d.) used vast amounts of call center data to train AI models, identify trends and set KPIs. Google (GOOG, n.d.) uses its VertexAI to store transcripts, emails, and call records for analysis. The data are made available to agents and analysts in real time. Nicastro (2025) sums up these benefits, writing that Contact Center as a Service has turned call centers into revenue generators. In short, AI coupled with contact center data is a powerful tool for driving business insights and competitive advantage.
Finally, Amazon argues that customers want the 24/7 service that only AI tools can provide. Human agents, they write, are expensive and not always available. Furthermore, by handling the majority of contacts through AI, human agents are free to focus on more complex and sensitive issues (Merritt, 2021). In a similar vein, Morrell (2025a) writes that customers will choose the method of contact that is most convenient for them, and the one that fully resolves their issue. In that, there isn’t one right way to service customers, except that they are served quickly and with the right information. AI meets this objective by being constantly available at scale. The benefits are not only increased agent efficiency, but reduced costs and better, more available customer service.
Customer voice
The benefits of AI to businesses may be clear, however, adopters must ask whether automation is actually what their customers want and whether there exists too much of a good thing. Morrell (2025a) writes that companies can’t assume that customers will use digital communication simply because it’s there. As Microsoft (Morris, 2016) and Harvard (Dixon & Tomand, 2010) pointed out, customers will choose the solution that’s easiest to use, regardless of the technology behind it. McKinsey data reinforced this point, noting that 94% of Boomers and 71% of GenZ say calling in was still the easiest way to reach customer care (Blackader et al, 2025).
Further underwriting these preferences may be a broader skepticism of AI. For example, more than half of customers surveyed are concerned that artificial intelligence will replace humans (Zdatny & Brown, 2024). This need for human contact is not new or unique to AI, however. Dichter (2019) argued in Forbes that people’s desire to speak to someone in their own country isn’t nativism, it’s psychological and instantly relatable. In other cases, simply implying a human connection removes psychological barriers. Stewart (2023) summarizes a conversation with Boston University professor Michelle Shell, who says just displaying the option to talk to an agent while interacting with AI puts people at ease. Moreover, the McKinsey data above suggests demographics play a role in shaping our expectations. Morrell (2025a) writes that GenZ generally have less experience with in-person engagements, are more apprehensive about phone interactions, and carry higher expectations of service as a result.
Yet human contact isn’t necessary or even preferred for many interactions. Forrester (Leggett, 2019) found that two-thirds of customers say valuing their time is the most important thing a company can do to provide them good service. Self-service, they argue, meets the customer while they’re engaged, is faster, cheaper, and often comes with higher satisfaction. It’s also worth noting that the nature of the company’s business defines what quality service means. Microsoft (Chang, 2023) writes that B2B customers are more likely to self-help and engage online resources, while as individuals, customers want something closer to white-glove service. In fact, 63% of customers expect service to be personalized, and 80% of customers were more likely to make a purchase if that need was met. Nonetheless, the rise of digital experience, chat, email, and now AI, have allowed companies to radically depart from traditional customer service models. Facebook, for example, famously has no customer service phone number. Reporters for Business Insider noted that the best way to get a human at the company, was to reach out via social media (John and Johnson, 2023). Squarespace (SQSP, 2024) publicly acknowledges its lack of a direct line writing, “we truly believe it wouldn’t be possible to provide the same effective help over the phone.” The company argues that the nature of their product often requires screencaptures, code snips, and other visual representations that are better resolved over email. Online scheduler Acuity (AS, 2025) makes a nearly identical statement, directing customers to its 24/7 chat platform; and Uber (n.d.) says it doesn’t offer a direct line for riders but does for Uber Eats and Business.
The risks of eliminating phone lines will be covered more in the discussion, but there are indications that the approach might not represent customer interests. At the very least, a company’s ability to not provide a direct line is highly dependent on the business it is in. Frontier Airlines for example was among the companies cited by Stewart (2023) who had eliminated their customer service numbers. Two years later, the company has reinstated its 800 line, presumably due to customer pushback. Indeed, the unpredictability of problems and human nature suggest companies should think carefully before eliminating phones. IBM (n.d.) and Amazon (Merritt, 2021) each stressed the importance of maintaining human agents to handle more complex issues. McKinsey (Blackader et al, 2025) went a step further, writing that such complex issues require human empathy and judgement. The higher the mix of complexity and emotion, the more the customer wants a live agent (Morrell, 2025a).
Finally, it’s worth noting that trends in the data support the persistence of phone lines. According to Morrell (2025a), phones still account for 60% of contacts and though that number plateaued in 2022, it hasn’t declined. McKinsey reported similar data, noting that while digital interactions have increased 6% since 2010, phone calls still gained 2% over the same period. In other words, they write, though AI might take a larger share of interactions, the total number of phone calls are likely to increase over time (Blackader et al, 2025). The resilience of human interactions was further supported by Qualtrics data. In the same survey of customers who voiced concern that AI would replace humans, 61% said they preferred a human contact. And 74% preferred humans when the problem was technical (Zdatny and Brown, 2024).
In summary, customers want fast, easy service that solves their problem. How these needs are met is highly dependent on our age, level of frustration, technical aptitude and relationship to the brand. It’s clear from the literature reviewed here, however, that while some companies have eliminated phone numbers, talking to a live person remains not only the dominant form of contact, but a highly persistent one as well.
Challenges
Getting AI right and deciding whether it should be used at all are two of the biggest challenges facing decisionmakers. Media coverage of AI and the gravitational lensing that comes with it, distorts perceptions and suggests that businesses are adopting AI at rates that match its growth. In fact, the opposite is true. MIT reported that while 60% of surveyed businesses said they’d reviewed enterprise AI solutions, just 20% developed a pilot, and only 5% reached production (Challapally, 2025, p. 3). Lindner (2025) cites better but still paltry statistics indicating that 74% of CRM initiatives fail to meet goals due to inadequate technical integration. These low success rates suggest that companies are either not finding value in AI or are unable to overcome the technical challenges inherent in enterprise-wide initiatives.
The aforementioned deluge of AI vendors and startups offering unproven products is one challenge reported by Blackader et al (2025). However, identifying a quality vendor is only part of the challenge. Technical and cultural resistance to change are major factors facing new initiatives. For example, Nicastro (2025) writes that lack of trust is a significant hurdle to AI adoption. Call center agents don’t fear AI so much as they fear a bad rollout of the technology. Employees want artificial intelligence to make their jobs easier, not micromanage them. Similarly, Challapally et al (2025) write that users prefer GPT for mundane tasks, but 90% of those surveyed still prefer humans for mission-critical work. These findings suggest that a solution might look good on paper but fail to achieve real value in practice.
Even if social resistance is mitigated, controlling costs remains a non-trivial concern. Enterprise-wide implementations of AI can run into the hundreds of thousands of dollars and be difficult to forecast. For instance, a Gartner study found that over half of companies surveyed underestimated the initial costs of implementation by 30-40%. Hiring data scientists, ML experts, and training existing staff also carries ongoing costs that are often under-appreciated (Piccolo, 2025).
Budget concerns aside, determining the right implementation of AI can be difficult. Vendors often don’t understand a client’s approval process or dataflows, and while generic chatbots have high pilot to implementation rates, they’re often brittle, over-engineered, and not aligned with workflows. One executive interviewed by MIT said that if the technology doesn’t integrate with Salesforce, no one’s going to use it (Challapally et al, 2025, pp. 7-15). Similarly business results can be equally elusive. For example, McKinsey (Blackader et al, 2025), found that very few companies were able to achieve meaningful reductions in contact center volume. Those who did, they write, were able to solve the aforementioned data and integration issues. Yet, even relatively simple engagements can be difficult to remediate. For instance, McKinsey found that 50-60% of contact center interactions remain transactional, despite efforts to reduce them (Blackader et al, 2025). The resilience of simple contacts should at minimum serve as a reality check to the limitations of AI.
Finally, politics and data are concerns for leadership and vendors alike. Dixon and Toman (2010) wrote in the Harvard Business Review that companies often fail to recognize the importance of providing basic service and instead focus on metrics that don’t matter. For example, Harvard’s survey of contact center leadership found that 89% of respondents said exceeding expectations was their primary strategy; yet the authors argue that this approach is flawed, noting that the data show that more satisfied customers are not necessarily more loyal customers. The challenge is no less for vendors tasked with implementing AI initiatives. In their case study of Deutsche Telekom, McKinsey (McK, n.d.) noted that the most challenging aspect of implementation was identifying relevant information. Morrell (2025b) covers this problem from a production perspective, writing that failures in automation often lead to callbacks, but fewer than half of companies said they had sufficient data to identify the root cause. These comments speak to the importance of having quality data, clear goals and understanding what matters.
In some cases, the difficulties start with articulating value to the CEO. Surveyed executives expressed trouble framing the value of AI outside of sales and marketing, even though backend applications of artificial intelligence often carry higher returns (Challapally et al, 2025, pp. 10, 20). In other cases, risks arise from expectations that don’t match the chosen technology. For example, a lawyer interviewed by MIT said GenAI’s lack of memory and access to prior conversations made it impractical to use. This lack of persistence was cited by researchers as a major reason AI projects fail (pp. 12-14). Yet, MIT data also found that there was near unanimous individual use of AI across the board, indicating that while corporate initiatives might fail, employees are still finding value in AI (p. 8). That said, these results are not necessarily representative of contact centers. For example, Qualtrics cited in-house research that showed only 20% of agents actively use AI (Zdatny & Brown, 2024), a far cry from the unanimous adoption rates cited by MIT. Regardless, these findings suggest that AI has a place in day-to-day workflows, but enterprise solutions often fail to meet the employee where AI is needed most.
In short, the risks to successful implementation of AI are equal parts social, political, technical, and operational. Simply throwing AI at a problem does not ensure that the problem will be solved or that the project will be completed. Competition between stakeholders and business units also plays a role in project success. Finally, technical challenges, lack of vendor domain knowledge, and process incompatibility all conspire to make forecasting costs difficult. How these challenges can be met will be discussed more in the next section.
Remediating challenges
Clearly there are challenges facing AI that are not easily resolved. Yet the gains in efficiency and revenue cited by MIcrosoft, Klarna, IBM, and others, show that AI can be done right. AWS provides a list of best-practices that begins with starting small. Amazon recommends focusing on transactional asks like password resets, address updates, and routing to the right person (Merritt, 2021). While McKinsey’s (Blackader et al, 2025) research shows that transactional contacts can be stubbornly persistent, focusing on these simple tasks is low risk while still providing value. Alternatively, IBM (n.d.) took a phased approach to their development of Camping World’s virtual assistant, a strategy that both limited risk and proved essential to the project’s success. And, in their case study of Definity, Google (GOOG, n.d.) focused their initial rollout on documenting calls before taking on other tasks. Definity says it is using the success of this program to pursue enterprise applications of AI. In a nod to MIT’s recommendation that adopters focus on backend systems, Google says the call center’s low risk and high ROI made it an ideal starting point for AI adoption. Their tight focus on streamlining documentation further aided the project’s success.
Though starting small might encourage companies to develop solutions in-house, finding a quality partner remains essential. According to MIT, strategic partnerships between vendors and clients are twice as likely to succeed as internal builds (Challapally et al, 2025, pp 19-23). Indeed each of the case studies above involves a major technical partner, whether Google, IBM, or McKinsey. Vendors can also help solve requirements and technical gaps. For example, in their work with Deutsche Telekom, McKinsey (McK, n.d.) cited their commitment to spending time with agents, understanding workflows and the team’s learning process as central to their approach. This effort helped them identify information overload as a major inhibitor to agent learning. Separately, IBM worked with Camping World to tailor solutions to their various customer segments, and Google’s (GOOG, n.d.) work with Definity addressed specific operational pain points. In all cases vendors took the time to understand the client’s objectives and craft solutions that resolved front-line issues.
Focusing on making the process easy was also a recurrent theme throughout the literature. In their best practices, Amazon writes that solutions need to communicate who the customer is, what they’re doing, and where they are in the system and pass that information to the agent. The customer, they say, should not have to repeat themselves (Merritt, 2021). Dixon and Toman (2010) make a similar point, arguing that while good service does little to reinforce a brand, bad service can do a lot to undermine it. Therefore, it’s important to make interactions with the company as easy as possible. Finally, Microsoft writes that customers want the fastest solution for the least amount of effort, pointing out that 28% of respondents cited agent ineffectiveness and 25% cited lack of self-help as their top complaints. Solutions, they say, should not only reduce friction but empower employees and customers to solve problems (Morris, 2016).
Finally, it’s important to find the right balance between AI initiatives and organic use. MIT argues that successful implementations do three things: buy not build, select tools that integrate deeply and adapt over time, and empower managers to make decisions instead of adopting a centralized solution (Challapally et al, 2025, p. 23). Microsoft (Chang, 2023) offers slightly different criteria, defining success as keeping solutions lean and consolidating initiatives around a single cloud platform. It’s not clear to what extent the case studies above follow these recommendations. Certainly Definity’s desire to implement enterprise applications of AI suggests centralized planning. However, the high personal use of AI cited by Challapally et al (2025) indicates that decentralized approaches have merit. Nicastro’s (2025) argument that agents don’t fear artificial intelligence but a bad rollout of the technology, alludes to a similar point. In that, a more organic approach to AI might help drive adoption. Klarna (2024b), for example, has embraced a ground up approach, encouraging employees to test applications of AI in their daily work. This decentralized model, the company says, has resulted in a 90% employee adoption rate of artificial intelligence.
In summary, successful initiatives will tailor solutions to customers’ and employees’ needs. They will be decentralized and focus on making interactions with the company as easy as possible. Their success hinges on finding a quality partner, starting small, and overcoming organizational resistance, both to using AI and to how it’s deployed.
Discussion and outlook
I am admittedly biased in my view of contact centers. On the one hand, I’m sympathetic to the nature of the role and the ownership one takes over issues they had no part in creating. It’s drudgerous, yes, but important, I believe, to own your mistakes, not outsource them to someone else. The notion of deflecting responsibility to third parties or a fleet of robots, offends my sense of accountability and certainly wouldn’t fly in my career leading enterprise initiatives. On the other hand, all of those contacts are very likely statistically irrelevant. In a free market, real risk comes from over-emphasizing what doesn’t matter, and so companies are incentivized to make decisions that benefit the business, not a fractional subset of individuals. Yet this remains an abrasive somewhat cynical pill to swallow, regardless of how true it may be. And so, perhaps it’s foolish to say, but surely there exists a reasonable middleground, where cake can be had and eaten too.
To begin with, no one wants to feel as though they don’t matter, whether that’s customers or contact center representatives. The expectations of personalized white-glove service stated by Chang (2023) and the cross-generational preference for human contact covered by Blackader et al (2025) are indicators of that fact. That said, one could be forgiven for believing that if every company could do away with direct lines, as Facebook, Squarespace, and Uber have done, that they would. After all, a 2018 survey by the National Association of Call Centers, suggests that a non-trivial percentage of organizations are biased in that direction. According to Dichter (2019), 45% of companies were either totally or mostly cost-driven, viewing the primary purpose of the contact center being to benefit the company, not the customer.
It’s no surprise, therefore, that reduced costs is one of the most consistent benefits cited throughout the literature. Amazon, IBM, and McKinsey all pushed cost-savings as a principal benefit of AI (Merritt, 2021; IBM, n.d.; McK, n.d.). In an interview with Vox, Harvard professor Ryan Buell said companies recognize that these costs exist on a sliding scale. Humans are the most expensive, robots are the cheapest (Stewart, 2023). Yet there exists a second sliding scale between costs and customer expectations that prevents companies from abdicating all responsibility to meet customers in person. The preference for human interaction across multiple surveys, including Qualtrics (Zdatny and Brown, 2024) and McKinsey (Blackader et al, 2025) are indicators of this need. However, there are more pervasive factors that won’t be remediated simply by the passage of time. Dichter (2019) writes that the rise of smart phones and social media have allowed customers to reach millions of viewers the minute their expectations aren’t met. The implication being a PR disaster for companies who prioritize cost saving over quality service. United Airlines famously suffered this fate after refusing to reimburse musician Dave Carroll (n.d.) for damaging one of his guitars. The artist responded by writing a song that went viral and gained traction in the mainstream media.
Nonetheless, it’s foolish to expect companies to avoid finding ways to lower costs. AI uniquely applies to call centers in ways that it may not to other business units. It’s also true that very few agents will complain if artificial intelligence means seeing fewer calls in queue. However, this doesn’t mean that companies want to broadly eliminate all humans from the contact center. In fact, Dichter (2019) writes that the rise of offshoring that came about in the 1990s has since shown signs of reversing. For example, there are over 600 call centers, employing over a quarter million people in Texas alone; and Senator Gallego (2025) of Arizona recently introduced legislation to keep call centers in America. Furthermore, IBM (n.d.) acknowledged that while Watson reduces the need for human contact, it doesn’t replace it. And McKinsey (Blackader et al, 2025) says retaining quality human agents will become a competitive advantage over the coming years. In that, Facebook, Squarespace, and Uber are outliers that are allowed to forgo human contact because the nature of their business allows them to do so. Perhaps this is the slice of cake my inner idealist desires. The sentiments above and their permutations cited throughout this paper, suggest, if nothing else, a desire to keep humans involved. I believe, however, that human involvement is far more obligatory than that statement suggests, particularly in the age of social media. Ironically, however, its value may not be fully appreciated until it's nowhere to be found.
Outlook
Hanging over the human argument is the very fact that AI might be giving people exactly what they want. After all, Microsoft (Morris, 2016) and Amazon (Merritt, 2021) argued that customers want service that is fast, easy, and available. Not necessarily a human to talk to every time they need help. Admittedly it is tempting to apply these perspectives unilaterally across the spectrum of contacts. But as was shown in the literature review, context, age, technical aptitude, and frustration level all play a role in determining whether a human conversation is required. All else being equal, however, simply offering a path to humans is often enough (Stewart, 2023). These competing vectors present a compelling conversation for exactly what the future holds and to what degree AI will take over customer service.
To start, call centers are one of the departments most likely to be impacted by artificial intelligence. MIT, for example, concluded that AI was already driving 5-20% reductions in force across the industry (Challapally et al, 2025, p. 21). While it’s easy to view this impact as a much broader harbinger of things to come, that’s likely not the case. The surveillance required to gain sufficient data to inform intent, tone, and mood are beyond what most of us are willing to concede. Contact centers have no such barriers, however. Conversations are recorded, screens monitored, chats logged, feedback taken, and to a certain extent, shared across industries, in ways the salaried worker would vociferously resist. In the call center, however, this is standard practice. Even bathroom breaks are monitored. The environment, therefore, is uniquely positioned to deploy AI in ways that are difficult to replicate elsewhere. In that, they may be a laboratory for the AI apocalypse. Or it may simply be overblown. After all, 57% of customer care leaders told McKinsey they expected call volumes to continue increasing over the next several years, and the same study noted that while AI has helped remediate transactional calls, volumes have continued increasing due to new and different needs (Blackader et al, 2025). Morrell (2025a) echoes these findings in his article titled The Great Contact Center Standoff, noting the paradox between the rising use of AI assisted service and customers’ increased preference for phone interaction. Gartner (2025) argues that not only will no fortune 500 company have fully eliminated humans from customer service, but that to do so would be highly undesirable. While the company capped these predictions at 2028, the persistence of phone calls, the fact that a strong majority of GenZ still prefer human conversation, and the evolving nature of new and different reasons to call support, all suggest that agents will be answering phones for years to come.
This points to a future where AI dominates the low hanging fruit while organic intelligence handles the more sensitive issues. Moreover, the persistence of human agents in a field primed for AI takeover, bodes well for the resilience of human labor in other roles. If we can survive in contact centers, we should be able to find a toehold in engineering. More importantly, this evolution suggests a fundamental shift in what the role demands. Nearly every source cited here referenced the importance of maintaining a human connection particularly for complex scenarios. This suggests that agents of the future will need to possess high emotional maturity, sound judgement, and empathetic self-awareness. It’s also possible that AI might alleviate our need to call contact centers in the first place. McKinsey (Blackader et al, 2025) hypothesizes that AI assistants will call companies, make dinner reservations, and square our utility bills on our behalf. Indeed, such interactions are already taking place in limited fashion. OpenTable, for example, is already using AI to help customers with their dinner plans. The budding rise of AI assistants even garnered attention from the New York Times (Guay, 2025) who ironically concluded that the best assistants had the least amount of AI.
That said, predicting the future amidst all the hyperbole and fear surrounding AI is next to impossible. However, a few general landmarks can be established. First, AI isn’t going away. Its impact on contact center jobs is real and potentially substantial. Second, AI has a use case, in part, because it provides the fast, easy service consumers want. Lastly, humans aren’t going anywhere. Despite all of our intelligent automation, devices, and need for instant gratification, sometimes what we need most is simply another human to talk to.
References
AS. (2025). Why Acuity doesn’t offer phone support. Acuity Scheduling.
Babu, J. (2025). Microsoft racks up over $500 million in AI savings while slashing jobs,
Bloomberg News reports. Reuters. https://www.reuters.com/business/microsoft-racks-up-over-500-million-ai-savings-while-slashing-jobs-bloomberg-2025-07-09/
Blackader, B., Buesing, E., Amar, J. & Raabe, J. (2025). The contact center crossroads:
Finding the right mix of humans and AI. McKinsey & Company. https://www.mckinsey.com/capabilities/operations/our-insights/the-contact-center-crossroads-finding-the-right-mix-of-humans-and-ai
Carroll, D. (n.d.). United breaks guitars. Dave Carroll Music.
https://www.davecarrollmusic.com/songwriting/united-breaks-guitars/
Carroll, D. [sonofmaxwell]. (2009, July 6). United Breaks Guitars. [Video]. Youtube.
https://www.youtube.com/watch?v=5YGc4zOqozo
Chang, A. (2023). 5 customer service trends to watch in 2023. Microsoft.
Challapally, A., Pease, C., Raskar, R. & Pradyumna, C. (2025). The GenAI divide: State
of AI in business 2025. MIT. https://mlq.ai/media/quarterly_decks/v0.1_State_of_AI_in_Business_2025_Report.pdf
Coleman, J. (2023). Amazon CEO says generative AI will ‘change customer
experience’. CNBC. https://www.cnbc.com/2023/12/07/amazon-ceo-says-generative-ai-will-change-every-customer-experience.html
Desmarais, M. (2025). First call resolution: A comprehensive guide. SQM.
https://www.sqmgroup.com/resources/library/blog/fcr-metric-operating-philosophy
Dichter, T. (2019). Call centers return to the U.S. - More companies get the link between
customer service and profit. Forbes. https://www.forbes.com/sites/thomasdichter/2019/03/30/call-centers-return-to-the-u-s-more-companies-get-the-link-between-customer-service-and-profit/
Dixon, M., Freeman, K. & Toman, N. (2010). Stop trying to delight your customers.
Harvard Business Review. https://hbr.org/2010/07/stop-trying-to-delight-your-customers
Ellis, L., Tucker-Smith, O., & Pohle, A. (2025). Tens of thousands of white-colar jobs are
disappearing as AI starts to bite. The Wall Street Journal. https://www.wsj.com/economy/jobs/white-collar-jobs-ai-324b749c
Gallego, R. (2025). S.2495 - Keep call centers in America act of 2025. Congress.Gov.
https://www.congress.gov/bill/119th-congress/senate-bill/2495/text
Gartner. (2025). Gartner predicts none of the fortune 500 companies will have fully
eliminated human customer service by 2028. Gartner. https://www.gartner.com/en/newsroom/press-releases/2025-09-10-gartner-predicts-none-of-the-fortune-500-companies-will-have-fully-eliminated-human-customer-service-by-2028
GOOG. (n.d.). Definity: Modernizing call center experiences with AI. Google.
https://cloud.google.com/customers/definity
Guay, M. (2025). I tested 9 AI-powered scheduling assistants. My favorite is the one with the least AI. The
New York Times. https://www.nytimes.com/wirecutter/reviews/best-ai-scheduling-apps/
IBM. (n.d.). Driving a reimagined customer experience with an AI-powered virtual
assistant. IBM. https://www.ibm.com/case-studies/camping-world
John S., & Johnson, A. (2023). How to contact Facebook support and get help for
issues with your account. Business Insider. https://www.businessinsider.com/guides/tech/how-to-contact-facebook-problems-with-account-other-issues
Klarna. (2024a). Klarna AI assistant handles two-thirds of customer service chats in its
first month. Klarna. https://www.klarna.com/international/press/klarna-ai-assistant-handles-two-thirds-of-customer-service-chats-in-its-first-month/
Klarna. (2024b). 90% of Klarna staff are using AI daily - game changer for productivity.
Leggett, K. (2019). Your customers want to self-serve – It’s good for them and good for
you. Forrester. https://www.forrester.com/blogs/your-customers-want-to-self-serve-its-good-for-them-and-good-for-you/
Lindner, J. (2025). Digital Transformation in the CRM Industry Statistics. WorldMetrics.
https://worldmetrics.org/digital-transformation-in-the-crm-industry-statistics/
McK. (n.d.). Empowering telecom employees with personalized AI-powered training and
coaching tools. McKinsey & Company. https://www.mckinsey.com/industries/technology-media-and-telecommunications/how-we-help-clients/empowering-telecom-employees-with-personalized-ai-powered-training-and-coaching-tools
Merritt, A. (2021). Best practices in customer service automation. AWS.
https://aws.amazon.com/blogs/machine-learning/best-practices-in-customer-service-automation/
Morabito, C. [CNBC]. (2025, Nov. 4 ). The truth about AI and mass layoffs [Video].
Youtube. https://www.youtube.com/watch?v=8g5img1hTes
Morrell, S. (2025a). The great contact center standoff. Contact Center Pipeline.
https://www.contactcenterpipeline.com/Article/the-great-contact-center-standoff
Morrell, S. (2025b). From hearing to listening. Contact Center Pipeline.
https://www.contactcenterpipeline.com/Article/from-hearing-to-listening
Morris, T. (2016). 60% of consumers have stopped doing business with a brand due to
poor customer service. Microsoft. https://www.microsoft.com/en-us/dynamics-365/blog/business-leader/2016/07/22/4-frustrating-customer-service-experiences-and-how-to-fix-them/
Nicastro, D. (2025). AI-powered contact center as a service vendors battle for market
share in 2025. CMSWire. https://www.cmswire.com/contact-center/ai-powered-contact-center-as-a-service-vendors-battle-for-market-share-in-2025/
Piccolo, V. (n.d.). Call center outsourcing pricing in 2025. Callin.
https://callin.io/call-center-outsourcing-pricing/
Stewart, E. (2023). The death of the customer service hotline. Vox.
https://www.vox.com/23571375/no-call-center-phone-number-frontier-facebook
SQSP. (2024). Why we don’t offer phone support. Squarespace.
Uber. (n.d.). Support resources for Uber riders. Uber. https://help.uber.com/en/riders
Zdatny, I. & Brown, L. (2024). Contact center trends 2025: The three big shifts shaping
customer experience. Qualtrics. https://www.qualtrics.com/articles/customer-experience/contact-center-trends/