When a flight gets rescheduled, a food delivery arrives damaged, or an online purchase goes missing, customers expect swift resolution. Instead, many are encountering the same dead end: AI-powered chatbots that repeat the same unhelpful responses before eventually abandoning them back to a confused human agent who knows nothing of their problem. This pattern has become so common that frustrated users across Malaysia and beyond regularly take to social media to share their experiences, painting a bleak picture of how automation is eroding customer service quality rather than enhancing it.
The Malaysia Cyber Consumer Association has documented a significant surge in complaints about automated customer support systems in recent years. According to MCCA president Siraj Jalil, the core issue centres on what experts call the "infinite loop" phenomenon—a technological trap where chatbots programmed to recognise only specific keywords repeatedly direct customers to the same FAQ pages when confronted with problems that fall outside their narrow parameters. When a customer's issue involves any degree of nuance or falls outside standard categories, the bot becomes useless, cycling endlessly through identical suggestions while offering no pathway to human assistance.
Henrick Choo, managing director of IT services firm NTT Data Malaysia, identifies the root cause: many companies deploy chatbots not to solve problems but to keep customers away from expensive human agents. The metric that drives these implementations has become backwards, he argues. Rather than measuring success by problems resolved, companies measure it by how many customer interactions they prevent from reaching live staff. This cost-cutting mindset, particularly common among Malaysian firms operating under tight budgetary constraints, creates the opposite of its intended outcome. Instead of reducing expenses through deflection, companies generate higher frustration levels, more repeat contacts, additional complaints, and ultimately damage to their reputation.
Research from Johns Hopkins University in the United States has documented this phenomenon as "gatekeeper aversion," a deeply ingrained customer psychology where people instinctively distrust automated systems positioned as barriers rather than helpers. In experimental settings, researchers including Associate Professor Evgeny Kagan found that users immediately perceive the risk of chatbot failure as high and actively resist engaging with them. This resistance intensifies dramatically when the chatbot design offers no immediate escape route to a human agent, forcing customers to exhaust all automated options before gaining access to actual help. Customers sense immediately whether the bot exists to assist them or block them, and this perception shapes their entire interaction experience.
The frustration multiplies when customers finally connect with a human agent, only to discover that their entire conversation history has vanished. Many systems lack any mechanism to transfer context from the chatbot to the human representative, forcing customers to recount their entire situation from scratch. Siraj explains that consumers describe this experience as exhausting and disrespectful of their time. What begins as a simple complaint compounds into a sequence of repetitive explanations across multiple interactions. If the connection drops and the customer is returned to the queue, they may be forced to repeat the entire process again, compounding their frustration exponentially.
Choo emphasises that the handoff between automated and human support represents precisely where most companies lose customer trust. The solution, he stresses, lies not in abandoning AI but in fundamentally redesigning how these systems integrate with human-centred support. When a customer reaches a live representative, that agent should have access to the complete conversation transcript, the customer's profile, their transaction history, emotional sentiment analysis, and recommended next steps. This contextual awareness transforms the experience from frustrating repetition into genuine problem-solving. Without this integration, even well-intentioned human agents cannot deliver the seamless service customers expect.
The underlying problem extends beyond the chatbot interface itself. Choo identifies systemic design failures in the infrastructure supporting these systems, including data architecture, escalation protocols, and integration depth. Many companies make the fundamental mistake of deploying chatbots capable only of retrieving information from knowledge bases without granting them access to the actual systems where problems get resolved. A chatbot might easily pull an FAQ answer, but resolving a billing issue requires access to customer relationship management systems, billing databases, identity verification tools, approval workflows, and compliance frameworks. If the AI lacks integration with these systems of record, it cannot take meaningful action—it can only provide information, which frequently proves insufficient.
Khalil Nooh, CEO and co-founder of Malaysian language model firm Mesolitica, adds another critical dimension to this problem: the quality of the underlying data feeding these AI systems. Many companies assume they can simply upload all their internal documents into a large language model optimised for information retrieval, and the system will function perfectly. This fundamentally misunderstands how these technologies operate. Legacy knowledge bases often contain what Nooh calls "knowledge-base rot"—outdated pricing information, conflicting policies, expired terms, and obsolete procedures. When AI systems trained on corrupted data attempt to retrieve information, their precision collapses, and the models resort to hallucinating plausible-sounding but entirely fabricated answers. Customers then receive confidently delivered misinformation, creating secondary problems atop their original issues.
The strategic error underlying many of these failures stems from flawed assumptions about AI's proper role in customer support. Some organisations view chatbots as wholesale replacements for human customer service, without considering proper escalation pathways for genuinely difficult issues. This approach inevitably fails because no single system can handle the full spectrum of customer problems. Complex issues involving product defects, billing disputes, or service failures require human judgment, empathy, and access to decision-making authority that automated systems simply cannot replicate. The most successful implementations view AI as a triage tool that handles straightforward, frequently asked questions while seamlessly escalating complex cases to frontline human agents equipped with proper training and system access.
For Malaysian consumers and businesses, these failures carry significant implications. As competition intensifies and customer expectations rise, companies that fail to resolve complaints efficiently lose market share to competitors offering superior service. The reputational damage extends beyond individual dissatisfied customers to collective online discourse where negative experiences accumulate. Regulatory attention is also increasing, with consumer protection bodies worldwide beginning to examine whether automated gatekeeping mechanisms constitute deceptive practices. Malaysian firms that implement customer service AI without proper human integration and system access risk falling behind regional competitors while simultaneously exposing themselves to regulatory and reputational risks.
Addressing these issues requires fundamental changes in how companies approach customer service automation. Rather than viewing AI as a cost-reduction mechanism, forward-thinking organisations must design systems that genuinely empower both customers and agents. This means implementing chatbots that possess genuine decision-making authority rather than mere information retrieval, ensuring complete data integration across all customer-facing systems, maintaining high-quality knowledge bases free from corruption and obsolescence, and designing escalation pathways that smoothly transition customers to human agents while preserving all contextual information. When implemented thoughtfully, AI can genuinely enhance customer service. When deployed primarily as a deflection mechanism, it creates the "doom loops" that increasingly characterise Malaysian customer service experiences today.
The path forward requires Malaysian companies to recalibrate their success metrics entirely. Rather than measuring triumph by the volume of customer interactions prevented from reaching humans, companies should measure it by the proportion of issues actually resolved. This reorientation would naturally incentivise better AI design, more robust system integration, and faster escalation when customers genuinely need human assistance. As regional digital maturity increases and customer expectations evolve, companies that crack this code will gain substantial competitive advantages over those still trapped in the false economy of cost-cutting chatbot deployment.
