When airline schedules need rescheduling, food orders arrive damaged, or online purchases go missing, modern consumers increasingly turn to digital customer service channels hoping for swift resolution. Yet across Malaysia's digital landscape, a troubling pattern has emerged: AI-powered chatbots designed to handle these everyday disputes are instead funnelling frustrated customers into what industry observers call 'doom loops' – endless cycles of automated responses that neither solve problems nor provide human assistance.
The Malaysia Cyber Consumer Association has documented a sharp rise in complaints about inadequate automated customer support systems in recent years, with president Siraj Jalil identifying a specific mechanism at the heart of the problem: the 'infinite loop' phenomenon. Many chatbots operate within rigid parameters, recognising only predetermined keywords and unable to navigate the contextual nuances of non-standard issues. When customers present complex or unusual problems, these systems default to recycling the same FAQ links repeatedly, leaving consumers feeling abandoned within a frustrating cycle that offers neither resolution nor escape.
This design failure stems from a fundamental misalignment in corporate priorities, according to Henrick Choo, managing director of IT services firm NTT Data Malaysia. Companies have optimised their chatbot deployments around a troubling metric: how many customer interactions can be deflected away from human agents, rather than how many problems can actually be solved. Under cost pressures facing Malaysian enterprises, this approach has become disturbingly common, privileging bean-counting efficiency over customer satisfaction. The irony, Choo notes, is that this strategy typically backfires spectacularly, generating more frustration, repeat contacts, and reputational damage than the savings justify.
Research from Johns Hopkins University in the United States illuminates why customers instinctively distrust these gatekeeping systems. Associate Professor Evgeny Kagan and colleagues identified what they term 'gatekeeper aversion' – a deep-seated consumer resistance to interacting with chatbots perceived as barriers rather than helpers. Users arrive with heightened expectations of failure and a strong desire to bypass automation entirely. This suspicion deepens substantially when chatbots lack a straightforward pathway to human assistance, and intensifies further when customers eventually connect with a live agent only to discover that their entire conversation history vanishes, forcing them to restart their explanation from the beginning.
Siraj emphasises how this contextual blindness – the system's complete erasure of conversation records after disconnections or technical glitches – represents one of consumers' most acute frustrations. Users describe the experience as exhausting and disrespectful, particularly when reconnecting with human representatives who greet them with generic openers like 'How can I help you today?' without any awareness of the interaction that preceded them. Should the live chat subsequently disconnect, customers face the prospect of queuing again and repeating their entire narrative once more. This repetitive trauma transforms what should be a straightforward service interaction into an ordeal.
The underlying cause extends beyond chatbot limitations alone. Choo identifies systemic integration failures as the primary culprit: many organisations connect their chatbots exclusively to knowledge bases and FAQ repositories while leaving them disconnected from the actual operational systems where real work occurs. True problem resolution demands access to customer relationship management platforms, billing systems, identity verification protocols, approval workflows, and compliance frameworks – the same arsenal available to human agents. When chatbots lack these connections, they become answer-retrieval machines incapable of taking meaningful action. Customers quickly perceive this incapacity and lose faith in the entire system.
Among the most overlooked design failures, Choo points to the absence of permissions and tools that would empower AI systems to actually execute solutions. Retrieving an FAQ page requires minimal capability; resolving a billing dispute, account problem, or service issue demands integration depth that many organisations have failed to architect. The integration gap represents not a limitation of artificial intelligence technology itself, but rather a failure of organisational experience design. Companies seeking quick cost reductions have bypassed the harder work of truly integrating their automation with their operational backbone.
Khalil Nooh, CEO of local language model specialist Mesolitica, identifies a parallel and equally consequential failure: the assumption that organisations can simply upload their entire document repositories into large language models and expect flawless performance. In reality, most knowledge bases suffer from what Nooh terms 'knowledge-base rot' – obsolete pricing information, conflicting policies, expired terms, and outdated procedures that have accumulated over years or decades. When retrieval precision collapses under this weight of bad data, the AI systems compensate through what Nooh describes as 'hallucination,' confidently providing fabricated information that compounds customer frustration and erodes trust further.
For Malaysian companies navigating this landscape, the implications extend beyond individual customer irritation. The chatbot phenomenon reflects a broader strategic miscalculation: the conviction that automation can wholesale replace human judgment, relationship-building, and contextual problem-solving. Some organisations have pursued customer support through chatbots with virtually no human backup, eliminating the escalation pathways that should activate when automated systems reach their inevitable limits. This approach leaves no trained frontline staff capable of understanding complex system interactions or addressing escalated cases with the sophistication they demand.
The path toward genuine improvement requires fundamental reorientation. Organisations must reconceptualise chatbots not as cost-reduction mechanisms but as genuine problem-solving accelerators that capture information, reduce human agent workload through preliminary triage and information gathering, and then facilitate seamless handoffs to human representatives who inherit full contextual awareness. This demands integration with operational systems, permissions frameworks allowing meaningful action, and commitment to passing complete conversation histories forward. Technology enablement must serve customer experience rather than customer deterrence.
Choo emphasises that context represents the critical differentiator between genuine efficiency and manufactured frustration. When customers have already explained their issue to an AI system, subsequent human agents should immediately see full transcripts, customer profiles, previous transaction histories, sentiment analysis, and recommended next steps. Without this continuity, the system functions as a maze rather than a service delivery channel. Malaysian organisations operating under cost pressures must recognise that the savings generated by deflecting customers away from agents pale against the costs imposed through damaged relationships, repeat contacts, and diminished brand reputation.
The ultimate lesson transcends technology itself. AI chatbots can enhance customer service when designed with genuine resolution as the objective, integrated thoroughly with operational systems, and positioned as stepping stones toward human assistance rather than substitutes for it. The doom loops afflicting Malaysian consumers reflect not the limitations of artificial intelligence but rather failures of strategic vision and experience design. Companies that treat automation as genuinely complementary to human service rather than as a replacement mechanism will differentiate themselves through superior customer outcomes, while those pursuing purely extractive cost reduction will continue breeding the frustration that currently characterises Malaysian consumer sentiment toward these systems.
