My 2026 Thoughts on AI & Human Agency

It’s May 20, 2026, and I’m staring at a blank screen, trying to figure out how to start this piece. The irony isn’t lost on me. I’m writing about the future of human agency in a world increasingly shaped by AI, and I’m experiencing a very human block. Just last week, I asked an LLM for some blog post ideas – not for content, mind you, but for a kickstart. The suggestions were… predictable. “The ethics of AI in healthcare,” “The future of work,” “AI and creativity.” All valid, all well-trodden paths. It made me think: are we, as humans, becoming predictable in our discourse about AI because we’re leaning too heavily on what the machines themselves can generate?

My focus today isn’t on the broad strokes of AI’s impact, but on something much more specific and, I think, much more insidious: the slow, almost imperceptible erosion of our ability to formulate novel questions. Not just answers, mind you, but the very act of framing a problem in a way that hasn’t been done before. This isn’t about AI taking our jobs; it’s about AI taking our intellectual curiosity, or at least guiding it down pre-approved channels.

The Echo Chamber of Answers: When AI Defines Our Questions

Think about it. When you have a question, where do you go? Google, sure. But increasingly, it’s an AI chatbot. And what do these chatbots do? They provide answers based on the vast corpus of existing human knowledge. They synthesize, they summarize, they extrapolate. What they don’t do, at least not yet, is generate truly new lines of inquiry. They don’t look at a problem and say, “What if we completely redefined the parameters here?”

I remember a conversation with a friend, a brilliant data scientist, who was struggling with a complex optimization problem for a client. He’d thrown everything at it – traditional algorithms, machine learning models, even tried to phrase his queries to an advanced AI. Each time, the AI would give him variations of existing solutions, or point him to papers already published. He was getting stuck in an echo chamber of answers. The breakthrough came when he stepped away from the screens and just started doodling on a whiteboard, asking himself absurd, “what if” questions that no AI would ever suggest because they weren’t based on existing data. He ended up reframing the entire problem, which led to a solution no model would have found.

The “Optimal” Path vs. The Inventive Detour

This is where the agent philosophy really comes into play. An agent, by definition, has the capacity for independent action and decision. But what happens when the “optimal” path, as defined by an AI, becomes so overwhelmingly efficient and readily available that we stop looking for alternatives? The human mind, with its biases, its irrational leaps, its moments of pure serendipity, is often the engine of true novelty. AI, by its very nature, seeks patterns and efficiency within existing data. It’s a magnificent tool for optimization, but a potentially stifling force for genuine invention if we let it dictate our intellectual journey.

Consider the rise of AI-powered content generation for things like marketing copy, blog post outlines, or even initial drafts of technical documentation. It’s incredibly efficient. A marketing team can churn out dozens of variations of ad copy in minutes. But how often does that lead to a truly disruptive, category-defining campaign? Or does it just lead to increasingly polished, but ultimately similar, messaging that blends into the noise? The “best” ad copy, as defined by an AI, is usually the one that has historically performed well. It’s a self-fulfilling prophecy of sameness.

Cultivating the Unprompted Question

So, what do we do? How do we preserve our ability to ask truly original questions in an age where answers are always just a prompt away? It requires a conscious, deliberate effort to push back against the tide of readily available “optimal” solutions.

Practical Tactic 1: The “Un-AI” Brainstorm

My first recommendation is to institute what I call “Un-AI Brainstorm” sessions. When faced with a new challenge or a problem that seems intractable, resist the urge to immediately turn to an AI. Instead, gather your team (or just yourself, with a notepad). The rules are simple:

1. No digital devices allowed for the first 30 minutes.
2. Start with absurd questions. “What if gravity worked in reverse for this problem?” “What if our customers were all cats?” The goal isn’t to find an answer, but to break free from conventional thinking.
3. Focus on framing the problem in at least five completely different ways before even thinking about solutions.

This isn’t about shunning AI; it’s about making sure your initial framing of the problem comes from a human place of curiosity and unconstrained thinking, rather than being shaped by the patterns an AI has learned.

Practical Tactic 2: The “Deconstruct and Reconstruct” Exercise

When you *do* use an AI, don’t just accept its output. Take an AI-generated answer or solution and actively try to deconstruct its underlying assumptions. Then, try to reconstruct it from a completely different set of assumptions. This helps you understand the boundaries of the AI’s “thinking” and, more importantly, reveals the unexplored territory outside those boundaries.

For example, if an AI suggests an optimal marketing strategy based on past user data, ask:

• What if our target demographic completely changed overnight?
• What if we had zero budget, but unlimited creative freedom?
• What if our competitors suddenly vanished?

These kinds of thought experiments, which deliberately break the AI’s learned context, are crucial for fostering novel lines of inquiry.

Practical Tactic 3: Embracing “Productive Friction” in Development

In software development, there’s a growing trend towards AI-assisted coding. Tools that auto-complete, suggest functions, or even generate entire code blocks are becoming ubiquitous. While undeniably efficient, this can lead to a kind of intellectual laziness where developers stop thinking deeply about *why* a particular solution is best, or what alternative approaches might exist.

One way to counteract this is to deliberately introduce “productive friction.” When an AI suggests a piece of code, don’t just accept it. Take a moment to write a small, deliberately different version yourself, even if it’s less efficient or more verbose initially. Compare the two. Ask yourself:

• What are the edge cases my version handles differently?
• What assumptions did the AI make that I didn’t?
• Could a simpler, more elegant solution exist that neither of us found immediately?

Here’s a simple Python example. Let’s say an AI suggests this for finding unique elements:

def get_unique_elements_ai(items):
 return list(set(items))

It’s perfectly valid, and efficient. But before you just copy-paste, consider a more manual approach, perhaps for the sake of understanding or to explore different performance characteristics if the input is always sorted, for instance:

def get_unique_elements_human(items):
 unique = []
 seen = {}
 for item in items:
 if item not in seen:
 unique.append(item)
 seen[item] = True
 return unique

While the AI’s version is often preferred for its conciseness, building the manual one forces you to think about the underlying mechanism, the data structures involved, and potential trade-offs. This isn’t about rejecting efficiency; it’s about maintaining a deeper understanding and the ability to innovate beyond the “optimal” suggestion.

The Future of Agency: Not Just Answering, But Asking Better

The future isn’t about humans competing with AI to provide answers. AI will always be faster, more comprehensive, and less prone to factual error in that domain. The future of human agency, our unique contribution, lies in our capacity for imaginative, unprompted questioning. It’s about the audacity to look at a perfectly good answer and say, “But what if we asked something entirely different?”

This isn’t just an abstract philosophical point. It’s a critical skill for innovation, for problem-solving, and for maintaining our intellectual sovereignty. If we let AI define the boundaries of our questions, we will, inevitably, limit the scope of our potential answers, and by extension, our future.

So, the next time you’re faced with a problem, before you type it into a chatbot, take a moment. Grab a pen. Stare at the ceiling. Ask yourself a question that feels a little bit silly, a little bit irrational, and definitely not something an algorithm would suggest. That’s where true agency begins.

Actionable Takeaways:

• Prioritize Pre-AI Brainstorming: Before turning to AI for solutions, spend dedicated time (even just 15-30 minutes) brainstorming problems and potential angles without any digital assistance. Focus on framing novel questions.
• Deconstruct AI Outputs: When you use AI, actively analyze its suggestions. Identify its underlying assumptions, and then challenge those assumptions by re-framing the problem with different premises.
• Introduce “Productive Friction” in Creative/Technical Work: Don’t blindly accept AI-generated code or content. Take time to manually create alternatives, even if less efficient, to deepen your understanding and explore different approaches.
• Cultivate Absurdity: Encourage yourself and your teams to ask “what if” questions that defy current logic or data. These seemingly irrational queries are often the seeds of genuine innovation.
• Reflect on Your “Question Diet”: Periodically assess where your questions are coming from. Are you mostly asking questions that an AI could easily answer, or are you pushing into truly unknown territory?

🕒 Published: May 20, 2026