Before You Buy Another AI Tool, Give It One Useful Job

By Askia Jones · About a 10-minute read

A practical guide for small-business owners who want AI to save real time without putting their reputation, their customers’ trust or their standards at risk.

Every week brings a new AI app promising to answer your phones, write your marketing and “run your business while you sleep.” Most owners don’t need another app. They need one recurring task off their plate, done well enough to put their name on.

So for a small-business owner, the useful question about artificial intelligence is not “How do I keep up?” It is “What task could I make easier without lowering my standards?”

Start there. Not with another subscription. Not with a promise to automate your entire business. Choose one recurring task, give an AI assistant clear instructions, and check whether the result actually helps.

Below: how to pick that task, five places to start, the rules worth knowing and a scorecard to judge the results. In every case, a person stays responsible for the final work.

You are not as far behind as the ads suggest

It’s easy to feel like everyone else has already figured this out. The numbers say otherwise.

• The government’s count. In the U.S. Census Bureau’s most recent biweekly business survey, collected October 19 to November 1, 2026, 24.2% of U.S. businesses said they had used AI in any business function during the previous two weeks. Another 13.9% said they didn’t know. [1]

• A business-group survey. The U.S. Chamber of Commerce surveyed 3,870 small businesses online in June 2025. That year, 58% said they use generative AI, up from 40% in 2024. [3]

Why the gap? The Census asks a nationally representative sample [2] about a specific two-week window; the Chamber asked whether owners use generative AI at all. Both can be true: plenty of owners have tried a chatbot, and far fewer use AI as a regular part of the business.

What about Black-owned businesses? The Federal Reserve’s Small Business Credit Survey, fielded September to November 2025 among small employer firms, found that 35% of Black-owned firms were using AI, compared with 48% of white-owned firms. But another 31% of Black-owned firms planned to start within the next 12 months. That was the highest share of any group shown. [4] Among Black-owned firms already using AI, 46% said they were still experimenting; 8% called AI fully integrated. [4]

Among Black-owned firms using AI, 88% used it for writing or marketing, 73% for individual productivity and 45% for customer service. Their top challenges were accuracy (36%) and adapting tools to the business’s needs (36%), then cost (28%). Among Black-owned firms with no plans to use AI, 25% pointed to finding the right tools and 24% to cost. [4]

Interest isn’t the problem. The hard part is choosing the right first job and checking the work, and neither requires buying anything.

A note on vocabulary: after a September 2026 executive order, federal agencies, including the Census Bureau and NIST, began calling AI “super intelligence,” or “SI.”1,5 Same tools, new label.

Pick your first job

A good first job for AI passes most of these tests:

• It repeats. You do it every week, or you answer the same question again and again.

• You already know what “good” looks like. You could spot a wrong answer in seconds.

• Mistakes are cheap and easy to catch. You will review the result before any customer sees it.

• It’s mostly words or organizing. Drafting, summarizing, sorting or reformatting.

• The inputs are safe to share. You’d be comfortable if what you paste in became public.

Hold off on anything that moves money, makes legal, tax, medical or safety decisions, touches private customer details, or goes out without a person reading it first.

If a task fails two or more tests, pick another. You can come back to it later.

Five useful first jobs

1. Turn repeated questions into a customer FAQ

Look through the questions customers ask most often. They might concern booking, delivery, service areas or what happens after a purchase.

Give the assistant your actual policies, in the words you already use on invoices or your booking page, and ask it to organize them into short answers. Do not let it invent policies to fill the gaps.

Try this prompt: Turn the business policies below into a customer FAQ. Use plain language and a welcoming tone. Use only the information I provide. If an answer is missing, mark it “Owner needs to confirm.” Do not invent prices, guarantees, delivery times or refund terms.

Before publishing, check every answer against how your business actually operates. A polished answer is still a problem if it promises something you cannot deliver.

That warning matters because AI tools can state false things with total confidence. The National Institute of Standards and Technology (NIST) calls this “confabulation” and describes it as a natural result of how generative models are designed. [6] Clear instructions reduce the risk. Your review catches the rest.

Before and after: an FAQ for a catering business

Illustrative example. The business, its policies and both drafts below are fictional, written for this article to show a common pattern. They are not output copied from any specific tool.

The vague prompt: “Write an FAQ for my catering business.”

A typical draft:

How far ahead should I book? We recommend booking at least two weeks in advance, but we can often accommodate last-minute requests!

Do you offer refunds? Absolutely. Your satisfaction is our top priority, so we offer full refunds on cancellations.

It sounds friendly. But this owner never set a two-week window, doesn’t take last-minute orders and keeps the deposit when a client cancels late. Each sentence is a promise the business never made.

The specific prompt above, plus real policies: “Orders need 10 days’ notice. A 50% deposit holds the date and is not refundable within 7 days of the event. We serve events within 25 miles of downtown.”

The better draft:

How far ahead do I need to order? We need at least 10 days’ notice for every order.

What happens if I cancel? A 50% deposit holds your date. If you cancel within 7 days of your event, the deposit is not refundable. Owner needs to confirm: what happens to the deposit if a client cancels more than 7 days out?

Do you deliver? We serve events within 25 miles of downtown. Owner needs to confirm: delivery fees, if any.

Less exciting, far more useful, and it shows the owner exactly which policies still need a decision.

2. Build a first draft of a service description

You know what you do. Putting it into words can be harder, especially when customers are unfamiliar with your industry.

Start with the essentials: who the service is for, what is included and what someone should do next.

Try this prompt: Write a clear service description using these details. Explain who it is for, what is included and the next step. Avoid hype, unsupported claims and industry jargon. Ask me about any missing information before drafting.

That last line turns the assistant into an interviewer, so you fill the gaps instead of letting it guess.

Compare the result with your existing website or business listing. Does it help someone decide whether you are the right fit? Remove anything that sounds impressive but says little, especially phrases like “guaranteed results” or “the best in the city.” If you can’t back a claim up, don’t publish it.

3. Write a follow-up without losing your voice

Following up after an inquiry or completed job can slip down the task list. An assistant can help draft the message, but the relationship should remain yours.

Try this prompt: Draft a short follow-up to someone who asked about our services. Thank them, restate the next step I provide and leave room for questions. Keep it friendly and low-pressure. Do not imply that they have agreed to buy anything.

Use placeholders such as [CUSTOMER NAME] rather than pasting customer names, contact details or private conversations. Read the message before sending, and adjust it until it sounds like something you would actually say. Then save it as your own template, so next time you start from your words, not a blank prompt.

4. Turn one real update into several useful posts

You do not need to manufacture a new announcement every day. A real update, such as a new service, an upcoming event or a product restock, can support several different messages.

Try this prompt: Using only this business update, draft three posts: one announcement, one answer to a likely customer question and one reminder. Give each a different purpose. Do not invent customer quotes, results, discounts or urgency.

Check that dates, availability and links are correct. Use photos you own or have permission to publish. If the update only needs one post, publish one; more content is not automatically more useful.

Want social proof? Ask real customers for real reviews, and never reward only the positive ones. That’s federal law, as the next section explains.

5. Turn your notes into a repeatable checklist

Think about a task you explain repeatedly: preparing an order, opening the shop or getting ready for a client appointment.

Write down how you currently do it, then ask the assistant to organize those notes.

Try this prompt: Turn these notes into a step-by-step checklist. Preserve my process. Flag unclear instructions and missing steps instead of guessing. Separate required actions from optional suggestions.

Test the checklist during the actual task. For legal, financial, safety or other regulated work, use qualified guidance, not an AI-generated checklist, as the authority.

Keep your marketing honest: two federal rules of the road

AI makes it easy to produce words that sound like proof. Two Federal Trade Commission (FTC) guardrails apply to every small business.

Fake reviews and testimonials are banned. The FTC’s Rule on the Use of Consumer Reviews and Testimonials took effect October 21, 2024. [7] It bars businesses from writing, creating or selling reviews or testimonials that misrepresent whether the reviewer exists, whether they used the product or service, or what their experience was. It also bars incentives tied to a review’s sentiment, positive or negative, and requires owners or managers who review their own business to clearly disclose that relationship. [8] In plain terms: an AI-written “testimonial” from a customer who doesn’t exist is a fake review.

Claims about AI need proof, too. The FTC has warned businesses not to exaggerate what an AI product can do and not to claim that AI makes a product better without adequate proof. It also notes that “merely using an AI tool in the development process is not the same as a product having AI in it.”9 If you used a chatbot to help write your website, that does not make your service “AI-powered.”

This is general information, not legal advice; when in doubt, ask a qualified professional before publishing.

Keep a firm boundary around business information

Before using any tool, review its current privacy and data-use settings. Do not assume your inputs stay private.

NIST’s November 2026 cybersecurity guide for one-person businesses recommends checking an AI tool’s settings for data retention, training use, access privileges and confidentiality terms before entering anything sensitive, and reviewing all AI-generated content before you use it. [5] In practice, answer three questions:

1. Are my conversations used to train or improve the system, and can I turn that off?

2. How long are they kept, and can I delete them?

3. Who else can see them: teammates, connected apps or the company itself?

Then keep these out of any AI tool: passwords and recovery codes; card and bank numbers; Social Security and other ID numbers; customer names paired with contact details, order history or health information; and confidential contracts or business plans. Use placeholders, and test with public business information and fictional examples.

Protect the account itself, too. NIST recommends phishing-resistant multi-factor authentication wherever it’s offered and strong, unique passphrases of at least 15 characters, or passkeys. It also warns that AI now helps scammers write more convincing phishing messages, so take a second look at any message asking you to click, log in, send money or share sensitive information. [5]

For free, plain-language help, start with NIST’s Small Business Cybersecurity Corner [10] and the SBA’s cybersecurity page, which links to free planning tools from the FCC and free services from the Cybersecurity and Infrastructure Security Agency (CISA).11

Measure the result, not the excitement

Choose one task this week. Note how long it normally takes, then try the assisted version, including the time you spend prompting and correcting. Run it at least three times before deciding; the first attempt is mostly you learning to give instructions.

Use a simple scorecard:

The example row is illustrative, not a measured result.

Then ask: Did it save time? Was the final work accurate? Did it make the customer’s experience clearer?

If the answer is no, change the instructions, or stop using AI for that task. You do not need to force it into every part of your business.

Your name is still on the work

In a business built on referrals, reputation is the asset. Every FAQ answer, follow-up and post is a small promise with your name on it. AI can help you make those promises faster. It can’t keep them for you.

So give it one useful job. Check its work. Keep what helps.

The goal is not to sound more automated. It is to free up attention for the work, judgment and relationships that make your business worth choosing.

Follow BlackRefer Insights for more practical, sourced guides for Black business owners. And if you know a Black-owned business that belongs in the BlackRefer directory, refer it.

---

About the author: Askia Jones is the founder of BlackRefer and the Black Ecosystem, a network of platforms covering Black business, health, scholarship, culture and innovation, including Melanated.ai. He builds and runs these properties with the help of AI tools for research, editing and planning, and he writes Insights to give Black-owned businesses practical guidance they can use right away.

Data in this article were current as of November 15, 2026. Survey figures describe the businesses each survey sampled and may change in later releases.

Sources

Source 1

U.S. Census Bureau, Business Trends and Outlook Survey (BTOS), national estimates file, collection period October 19–November 1, 2026 (published November 5, 2026), questions on AI/“SI” use in the last two weeks. Data page: https://www.census.gov/hfp/btos/data File: https://www.census.gov/hfp/btos/downloads/National.xlsx

Source 2

Adam Grundy, Cory Breaux and Dhanapati Khatiwoda, “Large Firms With at Least 20 Employees Biggest AI Users,” U.S. Census Bureau, May 26, 2026. https://www.census.gov/library/stories/2026/05/ai-use-businesses.html

Source 3

U.S. Chamber of Commerce Technology Engagement Center, Empowering Small Business: The Impact of Technology on U.S. Small Business (2025), key findings and methodology (Teneo Research survey of 3,870 small businesses with fewer than 250 employees, June 6–26, 2025). https://www.uschamber.com/assets/documents/Empowering-Small-Business-Report-2025.pdf

Source 4

Federal Reserve Banks, Small Business Credit Survey, 2026 Firms in Focus: Chartbook on Firms by Race and Ethnicity of Ownership (findings from the 2025 survey, fielded September–November 2025; employer firms only; AI data from an optional module answered by about 81% of employer-firm respondents), pp. 42–47. https://www.fedsmallbusiness.org/-/media/project/clevelandfedtenant/fsbsite/reports/2026/2026-firms-in-focus-chartbooks/sbcs_chartbook2026_raceethnicity.pdf

Source 5

Daniel Eliot, Jeffrey A. Marron and Savann Thorn, Small Business Cybersecurity: Non-Employer Firms, NIST Cybersecurity White Paper 50, National Institute of Standards and Technology, November 2, 2026, Sections 1. [9] and phishing/MFA guidance. https://nvlpubs.nist.gov/nistpubs/CSWP/NIST.CSWP.50.pdf

Source 6

National Institute of Standards and Technology, Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile (NIST AI 600-1), July 2024, Section 2.2, “Confabulation.” https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.600-1.pdf

Source 7

Federal Trade Commission, “Trade Regulation Rule on the Use of Consumer Reviews and Testimonials,” final rule, Federal Register 89 FR 68034, August 22, 2024 (effective October 21, 2024). https://www.federalregister.gov/documents/2024/08/22/2024-18519/trade-regulation-rule-on-the-use-of-consumer-reviews-and-testimonials

Source 8

Electronic Code of Federal Regulations, 16 CFR Part 465, “Rule on the Use of Consumer Reviews and Testimonials,” §§ 465.2, 465. [4] and 465.5. https://www.ecfr.gov/current/title-16/chapter-I/subchapter-D/part-465

Source 9

Michael Atleson, “Keep your AI claims in check,” FTC Business Blog, February 27, 2023. https://www.ftc.gov/business-guidance/blog/2023/02/keep-your-ai-claims-check (archived copy: https://web.archive.org/web/20240829094818/https://www.ftc.gov/business-guidance/blog/2023/02/keep-your-ai-claims-check)

Source 10

National Institute of Standards and Technology, Small Business Cybersecurity Corner. https://www.nist.gov/itl/smallbusinesscyber

Source 11

U.S. Small Business Administration, “Manage your business: Strengthen your cybersecurity.” https://www.sba.gov/counseling/manage-your-business/#strengthen-your-cybersecurity