American Express card members overwhelmingly want the option to choose between an AI-powered voicebot and a human representative, according to research the company conducted in February 2026. The survey of 528 US consumer card members found 95 percent said having that choice is important - a finding that complicates the CX industry's push toward voicebots and agentic AI.
The research, conducted among members of the American Express Customer Research Community, focused on attitudes and behaviors related to AI in customer service interactions. For customer support teams, the message is direct: automation may be inevitable, but customers are not ready to give up human access.
Accuracy beats speed in AI service
The findings challenge a common assumption that customers want faster service above all else. When asked what creates a positive experience with an AI-powered voicebot, 67 percent of card members pointed to having their issue resolved accurately. Only 35 percent said fast resolutions drove a positive experience.
"Card members prioritize accuracy over speed," said Pam Hopkins, executive vice president, global consumer servicing and fulfillment at American Express. "This reinforces one of our core customer service beliefs: accuracy is what earns trust."
Hopkins said these findings now shape Amex's channel-of-choice servicing strategy. The company deploys AI where it adds value for card members while keeping human expertise available for "the moments that matter most."
"Our servicing approach is centered around helping card members get the right answer, not simply the fastest one," she said. "AI can help drive efficiencies, but accuracy and trust remain fundamental to how we think about service."
Staffing, compliance, and the EU AI Act
The EU AI Act added another layer of complexity in August 2026, when new transparency obligations took full legal effect. Organizations must now disclose how, where, and when customer interactions are handled by AI, from chatbots in service queues to decision-making algorithms operating behind the scenes.
Amex, which employs 25,000 colleagues across 25 countries, has rolled out journey changes to comply with the Act. But the company is also investing in workforce planning tools to match staffing with demand. A new automated schedule optimization system is being implemented across markets.
"By better matching staffing to customer demand, we can help customers quickly get the support they need while also delivering a stronger overall servicing experience," Hopkins said. "We use AI and other technologies to enhance, not replace, the role our servicing colleagues play."
Longer-term, Amex is building toward a connected servicing experience across phone, chat, and email. The goal is to let card members move between channels without repeating their needs. "We want every interaction to feel like a single continuous relationship," Hopkins said.
Customers accept agentic AI on their own terms
While businesses struggle to show measurable returns from agentic AI, customers are quietly adopting their own AI agents. The Amex research found 71 percent of card members are comfortable with an AI-powered assistant acting on their behalf - with their confirmation - to schedule meetings and appointments. Sixty-five percent said the same for hotel reservations.
Comfort with generative AI recommendations was even higher. Seventy-nine percent of card members said they were somewhat or very comfortable with generative AI analyzing their data to make personalized restaurant recommendations, and 73 percent said the same for hotel recommendations.
Hopkins said Amex is "actively building, testing and learning" with agentic AI, guided by customer feedback. "Not every interaction should have the same level of autonomy," she said. "Our focus is on investing responsibly in technology that enhances the customer experience, while maintaining the right level of human oversight."
Internally, AI and automation are being applied to routine, repeatable tasks including KYC updates and reviewing certain card application information. Amex has also embedded generative AI into its servicing portal to summarize calls, identify customer issues, and surface relevant information in real time. For support professionals looking to build skills in this area, AI for Customer Support training resources cover the practical applications of these tools.
Why this matters for customer support teams
The Amex data offers a practical benchmark for support leaders: customers will tolerate AI, but they will not accept it as a replacement for human judgment. The 95 percent figure on wanting human access is not a rejection of automation - it is a demand for a functioning escalation path. Teams deploying voicebots or agentic AI should measure accuracy rates before touting speed gains, because customers in this study ranked accuracy nearly twice as high as fast resolution. For supervisors managing these transitions, an AI Learning Path for Call Center Supervisors can help build the operational skills required to balance automation with human staffing. The core lesson from Amex: prove the technology resolves issues correctly first, and customer trust will follow.
Your membership also unlocks:
