Why Accuracy is the #1 Thing Consumers Demand from AI Customer Service Agents

Key Findings from Decagon's Survey

A recent survey by Decagon, a conversational AI platform provider, reveals critical insights into consumer preferences for AI in customer service. The study, conducted by The Harris Poll in July 2025 among 2,077 U.S. adults, highlights age-based differences and universal demands.

Age-Based Preferences

• Younger consumers (aged 18-44) are more likely to view brands using AI agents as modern and innovative. They are also more inclined to pay a premium for exceptional customer service and to stop buying from a brand after just one negative support experience.
• Younger respondents prefer using AI agents over older ones, citing reasons like 24/7 availability, faster resolution, ease of use, and higher accuracy.

Universal Demand for Accuracy

• 84% of all respondents stated that receiving knowledgeable and accurate support from AI agents is very important or absolutely essential.
• When asked to choose the most important factor, 40% selected 'knowledgeable and accurate answers', with 24/7 availability coming in second at 16%.
• Jesse Zhang, co-founder and CEO of Decagon, noted that age differences in these results were minimal, leading to aggregated reporting for clarity.

Implications of Inaccuracy

• Inaccuracy in AI-led support is a key reason consumers disengage from brands. Robin Gareiss, CEO of Metrigy, suggests that poor bot training leads to negative experiences, with consumers leaving after an average of 2.4 bad interactions.
• A June 2025 Amdocs Studio report supports this, finding that 54% of consumers will abandon a brand after four or fewer bad experiences.
• Alarmingly, 60% of younger consumers and 56% of older consumers would stop buying from a brand after just one negative support experience, indicating a sharp increase in intolerance compared to previous years.

AI and the Next Generation of Customer Experiences, August 2025

Survey Methodology

• The online survey was conducted in the U.S. from July 29–31, 2025, with data segmented into age groups (18–24, 25–34, etc.) but combined into 18–44 and 45+ for clearer trend analysis. This grouping was based on a noticeable shift in attitudes around age 44.