When a flight gets rescheduled, a food order arrives damaged, or an online purchase goes missing, customers expect swift resolution through modern customer service channels. Yet across Malaysia and the region, a growing number of people are discovering that contacting support via AI-powered chatbots often leads nowhere—trapping them in what experts call "doom loops" of repetitive, unhelpful responses that ultimately damage company reputations rather than solve problems.
The frustration is palpable online. Social media platforms from X to Reddit overflow with accounts of customers battling unresponsive bots, unable to escape automated systems or reach a human operator. What binds these seemingly disparate complaints is a systemic design flaw: chatbots engineered to deflect inquiries rather than resolve them. The Malaysia Cyber Consumer Association has documented a sharp rise in such grievances, with president Siraj Jalil pointing to a specific culprit—the "infinite loop" phenomenon. These systems are programmed to recognise only predetermined keywords, and when confronted with nuanced or non-standard problems, they robotically regurgitate links to FAQ pages regardless of relevance.
This represents a fundamental misalignment of corporate objectives with customer expectations. Henrick Choo, managing director at IT services firm NTT Data Malaysia, identifies the core problem: companies have optimised their chatbot metrics for the wrong outcome. Rather than measuring success by problems solved, many organisations track how many customers they keep away from human agents. "The metric became 'how many customers did we keep away from agents?' instead of 'how many issues did we resolve?'" he explains, noting that Malaysian firms operating under tight budget constraints are particularly vulnerable to this cost-driven logic. While reducing labour expenses seems rational on a spreadsheet, the strategy backfires spectacularly—generating more frustrated repeat contacts, complaints, and lasting reputational harm.
Academic research validates these consumer complaints. A John Hopkins University study on AI chatbots identified a psychological phenomenon called "gatekeeper aversion"—users immediately perceive chatbots as obstacles rather than helpers, especially when they cannot easily bypass them to speak with a person. Customers sense the deception inherent in systems designed primarily to block them rather than assist them, creating an adversarial dynamic from the first interaction.
The frustration multiplies when customers finally breach the automated wall and reach a human agent, only to discover that their entire conversation history has vanished. Many companies fail to implement systems where chatbot transcripts automatically transfer to human representatives. Siraj describes this as a form of contextual blindness that violates basic customer respect—forcing people to re-explain complex issues from scratch, sometimes repeatedly if disconnections occur. What should be a seamless handoff becomes an additional ordeal, precisely when customer patience is already exhausted.
Choo emphasises that the handoff moment is where companies consistently lose consumer trust. Customers are often willing to experiment with self-service automation, but their tolerance evaporates when trapped in what amounts to an automated maze with no visible exit. The solution requires radical transparency: when escalation occurs, human agents must inherit complete context—full chat transcripts, customer profiles, transaction history, sentiment analysis, and recommended next steps. Context, he argues, is the crucial distinction between efficiency and exasperation.
The underlying problem extends beyond chatbot design into the broader systems architecture that supposedly supports these tools. Many organisations connect their chatbots only to knowledge bases while leaving them disconnected from the actual systems where work happens—customer relationship management platforms, billing systems, identity verification tools, approval workflows, and compliance frameworks. A chatbot can retrieve information easily, but resolving substantive issues demands deep integration with systems of record. This integration gap means the bot can answer questions but cannot act, leaving customers stranded when problems require genuine intervention rather than information retrieval.
Local language model specialist Khalil Nooh, CEO of Mesolitica, highlights another widespread failure: organisations dumping outdated or incomplete documentation into large language models and expecting perfect results. Many knowledge bases suffer from what he terms "knowledge-base rot"—obsolete pricing information, conflicting policies, expired terms, and stale procedures. When retrieval systems operate on corrupted data, artificial intelligence amplifies the problem through hallucination, generating plausible-sounding but entirely false responses that further frustrate customers.
Nooh also warns against the misconception that AI chatbots should replace human customer support entirely. Some companies implement these systems without establishing proper escalation pathways or maintaining adequate human staff trained on underlying systems and procedures. This reflects a failure of imagination—treating customer support as a cost centre to be minimised rather than a revenue-generating function that builds loyalty and brand equity.
For Malaysian consumers and businesses, the implications are significant. As companies across Southeast Asia rush to deploy AI solutions, many are committing the same design errors documented by researchers and practitioners. The competitive advantage will accrue not to organisations that most aggressively automate customer interactions, but to those that thoughtfully integrate automation with human expertise. The goal should be empowering customers with instant answers for simple queries while guaranteeing frictionless escalation to knowledgeable humans for complex problems—and ensuring those humans possess complete information about preceding interactions.
The current trajectory is unsustainable. Cost savings achieved through aggressive automation are being overwhelmed by customer frustration, repeat contacts, churn, and negative word-of-mouth in an increasingly connected region. Malaysian companies competing in regional markets must recognise that customer service is not an operational cost to be minimised but a strategic asset. Those that invest in thoughtful integration—bridging automation and human intelligence, ensuring data flow between systems, and designing explicitly for successful escalation—will emerge as market leaders. Those that continue deploying chatbots as defensive gatekeepers will find themselves trapped alongside their customers in those very doom loops they sought to create.
