For Small Businesses, AI’s Real Test Is Whether It Helps With the Work

By Srithan Gade

The AI-generated images contained spelling errors and low-quality graphics. However, the small-business leader who explained these issues did not give up on the technology. On the contrary, the business tested different models in order to achieve better results.

When asked what determines the usefulness of a technology, the leader gave a simple criterion: “It does not have to be perfect, but it has to provide a benefit.”

That is the difference. Using an artificial intelligence tool and making sure it fits into the workflow are two separate decisions.

As part of the exploratory study guided by Dr. Murugan Anandarajan of Drexel University, I analyzed survey responses of 155 small-business respondents from 15 states and conducted six interviews with seven small-business leaders. My research aimed to investigate AI use among small businesses, value perception, and reduction/stopping of specific tools/uses.

The results suggest that, in addition to asking whether a business uses AI, it is useful to ask what it does with it, how often, and whether the outcome is useful enough to repeat the experience.

Usefulness doesn’t always mean revolution

Out of 155 survey respondents, 104 said that their organization uses AI. Most common use cases included sales or customer communication, document generation and structuring, researching and spreadsheet analysis. Of the 104 AI users, 67 estimated that they saved at least one hour weekly. These were participants’ self-reported data, not independently measured productivity improvements.

One interview participant used ChatGPT to create a job description for a manager of a personal-training business.

“For the simple things I have used it for lately, I like it. I am going to continue using it,” the participant stated.

That was a simple use case but a very revealing one. The participant found a particular job, used the tool, and got a useful result. It doesn’t demonstrate any revolutionary business transformations. It shows the practical reason to reuse the tool.

In the survey, there was a clear difference between rare and frequent users. Out of 67 respondents using AI at least several times per week, none selected “no clear benefit” or reported reducing or stopping an AI tool or use case. Out of nine rare users, eight selected “no clear benefit,” and three reported reducing or stopping an AI tool or use case. 28 respondents used AI several times per month.

These results do not show if frequent use led to value generation or vice versa. Also, reducing use of one tool does not mean abandoning AI completely. What these results show is how much the “AI user” label can leave out.

Continuing to use AI also does not mean an absence of complaints. Inaccurate outputs and lack of employee training were the most frequent obstacles. Four respondents who reduced or stopped the use of some tool or use case mentioned inaccurate outputs, subscription price, or privacy concerns.

The practical question therefore is not whether the tool makes mistakes sometimes. The question is whether the result is good enough to be useful for the particular task, with checking and correcting the output.

Non-use is not always a sign of resistance

The results suggest caution about assuming that businesses which do not use AI simply lack motivation.

Out of 51 respondents who do not use AI or were unsure whether their organization used AI, poor fit with their workflow and satisfaction with existing solutions were the most common barriers, chosen by 13 respondents each. The clear examples of the business value of the technology were the most frequently selected incentives to try AI, followed by its integration with the existing software and specialist assistance.

These results suggest that a different discussion could start from the question “What problem should be solved?” and “What already works?”, instead of trying to promote the technology to all businesses.

A demonstration of how AI can help solve the particular problem of the business may be more valuable than just stating that every business needs AI. The study does not show what kind of help would be the most effective, but it shows what respondents said might encourage them to try AI.

Even the understanding of the AI itself is not universal

Before seeing the examples, 21 respondents mentioned the AI components in the business software in their definitions of the term. In the next multiple choice question, 47 respondents chose AI features integrated in the business software. At the same time, 24 respondents classified regular fixed-rule automation as AI.

These were two different question types. This contrast does not prove that examples increased awareness of the technology, or that respondents who do not use AI were actually using it without realizing that. This contrast just shows that asking people to define AI themselves gives different results than showing them examples of it.

That is important for interpreting adoption statistics. Before counting AI users, it is important to know what exactly the respondents think about the AI.

This study has several limitations. Even though respondents come from 15 states, 105 of them are from New Jersey and the sample is not nationally representative. Responses were collected once from each respondent, and the rare-use and reduced-use groups were small, limiting the conclusions drawn from comparisons. The results describe respondents’ reported experiences; they do not establish cause-and-effect relationships or track changes over time.

For small businesses, the results of this study imply that the evaluation based only on adoption rate of AI may be insufficient. The evaluation based on usefulness, output quality, and time saved on particular tasks may provide more informative data.

This evaluation could start with one task and several simple questions: Is the output usable? How many corrections are needed? Does it really save time compared to the previous solution?

The goal does not have to be AI usage everywhere. A more useful goal is to understand where it deserves to be used.

Srithan Gade is a senior at Lenape High School in New Jersey and co-founder of VenoxAI, which provides websites, automation, and other digital services to businesses and nonprofits. He has previously provided website services to Reporte Hispano.