Your Name Is On It — Honest Limits for Owners

The most important thing to understand about AI before you start using it in your business is not what it does well.

It is what it does wrong — confidently, plausibly, and without warning.

AI produces text that sounds correct. It does not know what is correct. There is a meaningful difference between those two things, and the gap between them is exactly where small business owners get burned.

None of this is a reason to avoid AI. It is a reason to understand how to use it so your business reputation and your customer relationships stay intact.

It Gets Numbers Wrong — Check Every Figure

This is the most important limit in this entire lesson, so it leads.

When you use AI for quotes, proposals, and invoices, the rule is: AI writes the words around your numbers. You supply the numbers.

You write the price. You write the hourly rate. You write the total. You write the payment terms. AI formats the professional paragraph copy around those figures — and then you check that every number you provided was preserved exactly. AI sometimes rounds, adjusts, or simply replaces a number you gave it with a different one.

The process: write your numbers first, separately. Then ask AI to write the professional wording. Then compare. Every figure in the final document needs to match what you intended.

This sounds like extra work. It is not extra work — it is the same review you were already doing. The difference is that AI handles the prose that used to take you thirty minutes, and you spend five minutes checking the numbers you already knew.

Don't Let It Talk to Customers Without You

AI can draft a reply to an angry customer that sounds calm, professional, and reasonable — and still commit you to something you didn't intend.

When a customer is upset about a delayed delivery, AI might draft: "We'll have this resolved within 48 hours and will offer a full refund if not." That sounds helpful. But you don't know yet whether 48 hours is possible, and you might not offer refunds for that type of situation.

AI cannot read the full context of your business. It doesn't know your policies, your supplier relationship, your current capacity, or the history with that specific customer. It produces a reasonable-sounding response based on what tends to work in similar situations. Sometimes that is exactly right. Sometimes it is subtly wrong in a way that creates a commitment you didn't mean to make.

The guard is simple: a human reads every AI draft before it goes to a customer. One pass. Thirty seconds. You are not checking for typos — you are checking whether what it says is actually true and actually something you're willing to stand behind.

Keep Customer Data Out of Random Free Tools

When you paste customer information into a general-purpose AI chat tool, you are submitting that information to that company's systems. Depending on the tool and its privacy settings, that data may be used to train future models or stored in ways that don't meet your customers' expectations of privacy.

A customer's first name for context is low-risk. Customer messages are moderate-risk. Full names, addresses, payment details, account numbers, health information, or any data that could identify or harm a person — those are high-risk and should not go into a public free tool.

The practical rule: draft your AI prompts in a way that describes the situation without including private specifics. Instead of pasting "Jane Smith at 123 Main St ordered a $450 repair and wants a refund," write: "A customer had a repair done and is requesting a refund. Draft a calm, professional response that acknowledges their concern and asks for their order number before committing to next steps."

You get a useful draft. The customer's private details stay private.

For business tools with explicit privacy commitments and enterprise-grade data handling, the risk profile changes — but the habit of not pasting private data by default protects you across all tools until you have reviewed the terms of each one.

It's a Tool, Not an Employee

An employee you can ask to explain their reasoning. An employee has stake in the outcome — if something goes wrong, there is an accountability conversation. An employee learns your preferences and adjusts over time.

AI is none of those things.

It generates plausible output based on patterns in its training data. It has no stake in whether the output is right for your specific business. It cannot be held responsible for a wrong quote or a reply that damaged a customer relationship. You can.

Treating AI as an employee is how owners end up in situations they didn't intend: a reply they didn't read, a quote with invented numbers, a tone that misread the situation. The tool is capable. The accountability is yours.

What These Limits Add Up To

Four guardrails, one pattern underneath all of them:

You stay in the loop.

AI is genuinely useful for the five jobs in this track — it saves you real hours and produces genuinely good drafts. But that usefulness only holds if you maintain the human glance between the draft and the send. Read it. Check the numbers. Confirm the tone. Then press send.

That one habit — the review pass — covers most of what can go wrong and preserves everything that makes AI useful to a small business owner.

---

Continue building this practice at academy.jeremyknox.ai.