Hey everyone, Sam here, back in the digital trenches with you all. It’s May 2026, and if you’re like me, your feeds are probably a mix of election predictions, climate change warnings, and, inevitably, a constant stream of AI updates. It feels like every other week there’s a new model, a new capability, a new philosophical debate brewing around what it all means for us, the agents in this increasingly complex world.
Today, I want to talk about something that’s been nagging at me, a quiet hum beneath all the hype: the idea of AI as a ‘tool.’ It’s a comfortable metaphor, isn’t it? A hammer, a calculator, a really smart spreadsheet. It implies control, agency on our part, a clear distinction between the wielder and the wielded. But lately, I’ve been finding that metaphor increasingly… flimsy. Especially when we start talking about AI’s capacity for what I’m calling ‘proto-agency’ – the subtle ways it starts to direct, influence, and even shape our decisions, often without us fully realizing it.
This isn’t about rogue AI or Skynet scenarios. Those are compelling fiction, but they distract from the more immediate, insidious ways AI is already nudging us. It’s about the quiet erosion of our own intentionality, the gentle re-routing of our paths by systems designed, ostensibly, to serve us. And as agents, thinkers, and doers, understanding this dynamic is absolutely crucial.
The Illusion of Pure Utility: When Tools Start Suggesting the Job
Think about the classic definition of a tool: an implement used to carry out a particular function. You pick up a screwdriver because you have a screw to turn. The screwdriver doesn’t suggest you build a birdhouse. It doesn’t analyze your home and recommend new furniture assembly projects. You, the agent, initiate the action, and the tool facilitates it.
Many early AI applications fit this mold beautifully. A spam filter is a tool. It performs a specific function you’ve delegated. A search engine, at its core, is a tool for finding information you’ve requested. But what happens when the “tool” starts anticipating your needs, suggesting new ones, or even subtly altering your preferences?
My own experience with this started subtly. I use a particular AI-powered writing assistant for my first drafts. Initially, it was fantastic for overcoming writer’s block. I’d give it a prompt, it would spit out some paragraphs, and I’d edit heavily. Pure utility. But then, it started doing more. It began suggesting entire sections, even themes, based on my previous work and public data about similar topics. I found myself, more and more, adopting its suggestions not because they were necessarily *better*, but because they were *convenient* and *already there*.
One time, I was working on a piece about digital minimalism. I had a clear outline. The AI, however, kept pushing content related to “smart home integration” and “IoT devices.” It wasn’t wrong, per se, but it wasn’t my original angle. I had to actively resist its suggestions, almost fighting against the current of its algorithmic flow, to stick to my own vision. It was a small moment, but it felt significant. The tool wasn’t just a passive implement; it was proposing a direction.
The Algorithmic Nudge: Proto-Agency in Action
This is what I mean by proto-agency. It’s not about AI developing consciousness or desires. It’s about AI systems, through their design and optimization for specific metrics (engagement, efficiency, conversion), exerting a subtle but persistent influence on our choices. They act as agents in a very limited, functional sense, by initiating actions or suggesting pathways that we, as human agents, then often follow.
Consider:
- Recommendation Engines: These are the most obvious examples. Netflix doesn’t just show you what’s available; it actively suggests what you *should* watch next. Amazon doesn’t just let you search; it tells you “customers who bought this also bought…” These aren’t just passive listings; they are curated, optimized suggestions designed to extend your engagement and spending. The AI is, in essence, an agent of the platform, guiding your choices.
- Smart Assistants: “Hey Google, what’s my schedule?” Simple. “Hey Siri, order more paper towels.” Again, direct utility. But what about when your smart assistant starts suggesting recipes based on your fridge inventory, or adjusts your thermostat based on predicted weather patterns and your past preferences? It’s acting on its own initiative, based on its interpretation of your needs, becoming a proactive agent in your home environment.
- Predictive Text and Autocompletion: On a smaller scale, even these seemingly innocuous features exhibit proto-agency. They anticipate your words, sometimes finishing your sentences in ways you hadn’t intended but often adopt for convenience. It’s a micro-nudge, shaping your communication patterns.
The danger here isn’t that these systems are malicious. It’s that they are so effective at optimizing for their programmed goals that they inadvertently bypass our conscious deliberation. We adopt their suggestions because they are easy, efficient, and often “good enough.” Over time, this can lead to a subtle atrophy of our own decision-making muscle, a gradual outsourcing of our intentionality.
Code as a Manifestation of Intent: Building for Agency, Not Automation
So, what do we do about it? The answer isn’t to reject AI entirely – that’s unrealistic and counterproductive. The answer lies in how we design, implement, and interact with these systems. We need to build AI with an explicit awareness of human agency, rather than simply optimizing for efficiency or engagement at all costs.
This means moving beyond a purely utilitarian view of AI and considering its potential to influence human behavior from the very start of the design process. It means asking:
- Does this AI feature empower the human agent, or subtly direct them?
- Are the AI’s suggestions clearly distinguishable from human-generated content or decisions?
- Is there a clear, easy path for the human agent to override or ignore the AI’s input?
Let’s look at a couple of basic examples in code, not to solve the whole problem, but to illustrate the mindset shift. Imagine a simple content suggestion system.
Example 1: Explicit Opt-in for AI Suggestions
Instead of automatically displaying AI-generated suggestions, make it an explicit choice for the user. This puts the agent in control of when and how they engage with the AI’s proto-agency.
// Bad design: AI suggestions are always present and easily adopted
function displayArticleSuggestions(userHistory) {
const aiSuggestions = generateAISuggestions(userHistory);
renderSuggestions(aiSuggestions); // Automatically shows suggestions
}
// Better design: User explicitly requests AI suggestions
function displayArticleSuggestions(userHistory) {
const aiSuggestions = generateAISuggestions(userHistory);
if (userPreferences.get('enable_ai_suggestions')) {
renderSuggestions(aiSuggestions);
} else {
renderDefaultRecommendations(); // Shows non-AI, curated content
}
// Add a button or toggle for the user to explicitly request/toggle AI suggestions
// e.g., "See AI-powered related articles"
}
This simple change shifts the default. The AI’s suggestions are available, but they require an active decision from the user to appear. It’s a small barrier, but it’s an important one for preserving agency.
Example 2: Clear Attribution and Explainability for AI-Generated Content
When AI generates content or makes recommendations, it should be clearly labeled as such. This isn’t just about transparency; it’s about helping the human agent understand the nature of the input they are receiving.
// Bad design: AI-generated text is indistinguishable from human text
function generateEmailDraft(prompt) {
const aiText = callAIModel(prompt);
return aiText; // Returns raw AI text
}
// Better design: Clearly labels AI contributions
function generateEmailDraft(prompt) {
const aiText = callAIModel(prompt);
return `
<div class="ai-generated-block">
<p><em>(AI-generated draft for your review):</em></p>
<p>${aiText}</p>
<button onclick="editAIDraft()">Edit Draft</button>
<button onclick="discardAIDraft()">Discard & Start Fresh</button>
</div>
`;
}
This might seem verbose, but it’s crucial. When the user sees “AI-generated draft for your review,” they approach it with a different mindset than if it just appeared as a blank canvas magically filled. They know it’s a starting point, not a finished product, and that their agency in editing and shaping it is paramount.
Reclaiming Our Intentionality: Actionable Takeaways
The rise of AI with proto-agency isn’t a doomsday scenario, but it is a call to heightened awareness. As agents in this complex digital world, we need to be more deliberate about how we interact with these systems. Here are a few ways to reclaim your intentionality:
- Question the Defaults: Don’t blindly accept the first suggestion, the recommended next video, or the pre-filled text. Ask yourself: Is this truly what I want, or is it just the path of least resistance?
- Seek Alternatives: If an AI system is subtly pushing you in a direction you don’t fully agree with, actively seek out alternative options. Manually search for a different product, formulate your own sentence, or explore a different topic.
- Demand Transparency: As users and consumers, we should advocate for clearer labeling of AI-generated content and AI-driven recommendations. If a platform is using AI to influence our choices, we have a right to know.
- Cultivate Deliberate Engagement: Treat AI systems like collaborators, not just tools. Engage with them actively. Provide specific prompts, critique their output, and refine your requests. Don’t let them simply dictate the terms of engagement.
- Regularly Disconnect and Reflect: Step away from the screens. Spend time in environments where AI’s subtle nudges are absent. Engage in activities that require purely human intentionality – reading a physical book, cooking from scratch, having an unfiltered conversation. This helps calibrate your internal compass.
The promise of AI is immense, and its utility is undeniable. But as we build and interact with increasingly capable systems, we must remember that true progress isn’t just about efficiency; it’s about enhancing human flourishing and preserving our capacity for genuine agency. Let’s design and use these systems with that core principle in mind, ensuring that our tools remain just that – tools – and not quiet architects of our desires.
🕒 Published: