Seven in ten Asia Pacific consumers surveyed by Twilio said they often have to start new AI customer-service conversations from scratch, despite 84 percent of APAC businesses believing their AI agents recognize returning customers and their history.

The findings are part of Twilio’s 2026 Customer Insights Series, which surveyed 7,652 consumers and 660 business leaders across 18 countries globally, including Australia, Hong Kong, India, Indonesia, Japan, Malaysia, the Philippines and Singapore.

Twilio describes the mismatch as “AI amnesia,” where systems fail to retain or retrieve enough context for a customer to continue an interaction without repeating information.

Context failures are affecting customer behavior

Seventy percent of APAC consumers in the study said they had abandoned an AI-powered customer-service interaction because it lacked basic information about them or previous interactions.

Five percent said they ended their relationship with the brand entirely after such an experience.

Handoffs to human agents also remain a weak point. Sixty-five percent of APAC consumers said they still had to repeat themselves or fill information gaps after an AI bot transferred them to a human agent.

The findings suggest that the practical value of customer-service AI depends not only on model quality but also on how well customer identity, conversation history and enterprise data move across channels and systems.

Transparency remains uneven

Seventy percent of APAC consumers said AI agents should identify themselves at the start of every conversation, while Twilio found only 22 percent of APAC businesses do so.

The survey also found that 68 percent of APAC consumers believe AI bots have improved over the past year. Sixty-five percent said they want AI agents from different companies to communicate directly to resolve issues faster.

That creates a tension for businesses: customers appear open to more capable and interconnected AI systems, but they also expect clearer disclosure and continuity across interactions.

Memory becomes an infrastructure problem

For enterprises, improving AI memory is not simply a matter of increasing a model’s context window. Customer data may sit across CRM systems, support tools, transaction platforms and identity layers, each with its own access rules and retention requirements.

Connecting those systems can improve continuity but also increases privacy, security and governance requirements. Companies need to determine which information an AI agent should remember, how long it should retain it, which systems it can access and when a human should take over.

Twilio’s findings indicate that customer trust may depend less on whether a company uses AI and more on whether the resulting experience feels coherent, transparent and accountable.

More than half of Singapore consumers report feeling less tolerant when interacting with AI – Twilio