Small Business AI Case Studies: From Writing Copy to Running the Back Office
When people talk about small business AI case studies, the example that still comes up most is "using AI to write social posts." But a survey released this October by Goldman Sachs' 10,000 Small Businesses program shows the wind is shifting: small business owners are starting to move AI out of front-of-house copywriting and into the back office—bookkeeping, customer service, and quality control, the work customers never see.
The report has hard numbers as well as named, real-world examples, and it happens to answer a question many owners quietly wonder about: What exactly are other businesses doing with AI, and how far have they taken it?
Today's Highlights: Small Business AI Moves Into the Back Office
The report is based on a survey Babson College conducted in August 2026 of alumni businesses from Goldman Sachs' 10,000 Small Businesses program, which drew 2,473 responses. The findings: 86% of the owners surveyed already use AI, up from just 56% two years ago; 63% use it every day, and 73% spend less than $100 a month on it—a low-cost way to get started.
Even more telling is how the uses are shifting. From 2024 to 2026, the share using AI for financial analysis and operational planning rose from 18% to 42%, and for data analysis from 26% to 42%, while marketing copywriting actually fell from 74% to 65%. In other words,
AI is shifting from "write this for me" to "run the numbers and look this up for me".
A campground's customer service agent
The owner of Shell Campground in Wyoming built his own AI customer service agent that answers guests' questions by text and phone. He noticed that roughly 80% of the questions are actually simple. He also uses AI to categorize expenses and answer employees' booking questions.
A pickle maker's quality data
Detroit's McClure's Pickles layered AI on top of the company's existing records, so employees can query quality, maintenance, and recall data directly, run mock recall drills, and forecast production.
A sign company's regulatory review
A sign company in Florida uses AI to check the regulatory requirements in bid documents, cutting a review that used to take days down to minutes—while still keeping a human review step.
A 10-person clinic's billing
An unnamed clinic in Georgia has an AI agent audit its bills every night, flag denied claims, and proactively contact insurers to follow up, giving a 10-person practice the back-office capacity of a large medical group.
Who's using it is interesting too. Among businesses that use AI, 14% have only the owner using it; but at small companies with 1 to 25 employees, 28% have everyone using it—a higher share than at larger companies. For those who haven't started, the most common reason is not knowing where to apply it (45%). And whether they use AI or not, 77% of owners say they need more training.
Beyond the highlights, though, the numbers are sobering: only 15% have fully integrated AI into their core operations, 74% have no AI usage guidelines for employees, and 84% of users say their headcount hasn't changed as a result. Goldman Sachs also announced a $500 million commitment to launch a Small Business AI Institute, partnering with Anthropic, Google, Meta, and OpenAI on training. CEO David Solomon also reminded owners that there's no need to rush, since AI still isn't cheap.
Jesse's Take: Start With the Question You Get Asked Most
After going through these cases, here's what I most want to point out: none of them started with a big slogan like "adopting AI." Instead, each one started with one small task that repeats every day and follows clear rules.
Tip 1: Start where "80% of the questions are simple"
That observation from the campground owner is key. Your business has these questions too: business hours, shipping fees, returns and exchanges, how quotes work. The answers are usually scattered across LINE groups (LINE is the messaging app most Taiwanese businesses run on), Excel files, and the owner's head.Start by compiling them into a single Q&A document—whichever AI tool you end up using, this document is your foundation.
Tip 2: Let AI read your data, not just write for you
What the pickle maker and the clinic have in common is that AI is connected to their own records. A common scenario: a small contract food manufacturer keeps its inspection records spread across several Excel files, and when a customer asks for the test results on a particular batch, a colleague spends ages digging. Standardizing the file formats and gathering them in one folder is already the step right before having AI help you look things up. Once the data is organized, asking AI questions like "Which batch had the most returns last month?" becomes much easier.
Tip 3: Keep a human in the loop, and write a one-page policy first
The sign company emphasizes having someone review the work, and that matters. In the report, 74% of businesses have no AI guidelines, and I'd guess most small companies in Taiwan are about the same. Which data must never be pasted into AI, and which outputs need a human check before they go out—one page is enough to write that down.
I should also be upfront: these cases come from alumni of Goldman Sachs' own training program, the response rate was only 16.5%, and the report itself notes the results may not represent everyone. The businesses willing to answer a survey, and the ones picked to be showcased, are usually the ones doing better, so take the numbers with a grain of salt.
A First Step You Can Take Today
Take a sheet of paper and write down the five questions customers or employees asked you most often in the past week, then note where you usually go to find the answers. Those five questions are your company's first AI use case.
Further Reading
Related posts on JCXO
Original sources