AI: The Power of Assistance Over Perfection

Published in May, 2025

AI is a revolution. But its true strength, in my opinion, lies in domains where perfection isn’t the goal, where a margin of error is acceptable, and the aim is to assist and guide rather than fully replace humans. In these areas, AI truly shines as a collaborative tool, a super charger, amplifying human capabilities rather than striving for flawless autonomy.

Take self-driving cars as a case study. In 2005, "experts" predicted we’d have fully autonomous vehicles by 2010s. Yet, here we are in 2025, and while progress is impressive, we still don’t have cars that can drive themselves 100% of the time without human intervention, legally. Maybe we are at 99.99%, but still not good enough! Why? Because self-driving is a “100% application.” Even a single error can have catastrophic consequences, so the bar for safety and reliability is extremely high. This pursuit of perfection has slowed progress, as the technology must account for every edge case on the road.

Contrast this with coding, a domain where AI excels precisely because 100% accuracy isn’t required to boost a software developer's productivity. If AI can handle 50-80% of your coding tasks, writing initial drafts, generating tests, or suggesting optimizations, it frees you to focus on higher-level decisions like tech stack, architecture, best practices, debugging, and fine-tuning. The margin for error allows developers to iterate and refine AI-generated code, making the process faster and more efficient. And while AI isn't going to replace human developers, teams will, in the very near future, require siginicantly less developers.

Within coding, AI’s value is even clearer in less critical tasks, like building isolated UI (User Interface) screens. A button slightly misaligned or a color scheme that needs tweaking? These are low-stakes non-breaking issues where AI can deliver functional drafts, and human oversight can polish the result. Compare that to, say, an automation backend process like a database trigger, where a single mistake could cascade into major failures. In these high-stakes scenarios, precision is non-negotiable, and AI’s role is better suited to support rather than lead.

I’ve been experimenting with AI-powered code assistants and agents, particularly in the Salesforce ecosystem, and their impact on developing rich, customized Lightning Web Components (LWC) for isolated UIs is remarkable. These tools outshine traditional low-code solutions like Salesforce Screen Flows by offering unparalleled flexibility in customization. Historically, developers shied away from LWCs due to their “code-heavy” nature, requiring more time to build, test, and maintain, while Screen Flows provided a faster, low-code alternative. But AI changes the game. With AI capable of generating advanced, fully customized LWC UIs at near 95% completion (from my experience), the need for restrictive low-code builders like Screen Flows diminishes. Why settle for limited "templates" when AI empowers you to create tailored, high-quality interfaces with minimal effort?

In conclusion: So far, AI’s sweet spot is in assisting, reducing the need for human labor, but not replacing (not yet!). It’s a partner and advisor that handles the heavy lifting in domains where “good enough” unlocks creativity and productivity, leaving humans to specialize and focus on specific areas.