Something has gone terribly wrong with how Malaysian and regional businesses approach artificial intelligence in customer service. Whether you're rescheduling a flight, reporting a food delivery mishap, or chasing down a missing package, the experience has become disturbingly uniform: navigate an unresponsive chatbot, hit dead ends repeatedly, and desperately search for a human being who can actually help. The technology companies deployed to streamline support is instead creating what insiders call "doom loops"—cyclical traps that leave customers exhausted, mistrustful, and more frustrated than when they started.
The Malaysia Cyber Consumer Association has documented a troubling surge in complaints about customer service systems in recent years, according to MCCA president Siraj Jalil. The core issue isn't the AI itself, but how companies have engineered these systems to deflect problems rather than solve them. Chatbots are often hard-coded to recognise only specific keywords, which means when a customer presents a slightly unusual problem, the bot defaults to repeating the same FAQ links endlessly. Consumers find themselves trapped in what the industry calls an "infinite loop"—asking the same question repeatedly while the system provides identical, unhelpful responses. There's no exit strategy, no acknowledgement that the issue falls outside the bot's limited scope, just circular frustration.
The underlying psychology behind this design failure is revealing. NTT Data Malaysia managing director Henrick Choo explains that many companies have fundamentally miscalculated their priorities. Rather than measuring success by "how many issues did we resolve," companies now obsess over "how many customers did we keep away from agents?" This metric inversion, he argues, particularly affects Malaysian businesses operating under tight cost constraints. Yes, efficiency matters. But when AI becomes primarily a tool for deflection rather than resolution, it produces the opposite result: more repeat contacts, more complaints, and eventual reputational damage. Customers are instinctively aware of this dynamic. They sense immediately that the chatbot exists to block them, not serve them.
Academic research confirms this intuition. A Johns Hopkins University study on AI chatbots in customer service identified what researchers call "gatekeeper aversion." The theory is straightforward: chatbots function as the cheapest first-line responders, protecting the time of more expensive human agents. But in experiments, researchers including Associate Professor Evgeny Kagan found that this dynamic is remarkably persistent and difficult to overcome. From the moment users encounter a chatbot, they expect it to fail and they resist engaging. This resistance intensifies dramatically when the bot lacks a transparent, easy option to immediately escalate to a human agent.
What transforms mild frustration into genuine anger is the handoff failure. When customers finally break through and reach a live representative, they discover their entire conversation history has vanished. The human agent greets them with a cheerful "How can I help you today?"—forcing the customer to explain their entire grievance from scratch. This isn't mere inconvenience; it's a profound sign of disrespect for the customer's time and emotional investment in resolving the issue. When chat sessions disconnect, customers rejoin the queue and repeat the ordeal. Siraj describes this experience as psychologically draining, a violation of the implicit social contract between businesses and their customers.
Choo is blunt about where this system breaks: "The handoff is where many companies lose trust." Customers, he notes, are generally willing to try self-service systems. Their breaking point comes when they cannot easily exit the automated doom loop to speak with a human. The fundamental problem is contextual blindness. If a customer has already spent twenty minutes explaining their billing error to a chatbot, the human agent should inherit the full transcript, the customer's profile, transaction history, emotional sentiment detected in the conversation, and recommended next steps. Instead, the human agent operates as if the customer is calling for the first time. This represents a massive design failure, though Choo emphasises these are not limitations of artificial intelligence technology itself but rather failures in how companies have architected their customer experience systems.
Beyond the chatbot interface lie systemic integration problems. Many companies have connected their chatbots only to knowledge bases, not to the actual systems where real work happens. A bot can retrieve an FAQ, certainly. But resolving an account issue requires access to customer relationship management systems, billing databases, identity verification protocols, approval workflows, audit trails, and compliance frameworks. This gap—what Choo terms "integration depth"—represents the actual barrier to effective AI customer service. The AI cannot take action because it lacks permission to access the very systems where problems are solved. Companies have installed expensive technology that can provide information but cannot resolve anything.
Khalil Nooh, CEO and co-founder of local language model firm Mesolitica, identifies a separate but equally damaging problem: the databases themselves. Most organisational knowledge bases contain what he calls "knowledge-base rot"—obsolete pricing information, conflicting policies, expired terms and conditions scattered throughout legacy systems. When companies assume they can simply dump thousands of documents into a large language model and expect perfect performance, they're fundamentally misunderstanding how these systems work. The retrieval precision collapses. The model, lacking reliable information to work with, essentially hallucinates responses. A chatbot trained on outdated information is worse than having no chatbot at all.
The philosophical mistake underlying many implementations is even more fundamental. Some organisations operate under the delusion that AI chatbots should completely replace human customer support. They've reduced their frontline staffing without establishing robust escalation mechanisms, without maintaining human agents familiar with company systems, and without accepting that certain issues will inevitably require human judgment and contextual understanding. This represents not a technological failure but a strategic miscalculation about what customer service actually requires. A well-designed system would use AI to handle routine inquiries efficiently while maintaining clear pathways to capable humans for complex problems—with complete context transferred seamlessly.
For Malaysian businesses specifically, these design failures carry particular weight. The region's competitive e-commerce and logistics sectors depend heavily on customer trust and repeat business. When customers experience repeated frustration with chatbots that block rather than serve them, when they feel disrespected by systems that force them to re-explain problems, they vote with their wallets. They switch platforms. They post negative reviews across social media. They join the growing chorus on Reddit and X documenting their doom-loop experiences. The technology was meant to reduce costs, but poor implementation actually increases them through reputation damage, churn, and the irony of customers finally reaching human agents who must spend extended time resolving issues that proper automation could have handled from the start.
The path forward requires acknowledging that customer experience design is not incidental to AI deployment but central to it. Context matters more than cost savings. A chatbot that can smoothly transfer a customer's complete conversation history to a human agent, that has permissions to actually resolve common problems within integrated backend systems, and that operates on current, accurate information will reduce contact volume more effectively than one designed primarily to deflect. Companies must resist the temptation to measure AI success by how many customers it keeps away from agents. Instead, they should measure it by how many issues it resolves on first contact. Until Malaysian businesses recalibrate these metrics and invest in proper system integration and data governance, customers will continue to be trapped in doom loops—frustrated, angry, and increasingly unlikely to return.
