What should a small business use AI for, and what should it avoid?

Use AI for leverage on tasks where a mistake is cheap and reversible: drafting emails and posts, summarizing, brainstorming, reformatting, first-pass research. Keep a human hand on anything that touches a customer, money, or a promise — quotes, invoices, contracts, anything sent in your name. The rule is 'leverage on a leash': let AI do the work it's genuinely good at, and never let it be the last step before a customer, a payment, or a commitment.

Is it safe to let AI talk to my customers directly?

Let it prepare, not send. AI can draft a reply, a quote, or a message beautifully — and confidently get a price, a policy, or a fact wrong, in your name, at scale. Keep a human approval between the AI and the customer. The point isn't distrust of the tool; it's that a customer, a payment, or a promise is exactly where a confident mistake is most expensive, so that's where the human hand belongs.

How do I stop AI from creating hidden liability in my business?

Two habits. First, the leash rule: a human approves anything touching a customer, money, or a promise before it goes out. Second, verify before you trust: AI states figures and claims with total confidence whether or not they're accurate, so check anything load-bearing — a price, a legal claim, a number that hits your books — before it leaves your hands. Most AI liability in a small business comes from trusting a fluent answer that was never checked.

Toolkit · AI for small business

The Small-Business AI Checklist: What to Automate and What to Keep on a Leash (2026)

Published July 4, 2026 · Vita Indarra

Short answer: for a small business, AI is real leverage — and a real liability pointed at the wrong task. You don't need a policy binder to get this right, just one rule: automate what AI is genuinely good at, and keep a human hand on anything that touches a customer, money, or a promise. Below is the decision card you run each task through, so you capture the hours AI can give you without handing it the three things a confident mistake would cost you most.

Leverage on a leash

AI's great gift to a small operation is leverage: it drafts, summarizes, researches, and reformats faster than any hire, and it never sleeps. Its great trap is that it does the dangerous things just as fluently as the safe ones — it will write a wrong price, a wrong policy, or a wrong promise in your name with exactly the same confidence it writes a good email. So you don't choose between "use AI" and "don't." You put it on a leash: full leverage where a mistake is cheap, a human hand where a mistake reaches a customer, your money, or your word.

The decision card — run each task through it

TASK: ____________________

[ ] 1. WORST CASE if the AI gets it wrong: ____________________
      cheap + reversible (a draft, a note)         -> AUTOMATE
      touches a CUSTOMER, MONEY, or a PROMISE       -> LEASH IT

[ ] 2. AUTOMATE (let AI run it, reclaim the time):
      [ ] drafting emails, posts, descriptions      [ ] summarizing long material
      [ ] brainstorming / first-pass ideas          [ ] reformatting & cleanup
      [ ] first-pass research (then verify)

[ ] 3. LEASH IT (AI prepares, a human approves before it goes out):
      [ ] anything sent to a customer in your name  [ ] quotes, invoices, pricing
      [ ] contracts, commitments, promises          [ ] anything that hits your books

[ ] 4. VERIFY before you trust (for anything load-bearing):
      [ ] a price / figure / date                   [ ] a legal or policy claim
      [ ] a fact a customer will act on
      AI is confident whether or not it's right — check it before it leaves your hands.

RULE: leverage where a mistake is cheap; a human hand on customer, money, promise.

Make the leash a real check, not a rubber stamp

The leash only works if the human review is real. A person who skims and clicks "approve" on a wall of AI output isn't oversight — they're latency with a signature. Keep the checks few and meaningful: don't route every draft to yourself (you'll stop reading), route only the customer-, money-, and promise-touching ones, and actually read those. A narrow leash you honor beats a wide one you ignore.

Frequently asked

Won't a human in the loop cancel out the time savings?

Not if the leash is narrow. Most tasks are cheap and reversible — automate those fully. Only the customer/money/promise slice needs a human, and that slice is small. You keep almost all the leverage and give up almost none of the safety.

What's the single most common AI mistake small businesses make?

Trusting a fluent answer that was never checked — a confident wrong price or claim that reaches a customer. The verify-before-you-trust habit in block 4 prevents most of it.

Where do I start?

Pick your three most time-consuming cheap-and-reversible tasks and automate them this week. For the intuition first, see using AI in a small business without getting burned.

Go deeper

The field guide behind this checklist

This decision card is the working core of The Leverage and the Leash — real AI results for small business owners, without getting burned. Which tasks to hand AI, where to keep a hand on the leash, and how to capture the leverage without inheriting the liability — in plain language, for people running a business, not a lab. Live on Amazon.

The Leverage and the Leash · $9.99 Which book should I read first?

← More field notes