Analysis

Exploring the gap between AI’s “theoretical capability” and its real-world performance.

AI has quickly become one of the most discussed technologies in everyday life. It shows up in conversations with coworkers, friends, and family. You overhear people discussing it while eating out, and it fills social media feeds with predictions about how it will reshape work and society. Much of the conversation centers on whether AI will replace human jobs, but my experience using these tools has led me to see the situation somewhat differently.

In my own work with AI systems, I’ve found them to be incredibly useful in certain areas. They excel at brainstorming ideas, expanding knowledge, summarizing large amounts of information, and organizing complex topics into more understandable formats. Used well, these capabilities can save a significant amount of time and allow people to process information much more efficiently. In that sense, AI clearly has the ability to expand human productivity and make many tasks easier.

At the same time, regular use of these tools also makes their limitations visible. One example that stands out is computation. AI systems may correctly identify the formula needed to solve a problem but still produce an incorrect result when executing the math itself. In situations where precision matters, that creates a clear need for verification. While highly specialized systems designed only for a narrow type of calculation could address this issue, doing so sacrifices the flexibility that makes general AI tools appealing in the first place.

Another limitation appears in the way AI approaches problem solving. AI often presents the first reasonable solution it generates and then continues expanding on that approach. Humans, by contrast, frequently question whether an entirely different method might produce a better result. Much of human progress comes from this instinct to challenge existing systems and reinvent processes, even when current solutions appear to work well. AI systems are trained on existing knowledge and historical patterns, but they do not independently generate the motivation to push beyond those patterns unless they are explicitly guided to do so.

Recently I encountered a visual chart circulating online that attempts to map which occupational categories AI could theoretically cover. The chart separates “theoretical capability” from “observed usage,” suggesting that AI already has the potential to perform many tasks in fields such as management, business and finance, computer and mathematical work, architecture and engineering, legal services, arts and media, and office administration. In practice, the observed use of AI appears to mirror these categories, but at a smaller scale.

Discussions surrounding the chart often emphasize that AI will not necessarily replace entire jobs, but instead automate specific tasks within them. Some commentators suggest that companies able to automate 20 to 40 percent of knowledge work will significantly outperform organizations that do not adopt these tools. Others argue that the real shift will occur as workers learn how to direct, audit, and integrate AI systems into existing workflows.

There is likely truth in parts of this perspective. AI clearly has the ability to assist with many tasks inside existing roles. However, interpreting this as evidence that AI will broadly replace workers oversimplifies how organizations and systems actually evolve.

From my perspective, the chart itself also overlooks something important. If anything, AI’s potential usefulness may be more broadly distributed across occupational categories than the model suggests. Fields such as management, architecture and engineering, life and social sciences, legal work, education, personal care, and office administration all share characteristics that make AI particularly helpful as a supporting tool. These areas often involve large quantities of information, historical knowledge, and complex interpretation, which are environments where AI’s ability to synthesize information can provide real advantages.

Office and administrative work provides a good example. Many small repetitive tasks within these roles can be automated, allowing workflows to move faster and reducing time spent on routine work. At the same time, automation tends to push human involvement toward the more complicated or unusual situations that fall outside predictable patterns. Rather than removing people from the process entirely, AI often shifts their role toward solving problems that require judgment and creativity.

Legal work presents a similar dynamic. The field contains enormous volumes of information, and AI tools can dramatically accelerate research, document review, and information retrieval. Yet small nuances in wording can determine the outcome of a case, meaning human interpretation remains essential. Education follows a related pattern as well. Many foundational tasks help people build expertise over time, and removing too many of these steps could ultimately weaken how professionals develop their skills.

One of the most important differences between humans and AI systems comes from experience. People move between organizations, industries, and social environments. Along the way they encounter ideas that have nothing to do with their immediate work but later become the source of an important breakthrough. Sometimes the insight that saves a company or creates an entirely new opportunity comes from something learned in a completely unrelated context.

AI systems, by comparison, often operate within more confined boundaries. A system trained heavily on a single organization’s data may become extremely efficient at performing tasks within that environment, but it also risks becoming limited by it. Humans introduce the unexpected connections that allow systems to adapt and evolve over time.

Because of this, I see AI not as a replacement for human workers, but as a powerful tool that works best when paired with human judgment. AI can process information, summarize knowledge, and automate repetitive digital tasks at remarkable speed. Humans contribute creativity, curiosity, and the ability to question existing approaches. Together, those capabilities can strengthen organizations far more than either could alone.

It is possible that some companies will attempt to remove large numbers of roles in pursuit of automation. If that happens, they may discover that systems relying too heavily on automation risk becoming stagnant over time. Growth often depends on experimentation, unexpected insights, and new ways of thinking that emerge from human interaction.

AI will almost certainly reshape many aspects of work, but the most meaningful changes are likely to come from how humans learn to use these systems as tools rather than substitutes. Machines can accelerate processes, but the curiosity and imagination that drive progress still belong to people.