Getting Started with AI for Small Businesses: Using AI Isn't the Same as Putting It to Work
Today's Highlights: Getting started with AI for small businesses. Using it isn't the same as putting it to work
When people talk about getting started with AI, many assume the first hurdle is deciding whether to use it. But several recent US surveys show most owners have already cleared that hurdle. Where they get stuck is the next one: after adopting AI, have they actually built it into their workflows and turned it into visible results?
An analysis published by the Federal Reserve Bank of New York on Liberty Street Economics on October 8 uses AI module data from 5,248 employer firms in the 2025 Small Business Credit Survey (SBCS). A few numbers are worth comparing against your own business:
46% of firms say they or their employees are using AI tools, and another 15% plan to start within a year.
Among users, 63% consider AI at least somewhat important to operations, but only 51% have partially or fully integrated AI into their business processes.
31% report that AI has helped sales grow, but 77% say their labor costs have not changed.
Firms using AI are noticeably more optimistic about hiring and revenue growth over the next year. Even after controlling for industry, size and past performance, the gap remains 10 to 14 percentage points.
The researchers also note a detail: among AI users, firms that reported technical challenges with their website, social media, e-commerce or cybersecurity actually had higher revenue expectations. Their interpretation is that AI may be filling a specific technical gap for small firms, rather than driving growth through a lot of hiring. But the authors caution that these are expectations, and whether they come true requires follow-up.
On the same day, the Small Business & Entrepreneurship Council (SBE Council) released a mid-September survey of 516 owners of firms with 2 to 99 employees and saw a similar trend: 58% say they use AI agents regularly or occasionally, and another 14% are considering it. Among owners whose business held steady or grew, 28% cited adopting new technology such as AI as one of the reasons.
So where are they getting stuck? Bluevine's Small Business AI Trends Report from July (942 owners) offers clues: 82% of owners hit at least one barrier when trying to use AI more deeply. The most common were data security and privacy (33%), distrust of AI's accuracy (31%) and tool costs (24%). Only 22% are fully comfortable letting AI handle entry-level work without supervision. On the other hand, 48% of firms already using AI save at least four hours a week.
Put the three together and the picture is clear: the barrier to entry is already low. The real divide is whether you can go from "asking AI now and then" to "AI having a fixed place in the workflow."
Jesse's Take: Pick one stuck process first, then talk about tools
What I see most often when advising small businesses: the owner has two or three AI apps on their phone, employees use their own too, but quotes still depend on the owner's memory and customer data is scattered across LINE groups and several Excel files. Everyone "uses AI," yet the company is no easier to run. That's exactly the gap in the numbers above between using AI and putting it to work.
Suggestion 1: Start with the most stuck process, not with a tool
The New York Fed analysis offers a practical insight: AI is often most valuable where it fills your weakest spot. A common scenario: a hardware wholesaler with about a dozen staff hasn't updated its website in six months and never manages to write product descriptions for new items. Work that "nobody knows how to do and nobody has time for" is the best place to start with AI. Ask where things are most stuck first, then pick a tool.
Choosing a tool is actually the last step. For most everyday work, trying it for two weeks with accounts you already have or free versions is enough. Once you're sure it saves time and the quality is stable, then consider paying or connecting it to company systems. That wastes far less than buying a whole package at once.
Suggestion 2: Write AI into the workflow and be clear about who inputs and who checks
You don't need a big system at the start, but you do need a simple process sheet: who gives the data to AI, which fixed prompt template to use, who checks the output, and where it's saved in the end. Once it's written down, AI goes from one person's trick to the company's way of working, and it stays even when people change roles.
Suggestion 3: Set three data red lines first, then let everyone use it
Security and accuracy are the two things owners worry about most. Rather than banning AI outright, make it clear: customer personal data, unpublished quotes and financial figures don't go into free tools; anything sent externally must be reviewed by a person; and any numerical output must be checked against the original data. Simple rules are what make people willing to use AI, and keep using it.
An honest caveat: Bluevine is a fintech company and SBE Council is an advocacy group, and both surveys are self-reported by owners, so selection and self-report bias are inevitable. The New York Fed study only shows correlation; using AI doesn't guarantee growth. These numbers are reference points, not guarantees.
A first step you can take right away
This week, find half an hour to list three tasks in your company that repeat every week and take more than an hour each time, then pick the most painful one. For the next two weeks, use AI to handle it every time, and jot down how long it takes and how many revisions it needs. After two weeks you'll know whether it's worth formally writing into your workflow, and you'll have taken the real first step.
Further Reading
Related posts on JCXO
Original sources
Comments