AI Adoption Not Even Across the Board

Reality Check: AI Adoption Remains Limited Across U.S. Businesses

While headlines often suggest that AI is revolutionizing every facet of the business world, recent research reveals a more restrained reality. According to a study from the National Bureau of Economic Research, AI adoption in the U.S. is largely concentrated in a few industries and major companies, especially in sectors like manufacturing and healthcare.

A Closer Look at AI Adoption

Despite AI’s transformative potential, the study shows significant variation in its adoption. Kristina McElheran, an MIT Initiative on the Digital Economy visiting scholar and the study’s lead author, reports that only 6% of U.S. businesses used AI in 2017, with adoption mostly clustered within large corporations and industries with high-tech demands, such as manufacturing and information technology. Notably, AI usage is more prominent in a few “superstar” cities like San Francisco, San Antonio, and Nashville.

“The narrative is that AI is omnipresent, but the data suggests it’s more challenging to implement than many assume,” said McElheran, highlighting the uneven arrival of the digital age.

AI: A Big Business and Sector-Specific Phenomenon

Traditional measures of AI adoption often rely on indirect metrics like patents or job descriptions referencing AI, but this study approached it differently. Collaborating with the U.S. Census Bureau, the researchers developed the Annual Business Survey to measure direct use of AI across 447,000 firms, representing over 4 million businesses nationwide. This research defines AI adoption as using AI in actual production—rather than in experimental, conceptual, or marketing capacities.

In 2017, only 6% of companies reported using AI—a figure that remains relevant today, with a recent Census Bureau survey indicating fewer than 4% of companies are currently deploying AI in production.

Key findings include:

Large companies lead in AI usage: More than half of companies with over 5,000 employees use AI, with this figure climbing above 60% for firms with more than 10,000 employees.
Sector disparities: While about 12% of firms in manufacturing, information services, and healthcare use AI, adoption is much lower in sectors like construction and retail.
Geographic clusters: AI use is concentrated not only in tech hubs but also in some unexpected areas, like certain Midwest manufacturing centers and Southern cities.

The Role of Younger Leaders in AI-Driven Startups

Exploring characteristics that contribute to AI adoption, the study analyzed 75,000 startups participating in the Annual Business Survey. Findings indicate that startups using AI often have younger, highly educated leaders and are more likely to attract venture capital and focus on process innovation.

“Companies with other competitive advantages are better positioned to adopt advanced technologies like AI,” McElheran noted, adding that reconfiguring workflows and production processes is a key factor in AI adoption.

Unlike comprehensive software transformations, AI is more akin to a targeted solution that enhances specific tasks. However, as McElheran warns, introducing AI to one part of a system can create a need for additional innovation across the entire process. For companies without process innovation capabilities, adopting AI can be more disruptive than beneficial.

AI Adoption Across Unlikely Sectors

AI isn’t just confined to tech-driven industries. For example, manufacturing’s longstanding use of robotics has paved the way for AI-driven process improvements, while healthcare organizations have deployed AI to streamline operations, from scheduling surgeries to automating billing tasks.

The Challenges of AI Adoption: Inertia and Adjustment Costs

McElheran emphasizes that adopting AI can be a slow process due to built-in routines and adjustment costs, similar to the adoption challenges faced with previous technologies like the internet and word processors. These costs, while often unavoidable, can lead to economic shifts, including job losses in certain sectors and skill gaps among workers over age 50 who may miss out on the same pay increases seen by younger, tech-savvy colleagues.

Reflecting on AI’s potential and its challenges, McElheran calls for a balanced approach: “To fully harness AI’s benefits, we need a realistic, evidence-based approach that accounts for both the advantages and the societal costs associated with adoption.”