Who Really Benefits from AI Research? (Hint: Not Small Businesses)

Every week brings a new claim about AI’s transformative power for business. Research shows gains in productivity and scale, but a closer look reveals that these results almost always come from studies of large organizations. The small businesses that employ nearly half of America’s workforce are barely represented in the evidence base. This is more than a blind spot: it means the decisions being made about AI adoption and competitiveness are grounded in data that excludes most businesses in this country.

In the United States, most of what we think of as “independent research” is actually produced through a post–World War II system where universities serve as the primary research arm of the federal government. Agencies like the NSF, NIH, and the Department of Defense provide billions in grant funding each year, and universities translate that money into projects, labs, and ultimately published papers. Large corporations supplement this model with their own sponsorships, funding endowed chairs, research centers, or industry-aligned initiatives. Even think tanks, which look like stand-alone institutions, are often financed by the same government agencies or corporate donors, operating as quasi-academic extensions rather than completely separate entities. The result is that the publicly available papers we read, whether in journals or cited in the press, reflect the priorities of the federal funding agencies and the corporations with the resources to underwrite big projects.

That brings us to the connection between research and what we read in the media or professional forums. When outlets like Futurism, the Wall Street Journal, or even posts on LinkedIn cite “studies,” those studies almost always come out of the research ecosystem just described. In other words, the way we approach AI in business, what gets measured, analyzed, and published, is largely determined by this Government–Education complex. And because the funding behind it is concentrated in federal priorities and large corporate sponsorships, the beneficiaries of that research are shaped by those same forces. The result is that the evidence base we all rely on, even when filtered through popular media, reflects the questions and interests of big institutions rather than the full spectrum of American businesses.

Just don’t take my word for it. Let’s ask ChatGPT to look at the last fifty (50) U.S.-authored papers on AI and business. When broken down, the numbers tell the story better than any opinion piece ever could. Out of those fifty papers, only two even touched on small business data, and even then, they were not truly about the competitive realities of small organizations.

Breakdown of U.S. AI & Business Research (last 50 papers)

And of those two papers tagged as “SMB,” neither offered a meaningful deep dive into how AI shapes the competitive position of American small businesses. One tracked adoption rates from Census surveys, while the other only included size breakdowns as a side note. In other words, the facts show clearly: the research agenda is tilted toward the interests of big institutions, and small businesses are barely a footnote in the conversation.

Right now, the story of AI and business in the United States is really a story of the haves and the have-nots. The research that exists, funded through federal priorities and corporate sponsorships, benefits large companies, while small businesses face a steeper barrier to entry. We can see this clearly in the UK, where government data shows that small businesses adopt AI at a much lower rate than larger firms. In the U.S., we don’t even have equivalent tracking at the same level of detail, which means the gap may actually be worse here. That’s partly a reflection of how we structure our education and research system, which leaves small businesses without a seat at the academic table. The result is that we are left with opinions rather than evidence, and opinion-driven decision-making is no substitute for real data. Until the research system itself changes, there is no clear path to closing this divide.