Half of the companies planning to replace customer service with AI are reversing their decision
Half of surveyed companies revert from AI plans in customer service, citing challenges.

Many organizations that once aimed to replace customer service staff entirely with AI are now reversing course. According to a recent Gartner survey, 50% of companies that had planned to automate customer service with AI expect to abandon those plans by 2027. Early ambitions around AI-driven support centers proved overly optimistic, often clashing with real-world complexities, unexpected costs, and underwhelming customer experiences.
Today, 95% of customer service leaders indicate they will retain human agents alongside AI. Rather than full automation, companies are embracing hybrid models where AI assists but does not fully replace human interaction. Analysts emphasize that AI still lacks the emotional intelligence and contextual understanding required in sensitive or complex situations.
Customer trust plays a key role in this shift. Over half of consumers say they trust humans more than AI when resolving issues, while only 7% say they trust AI more. Additionally, the promised cost savings of AI often fall short once integration, training, and maintenance expenses are considered, further dampening enthusiasm for complete automation.
Some companies have already reversed course. Klarna, for example, replaced hundreds of support roles with AI but later brought back human staff after quality declined and complaints rose. The company’s leadership admitted that over-reliance on AI had negatively impacted customer satisfaction and acknowledged the need to restore human-led support.
The current industry consensus favors a “digital-first, not digital-only” approach. AI can effectively handle simple or repetitive queries, but must work in tandem with skilled human agents who step in when empathy, creativity, or deeper insight is needed. In this new model, AI enhances productivity rather than acting as a full replacement.
Sources: TechSpot, The Register, TechRadar, Economic Times