Study shows employee initiative decline is linked to management attitude, not salary

Over the past two years, the public discussion about artificial intelligence has focused on the issue of job displacement. This is a understandable concern because tools like ChatGPT and Claude are capable of writing, researching, summarizing, coding, and synthesizing information at incredible speeds.

Anxiety fueled by industry leaders

Concerns were amplified by statements from Sam Altman of OpenAI and Dario Amodei of Anthropic. Amodei warned that AI could eliminate half of entry-level office workers, while Altman spoke of a serious threat to entire entry-level categories.

However, the reality in many organizations is different, and even Altman and Amodei have adjusted their views. Altman recently admitted that he was 'quite wrong' about AI eliminating entry-level jobs, and Amodei now believes that automation will expand human workload rather than destroy it.

The challenge of AI adoption

Analyzing work in public and private structures reveals an unexpected contradiction: employees who have undergone basic AI training often lack access to the necessary tools for its application. Some take advanced courses but do almost nothing with AI since their last training, and this is not due to unwillingness but due to a lack of tools or time for experimentation.

This points to a problem that receives far less attention than automation itself: most businesses struggle to enable the use of AI, creating a huge gap between the theoretical capabilities of modern AI systems and how they are actually applied. There is a vast chasm between companies that are truly AI-driven and those that are not, and this gap is larger than many IT executives realize.

AI as an accelerator, not a replacement

Modern AI models are undeniably impressive, and the current version is the weakest. Considered in isolation, this may seem threatening, but capability itself does not mean replacement. AI can automate routine tasks and assist with writing, coding, analysis, and research. However, it cannot provide organizational context, align stakeholder interests, exercise judgment, understand politics, manage relationships, or make decisions on behalf of a business.

It can write code better than almost anyone alive, but without a human to prompt, guide, evaluate, and verify the result, it achieves nothing significant on its own. The missing element is human judgment and context. A more useful approach is to view AI as an accelerator, not a substitute. When one part of a process becomes significantly faster, the bottleneck simply moves elsewhere. Research that once took days now takes minutes, but the information still requires evaluation, interpretation, discussion, and action.

Impact on software development

This pattern is constantly evident in software development. AI helps teams create initial versions much faster, but then comes a stream of feedback, change requests, testing, and new ideas. The faster teams move, the more opportunities they create, so the work often grows rather than shrinks. Many organizations find that if a team can complete two projects instead of one, the reaction is rarely to work half as much—rather, they take on more tasks. The principle is not about doing more with less, but about doing more with the same resources. This changes the work, but does not necessarily eliminate it.

Risks for entry-level roles

Nevertheless, there are workforce consequences, and entry-level roles deserve the most attention. Many professions rely on apprenticeship models where junior staff gain experience through repetition. If AI eliminates routine tasks for junior staff, such as preliminary research or administrative team work, fewer entry-level positions may be needed.

Talk to an intern, and they will understand working with AI, but ask a follow-up question: 'Without extensive practical experience, how do I know the answer is correct?' They are absolutely right. AI can democratize access to expertise, but experience still matters—perhaps even more, because someone must determine whether the machine's answer is correct.

The problem of shadow AI

Organizations that try to control AI adoption solely through policy are setting themselves up for disappointment. 'Shadow AI' already exists: employees use these tools regardless of company approval, and many simply do not disclose it. Studies have revealed an 'AI disclosure penalty,' where employees admitting to using AI are perceived as less competent, less diligent, or less worthy of recognition. This creates a dangerous dynamic where people continue to use the tools but become less transparent and seek workarounds. The solution is not stricter policy, but creating an environment where people can safely experiment, learn openly, and discuss AI usage without stigmatization.

What AI skills mean

Perhaps the most interesting question currently being asked by HR teams is what AI skills actually entail. The answer is likely not prompt engineering. The most valuable abilities are surprisingly human: defining the right problem, providing context, critically evaluating results, responsible data usage, applying judgment, and perhaps most importantly, transparent work.

These qualities determine whether AI creates value or merely generates more content. The companies that will benefit the most in the next decade will not always be those with the most advanced models. They will be those that give employees access to tools, time for experimentation, safeguards for safe use, and a culture that encourages transparency. Before worrying about whether AI will replace workers, many companies need to solve a more pressing issue: enabling its use.