.

But it does not make sense in the AI economy that is rapidly emerging.

We are entering a world where machines perform an increasing share of economic work. Not just factory labor, but cognitive work as well. Artificial intelligence is already assisting with research, accounting, coding, customer service, design, legal analysis, and even medical diagnostics. In many industries, AI systems are becoming productivity engines capable of performing tasks once done by large teams of people.

Yet our tax system still assumes that human labor is the primary engine of economic activity. The result is a strange contradiction. We tax the very thing we claim to value most: work.

Meanwhile, the productivity explosion created by artificial intelligence is flowing overwhelmingly to a relatively small number of companies that own the infrastructure of the AI economy. The firms building and operating large scale AI models are capturing enormous economic value while the tax system remains tied primarily to human wages.

That imbalance is not sustainable.

For years Andrew Yang has argued that we should rethink this model. His proposal is both simple and profound. Instead of relying almost entirely on taxes on human labor, we should begin taxing the productivity created by AI systems and automation.

The revenue from that productivity could help fund a universal basic income that ensures every citizen benefits from the technological revolution unfolding around us.

Some people dismiss this idea as radical. I see it as common sense.

Consider how the state of Alaska handles oil extraction. Oil companies generate enormous wealth by extracting a natural resource that belongs to the public. Through the Alaska Permanent Fund, a portion of that revenue is returned to the people of the state. Every resident receives a dividend because the resource itself is part of the public commons.

Artificial intelligence is increasingly similar. The foundations of modern AI were built through decades of publicly funded research, government grants, university laboratories, and taxpayer supported infrastructure. The internet, advanced computing, and much of the data used to train modern models all emerged from public investment.

The benefits of this revolution should not flow exclusively to a handful of technology companies. They should flow to society.

Let me be clear about something important. This conversation is not about punishing innovation. Quite the opposite. I am one of the strongest supporters of technological progress you will meet. Much of my career has been spent working at the intersection of energy systems, climate technology, and emerging innovations. Technology is how we solve many of the greatest challenges humanity faces.

But if artificial intelligence drives extraordinary productivity while simultaneously replacing large segments of the workforce, we cannot continue funding society primarily through taxes on wages.

That system simply stops working.

Imagine a company that replaces ten thousand workers with AI systems. Those workers once paid income taxes and payroll taxes that funded schools, infrastructure, healthcare programs, and public services. Once those jobs disappear, the tax revenue disappears with them. Yet the company may now be producing even greater economic value than before through automation.

This raises an obvious question.

In an automated economy, where should the tax base come from?

The logical answer is the productivity itself.

There are several ways this could be designed. Governments could place modest taxes on large scale AI compute. They could tax profits generated primarily through automated systems. They could introduce automation replacement taxes that mirror payroll taxes when machines substitute for human labor. Another approach would be an AI dividend model that captures a portion of productivity gains and distributes it back to citizens.

Different economists propose different mechanisms. But the underlying idea remains the same. As artificial intelligence becomes one of the most powerful economic engines in human history, a portion of that value should be shared by the society that made it possible.

Universal basic income funded by AI productivity is not about eliminating work. It is about preserving freedom.

Freedom to start new businesses without the fear of immediate financial collapse. Freedom to care for children, aging parents, and communities. Freedom to pursue education, creativity, entrepreneurship, and civic engagement. Freedom to take risks and build things that matter.

In a world where machines can produce extraordinary wealth, the question is not whether abundance is possible. The real question is who benefits from that abundance.

We are entering a moment where the foundations of the economic system itself must be reconsidered. The rules that governed the industrial age were built for a world in which human labor was the central driver of economic value.

Artificial intelligence is changing that equation.

Taxing labor made sense in the industrial era.

In the age of artificial intelligence, it may be time to start taxing the machines instead.

It is good. It is helpful. It is polite. But it does not quite sound like us yet.

The answers are correct, but they feel generic. The tone is close, but not fully aligned. It gets the job done, but it does not yet carry the instincts, nuance, or personality your customers associate with your brand.

That moment is where trust is either built or broken. So let me be very clear about something upfront. This is not a failure of the system. It is the starting line.

What Your AI Is Actually Doing on Day One

When an AI agent goes live for your business, it begins from a neutral baseline. It has been trained to be broadly competent across millions of scenarios, industries, and communication styles.

That means its first priority is safety, clarity, and correctness. Not brand voice. Not institutional knowledge. Not the unwritten rules your team follows instinctively.

Early on, your agent is speaking to your customers the same way it would speak to anyone.

That is by design.

The model is optimized to serve the average user until it has enough signal to do otherwise. And this is the part most vendors never explain.

Why It Feels Generic at First

AI systems are trained using large scale human feedback. Thousands of people rate responses and choose which ones they prefer. Over time, the system learns what tends to be selected most often.

What wins in that environment is not originality or specificity. What wins is broad acceptability.

Clear.

Neutral.

Inoffensive.

Generally useful.

So when your agent is new, it defaults to that statistical center. A voice designed for everyone and no one. Your customers are not hearing your company yet. They are hearing the starting template. That is not the endpoint. That is the raw material.

Optimization Is Not Tuning It Is Training

Here is where patience becomes leverage. Every real conversation your AI agent has creates signal. When a customer asks a question and the answer lands well, we keep it.

When it misses the mark, we correct it.

When it sounds too stiff, we soften it.

When it over explains, we tighten it.

When it lacks confidence, we add authority.

Those corrections do not disappear. They accumulate.

Unlike a human hire who forgets yesterday’s coaching under pressure, an AI agent compounds learning. Once something is corrected and encoded, it does not regress. Over weeks, not years, the agent begins to reflect your priorities, your tone, your values, and your way of doing business. This is when clients usually say the same thing.

It finally sounds like us.

Why This Takes Time And Why That Is a Feature

You would never expect a new employee to sound like a seasoned team member in their first week. They need exposure. They need feedback. They need context. AI agents are no different, except they learn faster and forget nothing.

What you are really doing in the early phase is teaching the system how your organization thinks. How formal you are. How direct you are. How you handle objections. What you never say. What you always say.

That kind of knowledge does not come from prompts alone. It comes from lived interaction.

The Payoff Most Businesses Miss

Here is the part that matters.

Once optimized, your AI agent does not just answer questions. It scales your institutional voice. Every customer gets the best version of your messaging. Every interaction reflects your standards. Every response is consistent even at volume.

And unlike traditional automation, it keeps getting better instead of brittle.

The businesses that win with AI are not the ones rushing to perfection on day one. They are the ones who understand that early generic output is the price of long term precision.

What We Ask From Clients

We do not ask for blind patience.

We ask for informed patience. Engage with the agent. Review transcripts. Flag what feels off. Celebrate what lands well. That feedback loop is where the real value is created. Because the goal is not to deploy an AI that sounds like a machine. The goal is to deploy one that sounds like your best people on their best day, every day, at scale.

And that is worth doing right.

AI is now being cast as the villain in an energy crisis that existed long before AI and will not be solved by slowing it down. We are told data centers will overwhelm the grid, that artificial intelligence will erase climate progress, that efficiency gains inevitably trigger runaway demand. It sounds plausible. It is also incomplete.

The real issue is not how much electricity AI uses. It is how poorly our energy systems are designed. Countries that have modernized their grids understand this clearly. Countries that have not often prefer to blur the distinction.

AI does consume electricity. Training models and running data centers requires power. But framing this as an existential problem ignores where much of that energy goes today. Vast amounts are lost to inefficiency. Aging data centers waste heat. Redundant workloads run nonstop. Cooling systems lag decades behind current capability. Transmission drains power before it ever reaches demand. Grids built without flexibility struggle under even modest change.

What is often labeled an AI problem is, in reality, a system problem. And that distinction changes everything about how we respond.

What receives far less attention is that AI is already reducing energy use across the economy. It balances grids in real time. It optimizes industrial processes. It cuts waste in buildings and logistics. These gains rarely make headlines because they are invisible by nature. They show up as energy never consumed, money never spent, emissions that never occurred.

Electricity demand is not the enemy. Waste is. And electricity used intelligently in a clean system is not the same thing as electricity burned inside a fossil heavy, centralized one.

Denmark is frequently dismissed in American energy debates. Too small. Too simple. Too different. But that dismissal misses the point. In 2000, nearly ninety percent of Denmark’s electricity came from fossil fuels. Today it is under ten percent. That transformation happened in under a generation, not through miracles or moonshots, but through sustained choices.

Denmark invested steadily, peaking around one to one and a half percent of GDP during its transition years. It modernized the grid alongside generation rather than treating transmission as an afterthought. It prioritized efficiency before scaling new supply. It embraced decentralization rather than resisting it. Most importantly, it rewired incentives so utilities were rewarded for stability and efficiency instead of sheer throughput.

That incentive shift allowed progress to compound instead of stall.

“Denmark moved from nearly 90 percent fossil based electricity in 2000 to under 10 percent today by investing roughly 1 to 1.5 percent of GDP at peak years, modernizing the grid alongside generation, and rewarding efficiency over volume. The lesson is not that Denmark is small. It is that incentives decide speed.”

The United States is much larger than Denmark, and that fact is often used to shut down the conversation. It should not. Scale does not inherently make transition harder. When designed well, it makes it more powerful. The US has some of the strongest renewable resources on Earth, vast geographic diversity, a deep manufacturing base, technical expertise, and access to capital that few nations can match.

What it lacks is coordination and the political will to unwind rules designed for a different century. Every year the US spends tens of billions of dollars in direct and indirect fossil fuel subsidies, before even counting health costs, disaster recovery, and military expenditures tied to energy security. Redirecting even a fraction of that spending toward grid modernization would change the economics almost immediately.

The barrier is not affordability. It is who benefits when delay becomes policy.

When people imagine a modern grid, they often picture something futuristic or abstract. In practice, it is deeply practical. A modern grid means decentralized generation closer to demand. Microgrids that can island during failures. Storage located where it matters. AI managing load dynamically rather than humans reacting after systems break.

It means predicting peaks instead of overbuilding for rare worst case scenarios. It means siting data centers near clean generation and treating them as flexible assets instead of rigid liabilities. In that system, AI does not strain the grid. It stabilizes it. Fragility gives way to redundancy. Single points of failure give way to adaptive networks.

This is not speculative. It is already happening quietly in parts of the world. Just not fast enough.

In fossil heavy centralized systems, AI looks dangerous because it exposes inefficiency faster than institutions can hide it. In clean decentralized systems, AI becomes a force multiplier. It enables higher renewable penetration without instability, reduces reserve margins, detects faults early, and ensures infrastructure investment delivers outcomes rather than optics.

The deeper shift is philosophical. Legacy systems were built around averages and rigid assumptions. Modern systems are built around adaptability. That shift alone unlocks enormous savings in cost, energy, and resilience.

Strip away the rhetoric and the real conflict is not AI versus climate goals. It is centralized extraction versus distributed value creation. Short term guarantees versus long term system health. Regulatory capture versus public benefit. AI simply accelerates whichever model it is paired with.

Technology is neutral. Incentives are decisive.

Fear driven narratives about AI and energy serve very specific interests. They slow reform, protect incumbents, and distract from governance failure. A more honest story does not deny challenges. It puts them in context and focuses attention on where leverage actually exists.

The future is not about choosing between AI and sustainability. It is about designing systems where they reinforce each other. That future is not theoretical. It is already being built.

The only question left is whether we scale it intentionally or allow the past to keep vetoing the present until it collapses under its own weight.

History is very clear about which path costs more.