Good vs bad AI practices for small businesses

AI lets you move fast, and it lets you do damage fast. The difference is almost always how you use it, whichever tool you pick.

Good practices

• A human reviews anything customer-facing before it goes live, and spot-checks it afterwards.
• Clear scope. You know what the automation does, what it will never do, and who is accountable if it fails.
• Honest disclosure. If a customer is talking to an automated assistant, they are not misled into thinking it is a person.
• Data care. Only the data that is needed is used, and sensitive data is handled deliberately. See the privacy checklist.
• You own the accounts and data so you are never locked in.
• Written scope and quote before work starts.
• Real, checkable claims in your marketing.

Bad practices

• Publishing AI output unchecked, including facts, prices and legal or medical statements.
• Fake or incentive-skewed reviews, invented testimonials or made-up credentials.
• Mass-producing thin content to chase rankings. See AI content: what is fine and what backfires.
• Pasting confidential customer or business data into tools without understanding where it goes.
• Letting an automation make high-stakes decisions about people.
• Hidden text or instructions aimed at AI systems, which is manipulation.
• Vendors who promise guaranteed rankings or recommendations.

Warning signs in a vendor

• They cannot explain who reviews the work.
• No written scope or quote.
• They keep the accounts and you cannot export your data.
• Big promises, no specifics.

See questions to ask an AI vendor.

A simple test

Ask: if this goes wrong, who notices, who fixes it, and who tells the customer? If there is no clear answer, it is not ready.

Common questions

Keep reading

• AI for business

  10 Questions to Ask an AI Vendor Before You Sign

  2 min read
• AI for business

  Human in the Loop: Why Every AI Automation Needs Review

  2 min read
• AI for business

  AI Hallucinations: What They Are and How to Prevent Them

  2 min read