The AI and Human Intelligence Paradox: Superhumans or Zombies to the Machines

AI and Human Intelligence

There are plenty of contradictions and paradoxes as they relate to artificial intelligence, especially when looking at large language models (LLMs). Hypergrowth versus financial investment and ROI? Will it create or destroy jobs? And will it lead to better or werse outcomes for humanity? The tech CEOs themselves are unsure, as both Sam Altman and Dario Amodei have recently changed their answers on the jobs question.

The uncertainty and discourse go beyond the boardroom. Days before this article’s publication, the U.S. government citing national security concerns, issued an export-control directive ordering Anthropic to suspend its two most capable models (Fable 5 and Mythos 5). Fable 5 had launched three days prior, making this the first time a frontier model has been pulled from market by government order rather than by the company itself. Meaning, workflows, processes, or habits built around these tools were gone overnight. The episode shows that over-reliance on any single AI tool is its own form of risk. Humans who navigate this era effectively are the ones who bring the skills and judgment to adapt to these changes.

To bring things down to a human level, we can look at how people are currently using LLMs, how that behavior has changed, and what a person or leader can do to thrive amid the uncertainty surrounding LLMs, AI, and the fate of humanity (pardon the hyperbole it was too fun not to include).

The main contradiction we will review is that the same technology that can amplify human potential can also diminish human agency. Exploring the question: Will AI bolster humans to do greater things, learn anything they are passionate about, and enhance their lives? Or, whether it will make us zombies to the machines?

So, how are we using generative AI, and how has that shifted year-over-year?

Marc Zao-Sanders and his team have been studying this and publishing their results in HBR (1). The top use for both 2025 and 2026 was the only one to remain stagnant year-over-year: therapy/companionship. This is fascinating, but that is a topic for a different paper and a different author (top ten by year in figure).

The authors call out a trend they termed “thinkslop,” noting that in at least a quarter of the top use cases, users are asking AI to perform some portion of their thinking (therapy/ companionship [#1], relationship advice [#7], enhanced decision-making [#13], organizing my life [#14], drafting emails [#42], generating ideas [#47]).

For leaders and individuals alike, the question has become are these tools being used to compound thinking and judgment or erode them.

Stay Human

Early in the AI era (GPT-4), BCG set out to understand how people “…Can Create – and Destroy – Value with Generative AI (2).” The study found that when consultants used generative AI within its area of competency, they saw roughly a 90% improvement in performance. However, on tasks outside AI’s competence, such as business problem solving, those who used it performed 23% worse than those who didn’t use the tool at all. It is worth noting this was conducted on an older model, and it is possible these numbers have shifted with newer releases.

Shannon McKeen at Wake Forest took a different approach, allowing students to choose whether to use AI on a group project (3). Groups only had to disclose if AI was used or not. The groups that did use AI produced presentations better than the professor had seen in their tenure, while the groups that did not produced the standard undergraduate fare (an expected and unsurprising result). The more telling finding came when groups had to present and defend their work. Students who had done the work themselves could articulate the reasoning behind their conclusions, while those who had relied on AI were unable to defend their recommendations to a panel.

So, between consultants misusing AI, “thinkslop”, and students submitting work they cannot defend, what are we to do?

The answer appears to be keeping the human aspects human and utilizing AI for augmentation. For me, that means brainstorming independently before turning to AI. Which looks like forming a point of view, handwritten outlines, and a few long walks before the computer weighs in.

Zao-Sanders offers two tips to combat over-reliance: first, don’t start with AI; give yourself time to think and form an opinion. Second, draw boundaries around which parts of your work will and won’t be outsourced to AI (1).  A whiteboard and a strong team still have an irreplaceable place in business and remain a reliable antidote to the pitfalls discussed here and ahead.

Accountability

The term “workslop” seems to have entered the zeitgeist of the corporate world precisely because of how significant a problem it has become. Kate Niederhoffer et al. studied workslop and its impact on workplace productivity, finding that 40% of respondents had received workslop, leading to approximately two hours of additional rework per instance. When extrapolated to a 10,000-person company, they estimated $9M a year in lost productivity (4). The study also found that individuals rated those who sent workslop less favorably across every dimension measured (creativity, capability, reliability, trustworthiness, and perceived intelligence).

The BCG study made a related point worth underscoring adopting generative AI is fundamentally a change management challenge, and the leader’s job is to help people use new technology in the right way, for the right tasks, while continuously adapting as AI’s capabilities expand (2).

The counter to workslop is accountability. Leaders and organizations cannot allow employees to cognitively offload to others or outsource their thinking to LLMs. LLMs can be a legitimate part of the team and the collaboration while the employees using them remain responsible for the deliverable. Niederhoffer et al. argue that seamless collaboration going forward must include how AI work products are incorporated into shared workflows in service of outcomes, rather than used as cover for avoiding responsibility (4).

Collaboration paired with accountability will allow teams to own their work while leveraging AI to add real value and productivity.

Skills

Knowing what not to outsource is the first half of the equation. The other half is knowing which skills to actively build. As AI shapes the future of employment, leaders need to know which skills to prioritize when hiring, and individuals need to invest in skills that increase in value as AI further integrates into the workplace. Five skills that are becoming increasingly valuable are: discernment, systems thinking, design, storytelling, and authenticity.

Discernment

When the entire world’s knowledge base is one question away, it is imperative to know what to ask. Discernment is vital across all AI use. This starts with how, when, where, and what a person chooses to use AI for. Those answers depend on that person’s goals, competencies, and constraints, as discussed in the previous sections.

Skilled professionals will need to be discerning when evaluating the outputs they receive. Strong prompt writing and requiring the tool to cite its sources can help avoid hallucinations, but ultimately the accountable human will need to fully validate the output and be able to defend the point of view it supports. I personally require any LLM I collaborate with to cite its work. I suppose my semester as a TA in Research Methods never quite left me. This allows me to validate the primary sources, understand any potential bias from those sources, and dive deeper into the topic.

A discerning AI user will go one step further, using the tool to interrogate their own thoughts, ideas, and assumptions. The human should arrive with a point of view, and the tool should be used to sharpen and stress-test it. This surfaces false assumptions or biases that could undermine the overall argument.

Discernment applies to what gets built. With how far coding tools have come, individuals can now build prototypes, tools, and functional software in half a day. But capability does not equal necessity. I recently listened to the All In podcast, where a guest described an employee who built a tool that solved a real problem and was recognized at a company town hall (5). Shortly after, two other employees independently built the same tool. Anyone could, so three people did. What gets built should be guided by the problem being solved, not by the novelty of being able to build it.

Finally, leaders need to discern about which metrics they track when implementing AI. As several major companies have recently discovered, token usage is not a meaningful measure of value. Results and outcomes matter far more than activity.

Systems Thinking

The LLM has all the knowledge and none of the context. Context is crucial when writing prompts, and even more vital when determining where and how to incorporate AI agents into workflows. Individuals who can understand full systems will be essential to integrating AI effectively. They will be able to evaluate the cases for and against utilizing certain tools for certain jobs. As LLMs’ capabilities expand and more tools come to market through vibe coding. Those who understand the larger system will be the ones who identify where each tool can add the most value.

I started my career on a reporting and analytics team. The VP I reported to was one of the best I had ever seen at this. He could map a process and identify fixes for redundancy, reliability, and efficiency from a brief explanation alone. I learned more from him about systems than I did about reporting itself. Though he did teach me plenty about formatting a graph the right way.

Design

Creation is no longer a bottleneck. Images, PowerPoints, stories, and virtually anything else can be generated in minutes. The effort to create has shrunk to near zero, but refinement and taste remain distinctly human activities. Humans will still need to be the editor and the visionary.

This is an area I have worked tirelessly on as I continue to grow. I tend to design for purpose and clarity over style. Unfortunately, as it turns out, so do LLMs. Which means this is one area where AI will not push me toward better taste. I am fortunate to have colleagues with strong design instincts I can turn to for guidance. Claude may eventually go to design school, but understanding design in the context of the current cultural moment (what feels fresh versus dated, bold versus cluttered) will remain a distinctly human edge.

Storytelling

As much as this author enjoys facts and figures, they are rarely the crucial piece of persuasion or influence. Bringing the complexity of information together into a context that others can understand requires a story. And, the best stories come from those who lived it.

John Winsor pondered what has happened to the thought leader in the AI age (6). He introduces the concept of “thought doers” and argues that the stories and details a thought doer can share cannot be replicated by someone who hasn’t lived them.

Humans connect to stories, and those who can tell one from a first-person perspective will remain persuasive and earn lasting credibility. As someone who leads with numbers, I have learned this lesson more than once. I never present data without first having the headline and the story clearly defined.

Authenticity

My mom always loved the cliché attributed to Oscar Wilde: “Be yourself; everyone else is already taken.” And I begrudgingly admit she was right. AI has only amplified how much being yourself is a differentiator. This cuts across any use case, but especially anything writing or personal brand related.

Another area where authenticity matters is when providing a point of view on a decision. Anegelo Romansanta, Llewellyn Thomas, and Natalia Levina asked LLMs for strategic advice and found they got “trendslop” in return (7). They described this as a bias toward whatever is currently considered modern, progressive, innovative, or managerially fashionable, and resulting in recommendations driven by trend rather than context.

Your experiences and judgment can cut through that, helping identify what a company or situation needs rather than what is generically prescribed. That ability will become increasingly fundamental for those who want to remain relevant in business.

Conclusion  

Whether AI produces superhumans or zombies to the machine will not be decided by the technology. It will be decided by the choices humans make in using it.

LLMs and AI present a significant opportunity for individuals, leaders, and organizations, but only if utilized as tools to augment human intelligence rather than replace it. For all the predictions about what the future holds, nothing has been decided. The outcome will be shaped by the humans using these tools, how they choose to use them, and the culture, guidance, and guardrails leaders build into their organizations. Keeping humans central and fostering a culture of accountability and collaboration will determine who extracts real value from AI and who generates activity without impact. The skills that make us most human will be the ones that differentiate talent in the AI era.

References

• Chui, M., Yee, L., Hall, B., & Singla, A. (2026, June 1). How people are really using AI in 2026. Harvard Business Review.
• Candelon, F., Zuluaga-Martinez, D., et al. (2023, September 19). How people can create—and destroy—value with generative AI. Boston Consulting Group.
• McKeen, S. (2026, May 18). AI is creating a dangerous illusion of competence.
• Dolan, K. & Metcalf, L. (2025, September 22). AI-generated “workslop” is destroying productivity. Harvard Business Review.
• Anderssen, E. (Host). (n.d.). Anthropic’s digital god, Pope vs AI, job loss narrative flips, open source crackdown coming?[Video].
• Winsor, J. (2026, March). Has AI ended thought leadership?Harvard Business Review.
• Romasanta, A., Thomas, L., & Levina, N.(2026, March 16). Researchers asked LLMs for strategic advice. They got “trendslop” in return.Harvard Business Review.

SalesGlobe is a leading sales effectiveness and data-driven creative problem-solving firm. We specialize in helping Global 1000 companies solve their toughest growth challenges and helping them think in new ways to develop more effective solutions in the areas of sales strategy, sales organization, sales process, sales compensation, and quotas. We wrote the books on sales innovation with The Innovative Sale, What Your CEO Needs to Know About Sales Compensation, and Quotas! Design Thinking to Solve Your Biggest Sales Challenge.