For more than eighteen years, Wei Khjan Chan has pursued a career as a professional accountant—a discipline frequently characterized as being on the front lines of potential automation. Throughout his nearly two-decade tenure, he has witnessed countless headlines cautioning that artificial intelligence could soon replace many of the responsibilities traditionally carried out by accountants. Each new warning served as a reminder of the transforming economic landscape, intensifying his sense of vulnerability while simultaneously provoking deeper reflection on his own professional resilience. As Chan himself explained, early exposure to artificial intelligence would have offered an opportunity not merely to preserve his role but to take proactive control of his own technological transformation. In his words, if he had familiarized himself with AI earlier, he could have been the one to supplant his position with an enhanced, AI-augmented version of himself, rather than wait for others—or new technologies—to render his role obsolete.
Recognizing the urgent need to stay ahead of such disruptions, Chan chose to immerse himself in what he describes as ‘vibe coding,’ a creative practice that blends coding intuition with the support of AI-driven development tools. Through this approach, he learned to generate code, design applications, and automate complex workflows despite never having received formal technical training. Now an audit partner at an established accounting and advisory firm in Malaysia, Chan first encountered the concept of vibe coding during a series of hands-on weekend coding workshops in both Singapore and Malaysia, which he attended out of curiosity and self-improvement. What began as a modest experiment evolved into a sincere dedication to learning how to harness machine intelligence for solving practical business problems.
With no prior experience in software engineering, Chan succeeded in building a web application designed to address one of his most persistent professional frustrations: the cumbersome, time-consuming process of managing and submitting expense claims after business trips. The solution he created employs AI-powered optical character recognition to automatically scan, interpret, and process printed receipts, organizing them into standardized digital records easily transferable to his company’s finance teams. In addition to automating expense submissions, he has expanded his use of AI to other elements of his accounting workflow, including the automatic generation of invoices and reports. Presenting the application that he developed, Chan remarked that the underlying code—composed largely in JavaScript—was completely unfamiliar to him before he began this journey. Without intuitive coding tools and an openness to learn, he admitted, a traditional accountant would have found such tasks nearly impossible to complete on their own.
In Chan’s view, artificial intelligence is not a threat that signals the decline of accounting; rather, it represents the means through which the profession can be reimagined and preserved for the future. He never intended his newfound coding skills to serve as a means of professional escape. Instead, he regards AI literacy as an indispensable competency—comparable to spreadsheet mastery—that every modern office worker must cultivate. His firsthand experience building AI-driven applications has underscored just how transformative these technologies are: tasks that once demanded weeks of coordinated effort across multiple team members can now be prototyped independently within the span of a weekend.
Motivated by these results, Chan has become an advocate for broader educational initiatives within his industry. As a committee member of his local accounting institute in Malaysia, he actively campaigns for the widespread inclusion of AI training in professional development programs. He observes with concern that, while demand for high-quality accounting services continues to rise, the number of aspiring accountants entering the field has noticeably declined. This imbalance between labor supply and market need, he argues, could be mitigated through AI, which allows limited human resources to accomplish far more in less time without sacrificing accuracy or oversight.
When reflecting on the lessons he has learned through vibe coding, Chan recalls that in the early stages of his experimentation, he was advised to provide the AI models with long, exhaustive prompts covering the entire context of the problem. Over time, however, real-world practice taught him the opposite: brevity and precision, applied iteratively, yield better and more manageable results. The key, he emphasized, lies in ensuring that the initial instruction is carefully constructed—effectively setting the framework within which all subsequent adjustments occur. Once the core prompt establishes the proper foundation, incremental fine-tuning of individual components leads to more coherent outcomes than overwhelming the system with numerous simultaneous requests.
He compares this process to supervising a junior intern. The principle is straightforward: by dividing a large, complex objective into discrete, clear, and well-defined subtasks, one greatly increases the likelihood of achieving satisfactory results. The more specific and methodical the direction, the better the AI’s performance—just as a precise manager brings out the best in a novice employee.
Not every project progressed smoothly. During one particular development effort, Chan inadvertently structured his database architecture around the needs of a single organization, neglecting to plan for multi-company compatibility. Later, when a user requested support for more than one firm, he discovered that the entire foundation of the system would need to be redesigned from scratch. This mistake, although frustrating, became a pivotal learning experience: he came to understand that getting the system’s architecture right from the outset is absolutely essential, since additional features and capabilities can always be added later, but a poorly conceived framework can undermine even the most promising project.
When it comes to debugging the programs he creates, Chan approaches the process with good humor. He jokes that debugging often feels like ‘complaining to the AI’—an iterative conversation in which each new error message signals incremental progress. If the nature of the message changes, he takes it as evidence that the AI system is actively working through the issue. However, if the same error persists repeatedly, he resets the dialogue and reformulates his instructions, using new examples or refined contexts to redirect the AI’s reasoning path.
Contrary to common stereotypes about coding, he insists that vibe coding does not demand endless nights of intensive labor or countless hours of frustration. Instead, coding has become a late-evening pastime for him, an engaging intellectual exercise he pursues after his children have gone to bed. During these quiet hours, he methodically adds new features, refines specific functions, or experiments with alternate interfaces. The process, he says, feels almost like playing a strategic game—one that rewards creativity and persistence over brute force. Each small improvement builds upon the last, and over time the collection of minor refinements coalesces into a cohesive, functional tool. With consistent curiosity and guidance from AI systems, his projects gradually evolve from simple experiments into tangible, value-generating assets.
Chan’s story illustrates a profound shift in how seasoned professionals can approach emerging technologies. By refusing to succumb to fear of obsolescence and instead using AI as a learning companion, he demonstrates that even in traditionally conservative fields such as accounting, there is space for innovation, reinvention, and renewed relevance in the age of automation.
Sourse: https://www.businessinsider.com/accountant-how-to-learn-vibe-coding-ai-tools-workshop-2025-10
