The Crisis of Autonomy: Why “Digital Agency” May be the Most Critical Skill of the AI Era

Last week I moderated a panel on how AI is changing secure software development.  Several interesting topics emerged, but something struck me: panelists repeatedly returned to the continued importance of the role of humans.  The necessity of being ever vigilant, despite the temptation of offloading an increasingly large workload to autonomous software.

In the rush to integrate Generative AI into our workflows, we are witnessing a profound shift in the workplace. While we focus on efficiency gains and prompt engineering, we often overlook a looming risk: the erosion of human agency.

As someone whose career spans education and digital transformation, with a decades-long interest in history and international relations, I’ve spent my career watching how large-scale systems impact human behavior. Today, the challenge isn’t just learning how to use AI; it’s learning how to remain the “agent” in an increasingly automated world.

What is Digital Agency?

Digital Agency is the ability of an individual to exert power and make intentional choices within a digital ecosystem. It is the difference between being directed by an algorithm and leveraging one.  And it is more than human-in-the-loop, it is about being part of shaping and guiding a process.

Without a rigorous strategy for agency, organizations risk falling into “Cognitive Offloading”, a state where we stop verifying, stop questioning, and eventually, stop thinking critically because the AI provides good enough (oftentimes fantastic) answers.

The Challenges We Face

1. The “Black Box” Dependency: When users don’t understand why an AI suggests a specific path, they lose the ability to course-correct. This creates a fragile workforce that breaks whenever the technology falters. Remember the CrowdStrike outage in July 2024?  The one that brought organizations to a standstill?  The control of systems had been outsourced, not properly understood, and when things went wrong, it ultimately required, ironically, human agents performing a one-by-one surgery on millions of machines to solve a problem the automated system could not see.
2. The Skills Atrophy Loop: If we outsource junior-level thinking to AI today, we are effectively burning the bridge to senior-level expertise tomorrow. Case in point: According to a study by Anthropic, junior engineers who used AI to delegate code generation scored 17% lower on subsequent conceptual and debugging tests than those who coded manually. 
3. The Compliance Trap: Rapidly shifting global regulations (like Canada’s Bill C-63 or the EU AI Act) create a fear-based culture where employees stop experimenting for fear of breaking a rule they don’t understand. This can also be true of organizational policies and procedures if they evolve in unpredictable or uninformed ways.

A note of caution: when the printing press essentially created an explosion of new data and access, the regions that flourished were those that encouraged critical skepticism and philology.  People learned to approach information with caution instead of credulity, paving the way for the scientific revolution.  Regions that attempted to control and codify the new data through expanded bureaucracy and overreach tended to stagnate both intellectually and economically.   

Building an “Agency-First” Culture

To overcome these challenges, organizations need to move beyond technical training (though this is also fundamental) and focus on three pillars of Digital Agency:

1. Strategic Skepticism: We must train teams not just to prompt, but to interrogate. Digital Agency requires the confidence to reject an AI output when it lacks the nuance of human experience or historical context.  See the refund Deloitte was forced to issue to the Australian government in 2025 because a key report on the national welfare system was drafted largely by GenAI without adequate human interaction, prompting an Australian senator to remark that the firm had a “human intelligence problem”. 
2. Intellectual Sovereignty: Organizations must define human-in-the-loop protocols that are not just checkboxes, but active decision-making milestones where the human agent must provide the “why” behind the AI’s “what.”  See the recent case of a prominent law firm that failed to remove AI hallucinations from a high stakes court filing. 
3. Transition Resilience: We need to build Adaptive Intelligence (AQ). This isn’t about mastering one tool or software; it’s about building the psychological and structural capacity to pivot as AI capabilities change every month.

We often think of friction as a bad thing. But in education, professional development, and digital transformation, friction is where learning happens. When we use AI to remove all struggle, whether it’s a junior developer writing a function or a lawyer drafting a brief, we aren’t just saving time; we are bypassing the very process that creates expertise. We are trading long-term Intellectual Sovereignty for a short-term productivity spike.

The Path Forward

The goal of digital transformation shouldn’t be to see how much we can automate. It should be to see how much we can augment the human agent.

History shows us that the civilizations and organizations that thrive during technological shocks are those that prioritize human judgment over mechanical rote (examples include adapting to the printing press, the Industrial Revolution, and the Appollo Program, though there are countless more). As we navigate this new frontier, our focus should be helping leaders build teams that don’t just survive the AI shift, but lead it with intention, autonomy, accountability, and agency.

How is your organization protecting human agency in the age of automation? Let’s start the conversation.