Companies are replacing traditional phone calls with AI
Between 2024 and 2025, many companies invested in fully automating their customer service, promising to reduce costs and operate faster. But in practice, this strategy began to have the opposite effect: frustrated customers and an increase in user churn, as indicated by reports such as the one from Quadrics, where one in five consumers who have used AI-powered customer service said they gained no benefit from the experience. On the corporate side, there’s the case of Klarna, a banking company that presented itself as AI-centric, but whose CEO had to inform Bloomberg that it was rehiring human staff after its AI-based approach resulted in “lower quality.”
This is what sources like Knowmax define as the AI tax, that is, the extra cost that companies have to pay due to customer frustration with an unsatisfactory customer experience (CX) caused by an artificial intelligence that struggles to solve their problems or needs.
Today, for many consumers, interacting with customer support has become exhausting. Bots that don’t understand the context, circular responses, and processes that never reach a solution end up eroding the relationship with the brand, which in the CX field is known as AI fatigue. And when a customer feels that no one is truly listening to them, trust quickly disappears.
The impact is already reflected in the metrics: bad CX experiences have led to nearly 73% of consumers abandoning a brand permanently, according to Zendesk statistics.
What went wrong?
Operations and customer experience teams in North America are seeing a clear trend: relying solely on automation, far from being the promised solution in savings and speed, is affecting customer satisfaction.
As we saw in the initial examples, many AI implementations aren’t generating the expected return because there’s a disconnect between the automated experience and human interaction. This makes the customer feel like they’re dealing with two different companies: one cold and automated, and another that arrives too late.
That’s why, as we’ve mentioned on our blog since the beginning of the automation craze, the change isn’t about abandoning technology, but about using it better.
One solution we’ve suggested on the CallFasst blog is to adopt hybrid models where AI streamlines repetitive tasks—such as gathering information, classifying requests, or reducing wait times—while human agents step in when judgment, empathy, and real problem-solving skills are needed.
However, this time we’ve added a key factor: the difference lies in the process design. When the bot can’t help, the transition to a human should be quick and seamless. In other words, technology speeds up the operation, while people build the experience.
Human factor when it really matters
It’s in moments of greatest frustration (when the customer has lost patience and needs a real solution) that the human element makes all the difference. We understand that our agents are not replaceable by automation, but rather complement it. They are trained to handle complex conversations, resolve conflicts, and connect with people through empathy and emotional intelligence.
However, for decision-makers, we also don’t believe the solution involves going back to the past and banishing AI. In this regard, there’s a concept called Human-in-the-Loop, where responsibility and crucial decision-making are delegated to humans instead of letting AI make them all on its own. While AI handles transactional processes and routine tasks, teams of advisors focus on what can’t be resolved with predefined responses: sensitive situations, disgruntled clients, and decisions that require human context.
When the algorithm costs more than talent
While the primary motivation for replacing humans with bots has always been efficiency and budget cuts, the stark reality of 2026 has proven the exact opposite. According to Gartner projections , full automation will be prohibitively expensive for most companies, estimating that the cost per resolution using Generative Artificial Intelligence will exceed $3, a figure even higher than the average cost of many human customer service agents in BPO models. Even leaders in the tech industry are already warning about this financial imbalance. Bryan Catanzaro, vice president of Applied Deep Learning Research at NVIDIA, recently admitted in an interview with Axios that, for his team, “the cost of computing is far beyond the cost of employees.”
Leading companies in North America are recognizing that adopting models that include advisors—like those offered by CallFasst—is the only financially sustainable and secure strategy. This way, they avoid the costly expense of excessive IT infrastructure and the massive churn of frustrated customers, transforming their customer service, collections, technical support, or telemarketing into an inexhaustible engine of loyalty, trust, and long-term profitability.