American businesses are investing heavily in artificial intelligence, purchasing software, automating administrative work and experimenting with technology that promises to transform productivity.
Yet a growing question hangs over the investment boom: How much of that spending is producing measurable financial benefits?
Technology companies supplying AI infrastructure, cloud computing and software have reported substantial demand. Nvidia, Microsoft and other major technology firms have benefited from the rapid expansion of AI-related investment.
But strong revenues for companies selling artificial intelligence do not necessarily mean that the businesses buying it are experiencing comparable financial returns.
For small businesses, nonprofits and community institutions operating with limited budgets, that distinction could determine whether AI becomes a valuable operating tool or another recurring expense.
The Productivity Question
Research has documented both AI’s potential and the difficulty of translating that potential into organization-wide financial results.
A 2023 study by researchers at Stanford University and MIT examined more than 5,000 customer-support workers using a generative AI assistant. Productivity increased by approximately 14% on average, with particularly strong improvements among less-experienced workers.
The findings demonstrated that AI can improve performance in specific workplace activities.
However, improvements in individual tasks do not automatically produce equivalent increases in organizational profits.
An employee who completes a report in 20 minutes instead of an hour may become more productive. But unless the organization uses the additional time effectively, reduces costs or increases output, the financial benefit may remain difficult to identify.
That distinction is especially important for organizations purchasing multiple AI subscriptions without establishing clear performance goals.
Why Smaller Organizations Face Different Challenges
Large corporations can dedicate significant resources to testing technology, redesigning workflows and measuring results.
Smaller organizations frequently lack those capabilities.
A neighborhood business might adopt AI to handle customer inquiries, develop marketing materials or manage appointments. A nonprofit might use it to assist with grant research, reporting or administrative tasks.
These applications can be useful. But the economic value depends on what actually changes.
Does the business serve more customers? Are employees spending fewer paid hours on repetitive work? Has revenue increased? Have administrative expenses declined?
For nonprofits, the calculation can be more complicated. Technology may improve service delivery without immediately reducing spending. In those cases, measures such as processing time, service capacity, error rates and cost per client served may provide a clearer picture.
The Difference Between Adoption and Return
A 2025 MIT research initiative report drew widespread attention by finding that most organizations in its sample had not achieved measurable financial returns from their generative AI initiatives.
The report’s methodology and definition of success require careful interpretation. Its findings should not be treated as proof that 95% of all AI investments fail.
Still, the research highlighted an important challenge: Purchasing advanced technology is easier than integrating it into everyday operations.
Independent economic research has also suggested that AI’s productivity effects may take time to appear across the broader economy, particularly when businesses must first change how work is organized.
A Financial Test for AI
Organizations considering AI investments can begin with a straightforward assessment.
Identify the cost of the technology, the work it is intended to improve and the financial or operational results expected within a defined period.
Then measure the results against the original baseline.
Labor hours saved, administrative costs reduced, additional revenue, fewer errors and improved service delivery are more useful indicators than the number of AI tools purchased.
The calculation should also include implementation costs, employee training, oversight and the expense of correcting mistakes.
Artificial intelligence may ultimately become one of the most consequential productivity technologies in decades.
But its economic success will not be determined solely by the billions invested in data centers, software and semiconductor chips.
For the organizations paying to use it, the more important question is whether the technology improves financial performance, strengthens operations and creates value that can actually be measured.