The Real Automation Problem Is Not Work, It Is Power

Jun 17, 2026

10 min read

1 views

87%

The Question Beneath Automation

What if the biggest obstacle to automation is not technical at all, but political?

That sounds like a strange leap: one moment we are talking about recording workflows in Loom, feeding them into an AI model, and improving operations; the next we are talking about democratic decay, elite influence, and global power structures. But the connection is deeper than it first appears. Both stories are about who gets to define the process.

A workflow, at its simplest, is a miniature government. Someone decides what the steps are, who can change them, what counts as success, and who is allowed to question the system. AI does not merely speed up that system. It can also freeze it, expose it, or quietly centralize control over it. The real question is not whether a process can be automated. It is: who benefits when the process becomes legible enough to automate?

That is why the same technology that promises liberation can also deepen inequality. AI can make organizations more efficient, but efficiency is never neutral. It always redistributes discretion. Once you see that, the leap from office SOPs to state power stops looking so bizarre.

Every Process Hides a Power Structure

Most people think of automation as a productivity story. You record the steps, turn them into an SOP, and let software or AI handle the repetitive parts. That is true, but incomplete. Every repeated process encodes assumptions about what matters, what can be standardized, and who has authority to intervene.

Consider a simple example: customer support.

If a team records its support workflow, an AI can quickly identify patterns, suggest macros, route tickets, or flag common failure points. That is efficient. But it also changes the balance of power inside the team. The person who understands the system best, or the person who can explain it most clearly to the machine, gains leverage. The people whose work is harder to formalize may become invisible, while the people who design the workflow become more powerful.

This is exactly what happens at larger scales too. Governments and institutions are often described as democracies because people can vote. But actual influence usually lives elsewhere, in the places where processes are written, funding flows are set, and defaults are chosen. The formal structure looks open, yet the operational structure is often narrow. In business terms, the interface is democratic; the backend is not.

The deeper an organization’s procedures become, the easier it is to automate them. The easier they are to automate, the more power shifts to whoever controls the procedure itself.

That is the hidden symmetry between office automation and political economy. A workflow is not just a sequence of tasks. It is a concentration of judgment. When AI enters the picture, it does not eliminate judgment. It relocates it upstream.

Why AI Loves Systems and Systems Love Elites

AI is uniquely good at working with explicit structure. It thrives when a process can be observed, described, and optimized. That is why recording a process and generating an SOP is so powerful. It turns tacit knowledge into machine readable form.

But this introduces a paradox: the more a system becomes machine readable, the more it becomes controllable from above. In organizations, this means managers can monitor, standardize, and scale operations more aggressively. In society, it means institutions that already have influence can amplify it by making the world legible in their own terms.

This is where the political analogy becomes sharp. When a small group controls the systems that define reality, they do not need to command every action directly. They only need to control the standards, the metrics, the platforms, and the interpretation layers. That is why elite domination often looks less like brute force and more like process ownership.

Think about a corporation that decides which tasks become AI assisted. The decision is not simply about efficiency. It determines whose work is valued, whose expertise is captured, and which roles are made more replaceable. The same pattern appears in public life. If policy is shaped mainly by those who can fund campaigns, influence institutions, or dominate discourse, then the system may still look participatory while becoming operationally closed.

This is the core tension: automation increases visibility, but visibility is not the same as accountability. A process can be perfectly mapped and still be governed by a narrow set of interests. In fact, mapping may make that control more effective.

Imagine a city where every traffic light, toll booth, and transit schedule is optimized by software. The system runs more smoothly, but who decided the optimization target? Faster commutes for whom? Access for which neighborhoods? What tradeoffs were baked in? The answer matters as much as the code, because code is just policy with a shorter feedback loop.

The Hidden Risk of Making Everything Efficient

Efficiency is seductive because it feels morally clean. Fewer clicks, less waste, fewer errors. Who could object? Yet the history of institutions shows that efficiency often masks a deeper value choice: efficiency for what, and for whom?

This matters because the most important systems in society are not neutral machines. They are contested spaces. The quoted discussions of elite influence, declining democratic responsiveness, and the transfer of decision making away from ordinary citizens all point to the same underlying danger: when systems become too complex or too technical, participation shrinks. People stop feeling like authors of the rules and start feeling like subjects of them.

AI can accelerate this trend if it is used only as a top down optimization engine. If a leadership team uses AI to compress the organization into cleaner outputs, they may unintentionally reduce the number of people who understand the whole system. The result is an efficient black box. It runs well until it does not, and then nobody knows where to intervene.

That is why the promise of AI should not be defined as automation alone. The more meaningful promise is institutional self understanding. When you record workflows and ask an AI to reason about them, you are not just trying to save time. You are making hidden assumptions visible. You are asking, “Why do we do it this way?” That question is dangerous in the best possible sense.

It is dangerous because it reveals whether a process exists to serve a purpose or to preserve a hierarchy. Many organizations and governments keep obsolete procedures alive because those procedures protect certain actors, not because they work well. AI can either entrench that inertia or expose it. The difference depends on who asks the questions.

The real threat is not that AI will do our work. It is that AI will make bad systems feel efficiently inevitable.

That is a profound risk. A broken process that is slow can still be challenged. A broken process that is fast begins to look natural.

The Best Use of AI Is to Expose the Rules, Not Just Execute Them

There is a more ambitious way to think about AI than as a productivity layer. Use it as an institutional mirror.

A mirror does not solve your problems. It reveals them. When you record a workflow, generate SOPs, and ask a model to recommend improvements, you are creating a mirror for an organization’s habits. But the real value is not in the automation suggestions alone. It is in the questions the mirror forces you to face:

Which steps exist because they are truly necessary?
Which steps exist because nobody has challenged them?
Which steps exist because some people benefit from ambiguity?
Which steps are so opaque that only a few insiders can navigate them?

This framework is useful far beyond business operations. It applies to public institutions, media organizations, universities, and even households. Wherever there is a process, there is an opportunity either to democratize knowledge or to centralize control.

A concrete example: a school district uses AI to streamline administrative approvals. That sounds harmless, even admirable. But if the AI is trained to privilege historical patterns, it may reinforce old inequalities in who gets resources and attention. If the district instead uses AI to surface bottlenecks, compare outcomes across neighborhoods, and reveal where approvals cluster, the same technology becomes a transparency tool.

That distinction matters. Automation is not the endpoint. Governance is.

The most valuable AI deployment may be the one that reduces dependency on gatekeepers. Not by removing human judgment, but by spreading it more widely. When more people can see how decisions are made, more people can contest them. And when more people can contest them, systems become harder to capture.

This is where the connection between workflow automation and democracy becomes unmistakable. Both are, at root, battles over legibility. If a process is only legible to elites, it will be governed by elites. If it becomes legible to many, it can be reformed by many.

A Practical Model: From Automation to Agency

Here is a simple mental model for using AI in a way that increases agency rather than merely efficiency.

1. Record the process

Capture what actually happens, not what the official chart says happens. Real power lives in the gaps between policy and practice.

2. Translate tacit knowledge into explicit rules

Ask what experts know instinctively but never write down. This is often where the hidden bottlenecks, dependencies, and privileges live.

3. Ask the model to find not only shortcuts, but chokepoints

Do not just ask where AI can save time. Ask where the process depends on a single person, a single approval, or a single interpretation.

4. Separate optimization from governance

Some steps should be made faster. Others should be made more contestable. Those are not the same goal.

5. Measure who gains discretion

Any automation project should answer one question: does this expand the number of people who can understand and shape the system, or shrink it?

This model reveals why some automation projects feel liberating while others feel quietly authoritarian. A tool that helps a team document how work gets done can spread capability. A tool that merely speeds up the execution of a bad structure can entrench it.

The same is true in politics. The health of a democracy is not measured only by the presence of elections or institutions. It is measured by whether ordinary people can still influence the rules that shape their lives. If they cannot, then the system may be operating efficiently while becoming hollow.

Key Takeaways

Treat every workflow as a power map. Ask who defined it, who can change it, and who gets left out of the decisions.
Use AI as a mirror before using it as a machine. First expose the structure of the system, then optimize it.
Beware of efficient bad systems. Speed can make broken processes harder to notice, not easier.
Optimize for legibility, not just output. A system is healthier when more people can understand how it works.
Measure agency, not only productivity. The best automation expands human discretion instead of narrowing it.

Conclusion: The Politics of Making Things Easier

The promise of AI is often described as a future in which work becomes easier. That is true, but incomplete. What becomes easier matters less than who gets to decide what “easier” means.

A company that records its procedures and asks AI to improve them is not only streamlining operations. It is deciding whether knowledge will remain concentrated or become shared. A society that lets elite institutions define its most important processes is not only losing democratic input. It is making itself easier to govern from above.

So the next time you hear about automation, ask a better question than, “What can we remove?” Ask, “What kind of authority does this create?”

That question changes everything. It turns AI from a productivity toy into a civic and organizational diagnostic. It reveals that the deepest issue is not whether machines can do our tasks. It is whether we can build systems that become more understandable, more contestable, and more humane as they become more efficient.

Because the real future of automation is not just about replacing labor. It is about deciding whether the rules of the world become visible to everyone, or only to the people already running it.