Michael A. McDonnell, Founder, Apex Strategic Intelligence · Prosper, Texas
I am optimistic by choice, curious by nature, and a survivor by design. I believe the next great chapter of American entrepreneurship belongs to small and medium businesses — and that it will be written with small language model AI.
Every pivot compresses time
Every generation or two, a technology arrives that does not merely improve the world — it reorganizes it. Each time, the window of disruption gets shorter. And each time, the value moves a little further down the economic ladder.
Era | Duration | What it changed | Who captured the value |
|---|---|---|---|
The Industrial Revolution | ~100 years | Steam, steel, and scale moved humanity from agrarian labor to organized production. | Owners of capital equipment. Entry required a factory. |
The Information Age | ~40 years | Computing turned knowledge into a resource that could be stored, processed, and sold. | Corporations with mainframe budgets and technical staff. |
The Internet Age | ~20 years | Connection collapsed geography. For the first time, a small business could reach a global customer. | Platforms — and, at the margin, the small firms willing to sell through them. |
The AI Age | <5 years, and accelerating | Cognition itself becomes a purchasable input, priced by the task rather than by the salary. | Undetermined. The first pivot in which the most powerful tools are available to the smallest players on day one. |
Table 1. Four pivots. The durations shorten; the point of access descends.
Notice what the last column actually says. In every prior era, the technology arrived at the top of the economy and took decades to reach the bottom. The sole proprietor in Tulsa got the steam engine, the mainframe, and the internet late — after the incumbents had already used them to build a lead.
That sequence has broken. The same open-weight models a Fortune 500 research team downloads on a Tuesday are available to a ten-person accounting firm on the same Tuesday, at the same price, which is zero. This has never been true before. It is true now. The advantage no longer belongs to whoever gets access first. It belongs to whoever applies it first.
The economy this actually touches
It is worth being precise about the size of the population in question, because the abstraction "small business" obscures how much of the country it describes.
- There are 36.2 million small businesses in the United States — 99.9 percent of all American businesses. (SBA Office of Advocacy, 2025 Small Business Profile)
- They employ 62.3 million people, roughly 46 percent of the private-sector workforce.
- Between March 2023 and March 2024, small businesses accounted for approximately nine of every ten net new jobs created in the country.
- Roughly six million of those firms have employees. The remaining thirty million are one- and two-person operations — the plumber, the bookkeeper, the independent consultant, the shop owner.
That last line is the one that matters. The overwhelming majority of American businesses have no legal department, no marketing team, no analyst, no IT staff, and no realistic path to hiring any of them. The owner is all of those functions at once, or the business simply goes without. For a century, that has been the structural ceiling on small enterprise. It has never been a ceiling on ambition. It has been a ceiling on capacity.
The constraint on small business growth has never been ambition. It has been capacity. This is the first technology that lifts the capacity constraint without requiring capital, headcount, or a data center.
The economics of creation and destruction
Two economic ideas explain what happens next, and it is worth separating them properly, because they are routinely conflated.
Jevons: cheaper does not mean less
In 1865, William Stanley Jevons observed that improvements in the efficiency of the steam engine did not reduce Britain's coal consumption. They increased it. When a capability becomes dramatically cheaper, demand does not hold steady and free up the difference — demand expands into the new price, often by more than the efficiency gained.
Apply that to cognition. The Jevons Paradox of Intelligence holds that when the marginal cost of competent analysis, drafting, research, and review falls toward zero, businesses do not buy the same amount of thinking for less money. They buy vastly more thinking, applied to problems that were never worth a professional's hourly rate. The forecast nobody ran. The contract nobody read closely. The customer nobody followed up with. Those tasks were not unimportant — they were unaffordable. Demand for them has been suppressed by price, and it is about to be released.
Schumpeter: the destruction is the mechanism
The second idea is often misattributed to Adam Smith. Smith described how self-interested exchange coordinates an economy through price — an indispensable insight, but not this one. The relevant framework is Joseph Schumpeter's creative destruction: the observation that capitalism advances by continuously dismantling its own structures from within, and that the obsolescence of existing arrangements is not a side effect of progress but its engine.
This distinction is not academic. Smith's framework predicts equilibrium. Schumpeter's predicts turnover — that incumbent advantage is perishable and that the reallocation is where the opportunity sits. For a small business, those are entirely different forecasts. One says the field will settle. The other says the field is about to be redrawn, and the redrawing is the opening.
"Every act of creation is first an act of destruction." — Pablo Picasso
Picasso arrived at the same conclusion from the other direction. AI will obsolete categories of work; some of that will be painful, and pretending otherwise is not optimism, it is evasion. But obsolescence is the precondition for what follows, not an argument against it.
Why small models, not just AI
Most writing on this subject stops at "AI helps small businesses." That is true and nearly useless, because it ignores the reason the previous four waves of enterprise technology never reached the bottom of the market: cost structure and control.
Small language models — compact models in roughly the one-to-fifteen-billion-parameter range, distilled for narrow domains rather than trained for universal competence — solve both problems at once.
Property | Why it decides adoption |
|---|---|
Fixed cost | A model that runs on hardware you own is a one-time capital expense, not a metered bill that scales with usage. An owner who has watched a cloud invoice behave like a variable-rate mortgage will buy predictability before capability. |
Sovereignty | The model runs on the premises. Client files, patient records, case matters, and payroll never leave the building. For a firm with a professional obligation of confidentiality, this is not a preference — it is the entire condition of use. |
No implementation project | No six-month deployment, no enterprise contract, no data-science hire. A single desktop-class machine, installed in an afternoon, is the whole infrastructure for a fifty-person firm. |
Narrow beats broad | A model distilled for accounts-payable exceptions will outperform a vastly larger general model on accounts-payable exceptions, at a fraction of the cost. Most business work is narrow. Generality is a luxury the SMB does not need to buy. |
Table 2. The four properties that move small language models from interesting to installable.
This is the actual mechanism of democratization, and it is worth stating plainly: the frontier model made the capability possible; the small model makes it ownable. A capability you rent from a platform on someone else's terms is leverage for that platform. A capability that runs on a machine in your own back office is leverage for you.
What it unlocks
The functions below were, until very recently, the defining advantages of scale. Every one of them is now available to a business with fewer than ten employees.
Capability | What SLM AI unlocks for the small business |
|---|---|
Marketing | A two-person shop produces content, campaigns, and customer targeting at a standard that previously required an agency retainer. |
Operations | Scheduling, invoicing, supplier coordination, and reporting handled without a back-office team — and handled at midnight if that is when the work arrives. |
Customer service | Continuous responsiveness that once required dedicated support staff and a shift schedule. |
Financial insight | Analysis, forecasting, and scenario planning of the kind that used to arrive with a CFO's salary attached. |
Legal & compliance | Contract review, policy drafting, and regulatory monitoring, accessible without a standing retainer or the reluctance to "bother the lawyer" over a small question. |
New markets | Research, localization, and outreach into markets that would previously have required a dedicated team to even evaluate. |
Table 3. Capabilities formerly gated by headcount.
None of this is about replacing people. It is about expanding what one person, or one small team, can accomplish — which is the only definition of productivity that has ever mattered to an owner-operator.
The Act 2 entrepreneur
There is a second-order effect that almost no one is pricing in, and it may be the largest one.
The United States has an enormous population of experienced professionals between fifty and eighty years old who possess deep domain expertise, established networks, and available capital — and who, until now, faced a brutal arithmetic when considering a new venture. Starting a business meant rebuilding an entire operating apparatus from scratch: the marketing, the books, the compliance, the systems. At sixty, with a finite runway, that overhead was often disqualifying regardless of how good the idea was.
SLM AI collapses that overhead. The apparatus that used to require a team and eighteen months can now be assembled by one experienced person in a matter of weeks. Act 2 Entrepreneurship — the launch of a serious enterprise in the second half of a career — stops being a romantic exception and becomes a rational financial decision. The domain expertise that took forty years to accumulate has never been more valuable, precisely because the operational overhead that used to dilute it has never been cheaper to eliminate.
That is a demographic dividend hiding in plain sight: the most experienced cohort in the workforce, newly able to act on what it knows.
What has to be true
Optimism that cannot name its own preconditions is just enthusiasm. Three things have to happen, and none of them are automatic.
- The technology has to arrive through a channel the small business already trusts. Thirty-six million businesses will not evaluate model architectures. They will adopt what their existing IT provider, accountant, or managed service provider installs and stands behind. Capability without a delivery channel is a demo, not an economy.
- Governance has to be built in, not bolted on. A small firm cannot absorb a data breach, a hallucinated legal citation, or a regulator's question it cannot answer. Audit logging, access control, and evidence that the system did what it claims are the price of admission in any regulated vertical — which is to say, in most of the verticals worth serving.
- The value has to be verifiable. The last technology cycle taught small business owners to distrust transformation narratives. The adoption that sticks will be the adoption where the owner can point at a specific hour reclaimed or a specific error caught. Everything else is board-level enthusiasm with a shorter half-life than the contract.
These are engineering and distribution problems, not reasons for pessimism. But they are the difference between a technology that is available to small business and one that is actually adopted by it.
The resurgence
The American dream was always a story about the individual outworking, outthinking, and out-hustling the system. It thrived when barriers to entry were low and access to tools was roughly equal. It struggled whenever capital concentration tilted the field far enough that effort stopped being sufficient.
Small language model AI levels that field in a way no previous technology has — not by taking anything from large enterprises, but by handing small businesses capabilities they have never had. The plumber running a smarter scheduling and follow-up system. The independent consultant producing work indistinguishable from a global firm's. The first-generation entrepreneur with a strategic partner available at every hour of every day.
Jevons tells us demand will expand into this new capability in ways we cannot yet predict. Schumpeter tells us the reallocation is the opportunity, not the casualty. Picasso reminds us that the clearing is what makes room for the building.
I choose optimism — not because I ignore what is being disrupted, but because I can see clearly what is being built. The new American dream is not arriving despite AI. It is arriving because of it. And it belongs, more than it has in a century, to the small business owner willing to pick up the tool.
The dream isn't dead. It's being rewritten — one small business at a time.
Michael A. McDonnell is the founder of Apex Strategic Intelligence, developer of the CDI™ framework for sovereign small-model AI. Statistics: U.S. SBA Office of Advocacy, 2025 Small Business Profiles.