The AI Cost Collapse Is Rewriting the Economics of Wall Street

The Cost of Intelligence Is Falling

My central economic thesis is simple: artificial intelligence is beginning to change the cost structure of financial businesses in a way that could be as important as the computerization of trading itself. The important question is not whether AI can replace every employee. It cannot. The more powerful question is what happens when one highly capable person can suddenly command the output of what once required an entire team.

That distinction matters because businesses are built around scarce human time. A financial firm can hire talented analysts, researchers, traders, engineers, and risk managers, but every additional person brings salaries, benefits, office costs, management overhead, coordination problems, and limits on how much work can actually be done in a day. AI attacks that constraint directly.

I think about this as a collapse in the price of intellectual labor. A financial operation that once required millions of dollars a year in employee costs can potentially be operated with a tiny fraction of that expense when AI agents handle research, monitoring, chart analysis, information organization, and other repetitive cognitive work.

This is why I do not view AI merely as a productivity application. Productivity software helps an employee do the same job faster. AI can change how many people are required to perform the job in the first place.

The Financial Firm Is Becoming a Human-Machine System

The most interesting organizational change is the movement from a traditional hierarchy toward what I would call a human-directed swarm. Instead of a portfolio manager managing a large group of analysts, I can envision a much smaller number of humans directing numerous specialized AI agents.

One agent can act as a research assistant. Another can examine charts. Another can challenge an investment thesis. Another can monitor markets. Another can organize incoming information. A central coordinating agent can then move information between these specialized functions.

The human does not disappear from this structure. The human moves upward.

My job becomes deciding what questions deserve to be asked, what assumptions deserve to be challenged, which opportunities matter, and where the system should look next. AI can dramatically expand the amount of work surrounding those decisions without necessarily possessing the judgment required to make the most important decision.

This creates a profound change in leverage. Historically, scaling a financial business meant adding people. Increasing the number of analysts meant hiring analysts. Increasing research coverage meant building larger teams. Extending monitoring meant adding shifts and personnel.

With AI, the scaling equation changes. The marginal cost of adding another digital worker can be dramatically lower than hiring another human worker.

The Real Productivity Story Is Bigger Than Job Replacement

I want to be careful with the phrase “AI replaces workers.” It captures only part of the economic story.

The more important possibility is that AI multiplies the output of existing workers. If one person can accomplish the work that previously required ten people, the economy does not necessarily end up with nine people doing nothing. It can end up with one person pursuing ten times as many opportunities.

That distinction is critical for investors.

If AI merely eliminates labor, the immediate economic effect is lower employment in affected occupations. But if AI substantially increases the productive capacity of businesses, it can also lower operating costs, increase profit margins, accelerate product development, create new businesses, and make previously uneconomic activities profitable.

Consider a financial research operation. If research once cost a large amount of money, management naturally had to ration it. Analysts could only investigate so many companies. They could only monitor so many markets. They could only test so many ideas.

If the cost of preliminary research falls dramatically, the economics of exploration change. I can investigate more ideas, discard bad ones faster, and spend more human attention on the few ideas that survive.

That is productivity in its most useful form: not simply doing yesterday's work faster, but making it economically sensible to do more work than was previously affordable.

The Corporate Brain Becomes a Competitive Asset

One of the most powerful ideas in this emerging model is the creation of a shared corporate memory.

Traditional organizations lose enormous amounts of information. Research sits in folders. Emails remain buried in inboxes. Analysts remember conversations differently. Old investment decisions become difficult to reconstruct. A useful insight discovered six months ago may never be connected to a new development.

An AI-driven organization can instead treat information as a connected network.

Every research report, trade, message, chart, observation, and decision can become part of an institutional memory that machines can search and connect. The result is something I think of as a corporate brain.

The economic value of this system is easy to underestimate. Information becomes more valuable when it can be connected to other information. A single data point may be unremarkable. That same data point connected to a prior investment thesis, an earlier market reaction, a corporate filing, and a historical pattern can become much more useful.

This creates a potential competitive advantage that compounds over time. The firm is not merely accumulating information; it is accumulating organized context.

This is particularly important in investment management because investment decisions are rarely isolated events. Today's opportunity may depend on something the firm learned months or years earlier. The ability to retrieve that context quickly can become part of the firm's edge.

Markets Never Sleep, and Neither Does the Machine

Financial markets create another important advantage for AI: they generate information around the clock.

Humans have biological limits. We sleep. We take vacations. We miss messages. We become tired. Attention deteriorates when we are overloaded.

A digital agent does not face the same constraints. It can monitor information continuously and respond whenever a human asks it to investigate something.

That changes the economics of market coverage. A small investment operation can potentially maintain a level of surveillance that once required a much larger organization.

More importantly, it changes the relationship between an idea and the research process. If I wake up with a hypothesis at an inconvenient hour, I do not have to assemble a team before testing the first version of that idea. I can immediately ask the system to investigate it, find counterexamples, examine relevant data, or attack the underlying assumptions.

The value is not simply speed. It is reduced friction.

Whenever the cost of asking a question falls, people can afford to ask more questions. And in investing, better questions can be more valuable than faster answers.

The Red Team Becomes More Important, Not Less

There is another financial lesson I consider especially important: AI should not simply reinforce an investment thesis.

Markets punish certainty. A trader who becomes emotionally attached to an idea can find endless evidence supporting it. This is one of the oldest dangers in investing.

An AI agent can be assigned the opposite job: attack the thesis.

Instead of asking a machine to explain why an investment should work, I can ask it to identify every major reason it might fail. I can force the system to search for contradictions, unfavorable scenarios, historical precedents, hidden assumptions, and alternative explanations.

This turns AI into an intellectual adversary rather than an echo chamber.

The distinction matters because the best investment process is not one that produces the most bullish ideas. It is one that eliminates weak ideas before capital is committed.

The New Scarcity Is Human Judgment

As AI makes research and analysis cheaper, the scarce resource moves elsewhere.

For decades, information was scarce. Then the internet made information abundant. Today, AI is beginning to make analysis itself increasingly abundant.

That means the premium shifts toward judgment.

I still need humans to decide what matters. I need someone to recognize that an apparently unrelated development could change an entire investment thesis. I need creativity to formulate questions that the machine was not instructed to ask. I need judgment to determine whether an output is insightful, meaningless, or dangerously wrong.

This is why I do not believe the immediate investment lesson is “replace humans with AI.” The more compelling lesson is “move humans toward the decisions where human judgment is most valuable.”

That is a classic economic shift: when one input becomes abundant, another input becomes relatively more valuable.

AI Could Reshape the Economics of Asset Management

The implications extend well beyond one trading operation.

Asset management has traditionally benefited from scale. Large firms can afford research departments, technology infrastructure, compliance teams, data subscriptions, global offices, and extensive operational support. Smaller firms have often struggled because they cannot spread those fixed costs across enough assets.

AI potentially attacks that advantage.

If sophisticated research, monitoring, information management, and analytical workflows become available at a much lower cost, smaller investment firms can perform functions that previously required institutional budgets.

This does not automatically destroy large financial institutions. Large firms retain advantages in capital, reputation, distribution, proprietary data, relationships, execution infrastructure, and regulatory expertise. But AI can reduce the importance of sheer headcount.

The result could be an industry with fewer people per dollar of assets managed, but dramatically greater analytical capacity per employee.

Lower Costs Can Become Higher Margins or Lower Prices

There is a broader corporate-finance principle underneath all of this: when technology reduces the cost of producing something, someone has to capture that economic benefit.

It can go to shareholders through higher margins. It can go to customers through lower prices. It can go to employees through higher compensation. It can also finance expansion and entirely new products.

Which outcome dominates depends on competition.

If every investment firm has access to roughly the same AI capabilities, then AI savings are likely to flow toward clients and competitive pricing over time. If a firm develops a unique AI workflow, proprietary data advantage, or superior human-machine decision process, it may retain more of the gains as excess profits.

This is why I view AI as a macroeconomic story as much as a technology story. It changes the relationship between capital, labor, productivity, and profits.

The Investment Implication: Watch the Cost Curve

When I evaluate the investment consequences of AI, I do not want to focus only on companies selling AI software. I want to examine every industry where the cost of knowledge work represents a meaningful share of expenses.

Financial services are an obvious example, but the same logic can spread into legal services, consulting, software development, research, customer support, accounting, advertising, operations, and corporate administration.

The winners may not always be the companies that advertise the most advanced AI. Some of the biggest beneficiaries could be ordinary businesses whose margins improve because a large portion of their back-office and knowledge-work costs fall.

That is where the second-order effects become interesting.

If a company can produce the same revenue with fewer workers, its profit margin can rise. If an entire industry experiences the same transformation, competitive pressure may eventually push prices lower. If lower prices stimulate demand, total industry output can rise even while labor requirements per unit of output fall.

That is the difference between a technological revolution and a simple cost-cutting exercise.

AI and Crypto Share a Deeper Economic Theme

There is also a broader connection between artificial intelligence and digital assets that I find useful: both technologies challenge traditional assumptions about how economic coordination is organized.

Crypto experiments with moving ownership, settlement, money, and financial transactions onto programmable networks. AI experiments with moving portions of decision-making, information processing, and organizational labor onto software systems.

Neither development eliminates the need for humans. Both potentially reduce the amount of human coordination required for certain economic activities.

That matters because coordination is expensive.

Whenever a business requires multiple people to communicate, verify information, reconcile records, authorize transactions, or maintain operational processes, there is a cost. Technology becomes economically powerful when it can reduce that friction without creating larger problems elsewhere.

For digital assets, the key question is whether decentralized networks can provide useful financial infrastructure. For AI, the key question is whether software agents can reliably perform economically valuable work. In both cases, the long-term investment question is not whether the technology is impressive. It is whether the technology creates durable economic value.

The Biggest Risk Is Not AI Failure; It Is AI Commoditization

I would also caution against assuming that productivity automatically creates extraordinary investment returns.

A technology can be enormously useful while generating disappointing returns for investors if its benefits are quickly competed away.

Suppose AI allows every financial firm to cut research costs dramatically. That is wonderful for the industry. But if every competitor has the same capability, no individual firm necessarily gains a lasting advantage.

The economics then become similar to other powerful general-purpose technologies: enormous benefits for users, but potentially intense competition among suppliers.

The durable advantage will therefore come from combinations that are difficult to copy: proprietary data, specialized workflows, strong distribution, trusted brands, unique intellectual property, superior human judgment, or organizational systems that improve through use.

The Financial System Could Become Smaller and More Powerful

My long-term picture of finance is not necessarily one dominated by enormous organizations filled with thousands of specialized employees.

I can instead imagine much smaller teams controlling extraordinarily large analytical machines.

A portfolio manager might sit at the center of a network of specialized digital workers. One continuously scans markets. Another researches companies. Another challenges investment assumptions. Another maintains institutional memory. Another examines technical patterns. Another monitors risk. A coordinating system moves information between them.

The human team becomes smaller, but the effective organization becomes larger.

This is the paradox of AI-driven business: the physical organization can shrink while the productive organization expands.

That is the economic transformation I consider most important. We are moving toward a world in which organizational scale is measured less by headcount and more by the amount of productive intelligence a small number of humans can command.

What I Am Watching as an Investor

My framework for evaluating this transition comes down to several practical questions.

• How much of a company's cost base consists of work that AI can perform or accelerate?
• Does AI reduce expenses, increase output, improve quality, or accomplish all three?
• Can competitors access the same tools, or does the company possess a genuine advantage?
• Does lower cost translate into higher margins, lower prices, faster growth, or greater market share?
• Does the business accumulate proprietary data and institutional knowledge as it deploys AI?
• Is the human organization being redesigned around AI, or is management simply adding AI tools to an old structure?
• Where does human judgment remain indispensable?
• Could the technology create entirely new demand rather than merely automate existing demand?

These questions help separate an AI story from an AI investment thesis.

The Next Economic Regime Is About Leverage

The deepest lesson I take from this shift is that AI changes leverage.

Financial leverage allows a small amount of capital to control a larger asset base. Operational leverage allows fixed costs to support much greater output. AI creates a new form of organizational leverage: a small number of humans can direct a much larger amount of cognitive work.

That can radically change what a profitable company looks like.

For decades, many businesses grew by accumulating employees. In the emerging model, the most productive companies may grow by accumulating software, computing capacity, proprietary information, and increasingly capable AI systems while keeping the human core remarkably small.

I do not see this as a simple story about machines taking jobs. I see it as a repricing of human attention.

Routine cognitive work becomes cheaper. Continuous monitoring becomes cheaper. Information processing becomes cheaper. Research becomes cheaper. Coordination becomes cheaper.

And as those things become cheaper, the economic value of creativity, judgment, original thinking, strategic direction, and trusted decision-making rises.

That is the regime shift I believe investors should understand. The ultimate advantage may not belong to the company with the most employees or even the company with the most AI. It may belong to the organization that best combines inexpensive machine intelligence with scarce human judgment.

01 Inside Bracket22, A Trading Firm Powered By Al Agents