Agent Sprawl and the Loss of Agency
Our obsession with plug-and-play shortcuts is driving a dangerous surrender of human agency and accountability
In case you haven’t been paying attention, there’s been an aggressive corporate push to move beyond basic AI prompting and jump straight to autonomous AI agents. However, it’s already showing massive, structural cracks, so many that it’s already been given an official name: AI agent sprawl. That’s right, organizations are wrestling with an epidemic of autonomous agents operating without a map or an off-switch.
Despite all the promises that agentic workflows would be the ultimate plug-and-play shortcut to automated magic, it’s proving to be more of an operational horror story. Engineers at Uber burned through their annual AI budget in four months. Elsewhere, unmonitored systems on continuous, automated re-work cycles have racked up jaw-dropping, multi-million-dollar compute invoices in a matter of weeks.
Two weeks ago, I addressed the problem of tokenmaxxing, which I’d encourage you to check out if you missed it. While born of the same corporate impatience, agent sprawl is the dangerous operational inversion of that trend. Rather than humans faking machine productivity, agent sprawl is the result of humans handing the keys to autonomous systems and walking away entirely.
In last week’s episode of Future-Focused (you can check it out on YouTube, Spotify, or wherever you listen to podcasts), I went into more detail explaining the trend and highlighting why autonomous tech isn’t a mythical alien intelligence. I also laid out the pragmatic, real-world risks of charging forward blindly, comparing this wave to a hyper-accelerated version of the 2000s SharePoint crisis.
However, like usual, my brain didn’t stop swirling on what lies just beneath the surface. The more I thought about it, the more I noticed that this is about more than blown corporate budgets, surprise IT bills, or operational pipelines running out of control. While serious, those are just the financial and organizational symptoms of a much quieter, far more hazardous rot taking hold of our corporate structures and our society as a whole.
With that, let’s get into it!
Key Reflections
Misunderstanding Friction
“When you automate away all the tasks you hate, you lose the context and knowledge required to operate.”
One of the biggest AI promises was that it would effortlessly handle all of our “grunt work.” On the surface, it sounds incredibly appealing, and recent market data confirms it’s exactly what people are doing. Countless people are automating everything they’d rather not do. However, we’re taking it too far. It immediately brings to mind people who declare themselves “strategists” as a cover for their refusal to participate in the messy operational implications of their grand ideas. We frequently make the mistake of treating any operational work as administrative noise, entirely convinced that deleting it from our calendars is the apex of success. Tragically, we fail to realize we’re practicing a form of willful ignorance.
I assure you that this friction-free promise is a dangerous trap. We were already living amidst an epidemic of leaders who were profoundly disconnected from the day-to-day realities of their operations. Unfortunately, the agent sprawl turns that pre-existing problem into an absolute crisis. The thing is, we often forget that the tedious, unglamorous parts of our workflows are often where organizational context, critical thinking, and operational visibility live. As a result, if everyone hands the keys over to unmapped autonomous agents simply because tasks are irritating, we end up completely severing our situational awareness. And, if we let it continue, we will end up with an entire ecosystem where nobody knows what’s actually happening.
Automating all the things you hate doesn’t magically transform you into a visionary leader. It makes you functionally blind. And, if we all keep trading our awareness for a dangerous sense of comfort, we will become entirely detached from our businesses.
Anchored to an Illusion
“If you consistently count on AI to give you something to ‘react to,’ you’ve already handed it the steering wheel.”
We’ve all been told that as long as there’s a “human in the loop,” you’re perfectly safe and in control. As a result, many have marched forward, implementing the occasional checkpoint, fully convinced they’re maintaining strategic agency. However, that’s not how it works, and it’s easy to get set adrift. A perfect example of this breakdown is when you ask AI to generate an idea, framework, or outline, “to have something to react to.” It feels harmless, even responsible. I mean, come on. It’s ridiculously easy to convince yourself that because you’re holding the red pen, you’re the ultimate author, and AI is merely a frictionless springboard to accelerate your daily workflow.
However, we have to recognize this is a psychological blind spot almost everyone is missing right now. When a machine frames the initial canvas, you’re not getting a head start. You’re being anchored. AI instantly narrows your field of vision, subtly shapes your direction, and silently dictates the boundaries of the final decision before you’ve even engaged your own critical thinking. You’re no longer evaluating your best work. You’re lazily negotiating with what a machine has statistically determined is good enough. Don’t believe me? Next time it gives you something to “react to,” tell it you believe the polar opposite of whatever it generated. You’ll quickly discover how easy it is for a statistical guessing machine to trick you into allowing it to establish the baseline architecture of your strategy.
Now, I recognize this might be minor if it stays confined to a single email or a routine presentation. However, when you multiply this across an enterprise of exhausted workers running at an unsustainable velocity, you can hopefully imagine how quickly the “humans-in-the-loop” philosophy falls apart.
Unsustainable Velocity
“You cannot solve a structural overload by surrendering human oversight and trusting a machine to police itself.”
We live in a frantic culture obsessed with raw velocity. When leaders look at autonomous agents, they often don’t see them as tools that require careful orchestration. They’re seen as industrial accelerators that can do more work and push things out faster. While true, this velocity problem is severely exacerbated by our tendency to constantly add new things to our employees’ plates while almost never taking anything off of them. We also often forget to effectively communicate when priorities change. As a result, we end up throwing agentic workflows at the ever-growing mountain of obligations, assuming we can automate our way over the structural overload we created.
Mitigating this risk requires us to acknowledge that the pace of work today has become fundamentally unmanageable. If we don’t start there, we will continue to do nothing to structurally fix it. Over time, we will inevitably watch our agency slowly be crushed by the velocity we enabled, which outpaces any human’s ability to audit or govern it. To survive the overwhelming deluge, we have to break the vicious cycle. If we don’t, we will have no choice but to surrender more control in order to survive. The solution to a broken organizational pace isn’t more or faster automation. It comes from leaders who are courageous enough to step back, slow down, and say no to doing more, sometimes to genuinely good ideas.
The catastrophic consequence that will come from this velocity loop won’t come from a blown compute budget. It will be the result of a systematic, top-down abandonment of work to a machine your people can no longer track.
Preventing Organizational Self-Termination
If we don’t stop treating people like temporary placeholders until AI finally arrives, we’re consciously accelerating inevitable collapse.
We need to have an honest conversation about how we got here. What we’re seeing is largely driven by a massive leadership failure. Influenced by the promises of AI, many executives quietly and lazily abandoned their employees, believing AI would either completely replace them or leave only a light burden that could be carried by anyone desperate enough to carry it. As a result, they stopped investing in meaningful human development, slashed training budgets, and overlooked the stall in upskilling. It has been a total abdication of the core responsibility of leadership, wrapped in the glossy veneer of technological progress.
However, the cracks are becoming too big to ignore, and leaders are discovering just how catastrophically wrong they were. Treating people like placeholders until AI finally delivered on its promises allowed foundational human competence to erode. Fearing for their livelihood, employees simply followed orders, checked out, and became dangerously overreliant on AI tools they were never trained to use effectively. Now, as the autonomous agent sprawl triggers massive operational fractures and astronomical compute invoices, many are finding they’ve lost the human safety net they need to survive the fallout. The unglamorous truth is that some are discovering their decision to turn up the autonomy volume on a broken corporate architecture has led to autonomous, operational self-termination.
The true work of a future-focused leader begins right now. If you want to reclaim your organizational sovereignty, you have to put a full stop to unmanaged enterprise rollouts, re-establish the trust you broke, and invest in rebuilding the human capability you tried to shortcut. If we don’t change course, it may prove to be the single worst choice we have made for our businesses and our society.
Concluding Thoughts
As always, thanks for sticking around to the end.
If what I shared today helped you see things more clearly, would you consider buying me a coffee or lunch to help keep it coming?
Also, if you or your organization would benefit from my help building the best path forward, visit my website to learn more or pass it along to someone who would.
I hope this article has shaken everyone brave enough to read it free from the hypnosis that has entranced so many. There are a lot of messes we can clean up and problems we can fix, but if we don’t get a handle on how quickly human agency is slipping right out of our hands, it’s hard to tell what it will take to get it back.
The good news is it’s not too late. The damage isn’t already done. We just have to stop and take an honest look at where we are right now, being willing to accept what we see even if we don’t like it. From there, we can begin to map out a new path and change our trajectory before we’re too far gone.
Oh, and if this article left you feeling like there’s no possible way things can go well, wait until my next one. Having spent a great deal of time highlighting risks, I wanted to intentionally shine a light on an incredible example of what good can look like, hopefully inspiring us not just to “stop” but to “do better.”
With that, I’ll see you on the other side.
This was fire