FEATURE – Between hype and skepticism, understanding AI’s true capabilities is essential for leaders seeking to applyit meaningfully within their organizations.

Words: Art Smalley

Every day, we encounter wildly different predictions about the impact of Artificial Intelligence: some believe it will eliminate huge numbers of jobs and reshape society in dramatic ways; others warn that the technology is overhyped, prone to hallucinations, and far less capable than the headlines suggest. Between these two extremes lies a confusing landscape for leaders trying to understand what AI actually means for their organizations.

To navigate this debate, it helps to step back and separate hype from reality. Artificial intelligence is not entirely new. In fact, many forms of what we might call “big AI” have been around for decades. Robotics, machine vision, automation, and expert systems have been used in manufacturing and engineering for more than 40 years. These technologies operate in the physical world (machines that move, grasp, inspect, and process real materials). You might think of this as the “torque side” of AI, as I call it, the domain where machines interact directly with the physical environment.

What caught the world by surprise in recent years was something different: the rapid emergence of large language models (LLMs). When tools like ChatGPT became widely available, millions of people suddenly experienced a new kind of artificial intelligence. Instead of controlling robots or factory equipment, these systems interacted through language, producing text, explanations, summaries, and code.

This distinction is important. Traditional AI dealt with physical actions and mechanical systems. Generative AI operates in a different world entirely: the world of words and language patterns. Understanding that difference helps explain both the excitement and the confusion surrounding modern AI.

THE HYPE VERSUS REALITY PROBLEM

Part of the challenge is that most people interact with AI only at its most basic level. For example, many users encounter large language models through simple chat interfaces. At this level, the results can feel unpredictable. Sometimes, the answers are impressive; other times they are incorrect, incomplete, or strangely off.

This inconsistency happens because language models don’t actually think. They generate responses by predicting the next most likely word based on patterns in the data they were trained on—primarily large portions of the internet. Researchers sometimes describe them as “stochastic parrots” that simulate reasoning, but don’t truly understand the world.

Yet despite this limitation, language models are surprisingly capable. When used properly, they can perform many tasks that once required significant human effort. The key is learning how to move beyond basic interactions.

In practice, there are several levels of AI usage. At the lowest level (the chatbot interface), almost everyone can participate. But as you move up the capability ladder, the number of people who can effectively use these tools drops dramatically.

The next level involves prompt engineering, where users carefully design instructions to guide the model’s responses. Above that lies API-level interaction, where developers connect directly to AI models using programming tools, such as Python. This allows them to integrate AI into applications and automate more complex workflows. Beyond that are advanced techniques like retrieval-augmented generation (RAG), which connects models to external data and fine-tuned systems designed for specific tasks.

While these advanced approaches can deliver impressive results, only a tiny percentage of users currently operate at these levels. Most people remain at the basic interaction stage. As a result, AI’s real capabilities often appear exaggerated.

WHERE AI EXCELS… AND WHERE IT STRUGGLES

To understand how leaders should approach AI, it helps to consider how humans and machines differ in their thinking processes. As Stanford researcher Balaji Srinivasan put it, humans are natural end-to-end thinkers: when faced with a problem, we can define the situation, develop hypotheses, execute solutions, and reflect on the results. In Lean Thinking, this cycle is often described as PDCA (Plan, Do, Check, Act). Language models operate differently: they are strongest in the middle of the thinking process (they are “middle-to-middle thinkers,” according to Srinivasan).

At the beginning of the problem solving cycle, the Plan stage, humans excel. We frame the problem, gather context, and decide which direction to pursue. AI struggles here because it lacks real-world understanding. Without clear guidance, it can easily produce irrelevant or misleading suggestions.

At the end of the cycle, the Act stage, humans also dominate. This phase involves judgment, reflection, and the ability to adapt strategies based on experience. Again, these are areas where AI has limitations.

But in the middle stages—analysis, synthesis, and structured reasoning—AI can be extremely powerful. Once a task is clearly defined, language models can process large amounts of information, generate summaries, and suggest possible solutions.

This pattern becomes even clearer when examining different types of problem solving, which I have written about in my book Four Types of Problems. AI performs best in procedural tasks, where both the problem and solution are well understood. In lean terms, this resembles Type-1 problem solving that entails rapid action and response to abnormal conditions. These are tasks, such as writing routine code, summarizing documents, or troubleshooting known issues.

For diagnostic problems, where the issue must be investigated, AI can assist by organizing information and proposing hypotheses. But humans must still verify facts and determine the root cause.

For future-state design (Type 3), AI becomes less reliable. Defining a compelling target state requires observation, imagination, and contextual understanding. These capabilities remain uniquely human.

Finally, in the domain of true innovation, AI struggles even more. While it can recombine existing ideas in interesting ways, it rarely produces genuinely original breakthroughs without strong human guidance.

TECHNOLOGY ALONE IS NOT ENOUGH

For leaders considering AI adoption (and let’s be honest, who isn’t?), the most important lesson may come from Lean Thinking itself. Technology alone does not create transformation.

A simple framework captures this idea: Impact = Technology × Behavior × Management

If any one of these elements is missing, the overall impact drops to zero.

Organizations have learned this lesson many times before. Enterprise software systems, such as ERP platforms, promised dramatic improvements but often failed because companies neglected the behavioral and management changes required to make them work.

AI faces the same risk. Even the most advanced algorithms cannot deliver value if employees do not understand how to use them, or if the organization lacks a management system to guide their application. Consider a familiar example from lean manufacturing: the andon system. Sensors and monitoring tools collect data from machines and production lines, but the technology alone does nothing unless people respond when a problem occurs. Workers must pull the andon cord, supervisors must investigate the issue, and leaders must ensure problems are solved and prevented from recurring. Without these behavioral responses and management routines, the technology becomes meaningless.

FOUR LEADERSHIP PRIORITIES FOR THE AI ERA

So, how can we ensure the adoption of AI produces real value? To me, there are four fundamental points to consider here.

1. Redesign workflows, not just individual tasks. Organizations that see the greatest benefits from AI are those that rethink entire workflows. Simply automating isolated tasks rarely produces major gains. Instead, leaders should examine how work flows from beginning to end and redesign the process around new capabilities.
2. Don’t treat AI as an IT project. AI adoption cannot be delegated entirely to technical teams. It requires strategic thinking about where the technology creates the most value and how it aligns with the organization’s goals. That’s a task for leadership.
3. Focus on augmentation rather than replacement. While some fear that AI will replace large portions of the workforce, the most promising approach is to use AI to enhance human capability. When employees use tools that help them work faster, analyze information better, and solve problems more effectively, organizations can grow rather than shrink.
4. Build trust through transparency and experimentation. Top-down mandates rarely succeed. Employees must understand how AI benefits them and their customers. Starting with practical applications that solve real problems can build credibility and encourage broader adoption.

A CAUTIOUSLY OPTIMISTIC FUTURE

Artificial intelligence will undoubtedly reshape many aspects of work, but its impact will likely unfold more gradually and unevenly than the headlines suggest.

Generative AI is powerful, but it is still a tool that requires thoughtful integration into human systems. When leaders combine technology with the right behaviors and management practices, that technology will accelerate learning, improve decision-making, and enhance productivity.

In the end, the real transformation doesn’t come from the technology itself, but from how organizations learn to use it. In the case of lean organizations, through disciplined experimentation, continuous improvement, and a commitment to developing human capability. Our ability to bring together tech, behavior, and management is what will determine whether AI becomes just another hype cycle or a genuine force for progress.

_THE AUTHOR