AI Agents Migrate From Tech Giants to Main Street — A Macro View

The Typical Tech Diffusion Model Just Broke

Technology usually trickles down. Mainframes went to banks, then mid-sized firms, then eventually everyone. PCs followed the same arc. AI was expected to run the same playbook: hyperscalers and Fortune 500 companies first, Main Street eventually.

That sequence has inverted. By mid-2025, the Federal Reserve found that small businesses were adopting AI faster than large firms — a reversal that had not happened before in the monitoring data. Enterprise adoption had plateaued while small businesses were still accelerating.

The numbers are striking. Adoption among companies with 10 to 100 employees jumped from 47% to 68% in a single year. According to an Intuit & ICIC survey, 89% of small businesses are now leveraging AI, mostly to automate repetitive tasks. The typical AI-powered small business today runs a median of five separate AI tools — not experiments, but core operational infrastructure.

This is not a gradual convergence. It is a compressed cycle that rewrites the diffusion curve we have relied on for half a century.

Why the Curve Inverted

The mechanism is straightforward: cost collapse. Tools that once required an engineering team now run on a $20/month subscription. For owners already spread thin, that changed the math.

McKinsey predicts AI agents could add $2.6 to $4.4 trillion in value annually across various business use cases. The key is that value is no longer locked behind enterprise sales cycles and six-figure implementation budgets. A bakery can now use AI agents to scrape its own spreadsheets and manage growth — the same class of workflow automation that required consultants and custom integration a few years ago.

The productivity gains are measurable. Companies that have adopted AI report 26 to 55% productivity gains in the specific functions where AI is deployed. Business owners report saving a median of five hours per week, while their employees save an average of 11.5 hours. For context, 66% of AI-using businesses report that revenue increased as a direct result of adoption.

This is not hypothetical efficiency. It is showing up in transaction-level data from JP Morgan Chase's small business banking clients.

The Competitive Dynamics Shift

What matters for investors is the second-order effect. A small retail operation competing without AI pricing, inventory, or marketing tools is now running at a structural disadvantage relative to AI-enabled competitors doing the same volume with fewer staff hours.

The perception gap is telling: 80% of SMBs that use AI believe it is now commonly used among their peers in commerce — but only a third of non-users agree with that assessment. Some small businesses may be significantly underestimating how quickly their competitors are building an advantage.

This creates a bifurcation dynamic similar to what we saw in the early e-commerce adoption phase of 2010-2015. The laggards did not disappear overnight, but the productivity gap widened until the stragglers faced a catch-up cost that exceeded their operating margins.

For sectors like retail, professional services, and healthcare, AI is already a competitive baseline. For construction, trades, and hospitality, lower adoption means first-mover opportunity remains — but the window is narrowing.

You can monitor sector-level shifts in real time via our [sector rotation dashboard](/sector), which tracks capital flows across industries as these adoption patterns mature.

What This Means for the Macro Setup

From a regime perspective, this matters for two reasons.

First, it is inherently deflationary for labor-intensive service sectors. If small businesses can deliver the same output with fewer hours, wage pressure in those segments softens even as headline unemployment remains low. This complicates the Fed's read on capacity constraints.

Second, it creates a capex-to-opex shift that does not show up cleanly in traditional investment data. A $240/year AI subscription replacing a part-time administrative role is not tracked the same way a forklift purchase is. The productivity boom may be underreported in official statistics until the measurement apparatus catches up.

The historical analog is the early 1990s, when business investment in software and IT began outpacing equipment spending but lagged in GDP accounting for years. By the time the Boskin Commission revised its methodology, the productivity gains had already repriced equity multiples in tech.

Credit conditions remain the tell. If AI adoption is genuinely expanding small business margins, we should see it reflected in default rates and credit line utilization over the next two to four quarters. Watch the regional bank earnings calls — they will have the ground-level data before the macro aggregates catch up.