Heatherstone

AI for Your Business

The handout from the lunch-and-learn. Everything here you can do yourself, today, for free, in ChatGPT, Claude, or Gemini. No code, no apps, no subscriptions beyond the chat tool itself.

This page grows over time, so bookmark it.

The one rule that fixes most bad AI output: the AI only knows what you tell it. Brief it like a sharp new hire who knows nothing about your business yet.

1. The pattern that fixes 80% of it

Before you hit enter, ask yourself three questions:

1. Did I give specifics? Who is it for, what tone, how long, what outcome, what to avoid. “Write an email” is a coin flip. “Write a 100-word, friendly-but-firm email that ends with one clear next step” is a result.

2. Did I give context? The facts the AI can’t possibly know (the customer’s situation, your terms, your prices, your business). Paste them in.

3. Did I show an example? “Here’s one I liked” beats any amount of describing. One good example teaches the AI your voice faster than a paragraph of instructions.

Before and after

Before (generic, useless):

Write an email to a customer who hasn’t paid their invoice.

After (specific, usable):

Write a short, friendly-but-firm email to a customer whose invoice is 15 days overdue.

Context:

  • We’re a landscaping company. The customer is a long-standing regular we value.
  • Invoice #1042, $480, was due on the 1st.
  • We can split it into two payments if that helps.
  • Our tone is warm, local, and direct, never corporate or threatening.

Keep it under 120 words. No legal language. End with one clear next step.

Here’s the tone I like, from a past email: “Hey Dave, just a heads-up that invoice #998 slipped past its due date. No worries if it got buried; happens to all of us. Mind getting that squared away this week? Holler if you’d rather split it up.”

Same task. The second one gives you something you can actually send.

2. Keep a “company brief” and paste it in

Stop re-explaining your business in every chat. Write one short block once, save it (in your notes app), and paste it at the top of any chat where you want on-brand output.

Fill-in template

COMPANY BRIEF

We are: [business name], a [what you are] in [location].
We do: [your main products/services, in one or two lines].
Our customers: [who they are].
Our voice: [3-4 words, e.g. friendly, plain-spoken, local, no corporate jargon].
Things we always do: [e.g. offer payment plans, mention we're family-owned].
Things we never do: [e.g. hard-sell, use legal threats, sound stiff].

Filled-in example

COMPANY BRIEF

We are: Heatherstone Landscaping, a family-run crew in Kingsport, TN.
We do: residential lawn care and seasonal cleanups.
Our customers: homeowners, mostly repeat regulars who know us by name.
Our voice: friendly, plain-spoken, local; we sound like a neighbor, not a corporation.
Things we always do: offer to split larger bills; mention we're family-owned.
Things we never do: use legal language; sound stiff or corporate.

Paste that on top of a request, then your request can be one short line and still come out right.

3. Give it permission to say “I don’t know”

The AI will confidently make things up (prices, policies, facts, dates) when it doesn’t actually know. This is the single biggest trust trap for business use.

The fix is one sentence, added to your request:

“If you’re not sure about something, say so. Don’t guess or make it up.”

Use this every time the answer involves facts you’re going to rely on.

4. Verify the load-bearing facts yourself

The honest map of what to trust:

Trust it for drafts, rewrites, brainstorming, summarizing, explaining, first attempts. A miss costs you nothing; you just edit it.

Check it yourself for anything that has to be exactly right:

Rule of thumb: if being wrong would cost you money or trust, a human checks it before it goes out.

5. Right-size the tool

Chat AI is brilliant for some things and the wrong tool for others. Knowing the difference saves you frustration.

Chat is the right tool when:

You’ve outgrown chat when any of these is true:

Section 7 is a quick way to tell which side of that line you’re on.

6. Make it not sound like AI

Here’s the catch nobody mentions: AI writing sounds like AI writing, and your customers can tell. A polished email that reads as machine-generated does the opposite of what you want. It signals you couldn’t be bothered to write to them yourself. The fix is a quick editing pass before anything goes out. A few high-value habits:

Kill the em dashes. The long dash ( — ) is the single biggest giveaway. AI loves it; most people don’t type it. Read your draft and swap each one for what actually fits: a colon if the second half explains the first, a period if it’s really two sentences, a comma for a simple pause. This one change does more than any other.

Cut the buzzwords. AI reaches for a stock vocabulary that real people rarely use in a customer email: delve, leverage, robust, seamless, navigate, foster, tapestry, testament, elevate, unlock, in today’s world, it’s worth noting that, at its core. When you see one, replace it with the plain word you’d actually say, or cut it.

Drop the wind-up and the wrap-up. AI announces what it’s about to do and then summarizes what it just did. Cut openers like “in today’s fast-paced world” and closers like “in conclusion” or “in summary.” Start with the point; end when you’re done.

Break the rhythm. AI writes sentences that are all the same length, which lulls the reader and feels robotic. Vary it. A short sentence after two long ones wakes the page up. Read it out loud. If it drones, it needs a break.

Watch the “groups of three.” AI defaults to listing things in threes (“fast, reliable, and affordable”), over and over. Once is fine. Every sentence is a pattern. Mix in twos and fours, or just pick the one thing that matters.

The bottom line: AI gives you a fast draft; you make it sound like you before it reaches a customer. That last pass is the whole difference between “this business uses AI” and “this business sent me a robot.”

Want the complete version? I keep a full editing checklist, the “AI Tell Scrubber,” with the entire buzzword list and the deeper structural fixes for longer writing. Get the full scrubber →

7. The “beyond chat” self-diagnosis checklist

Got a task in mind? Run it through these. The more boxes you check, the more it’s worth building into a real automation, and the less it’s something you should keep doing by hand in a chat window.

0-2 boxes: chat handles this fine. Use the techniques above. 3-4 boxes: borderline, worth a conversation about whether building it pays off. 5+ boxes: this is a built-solution candidate. You’re spending time (and risking errors) on something a system should handle. This is what I build →

8. A few more useful moves

Smaller techniques worth knowing:

Where this stops, and what I do

Everything above is yours, free. When you hit something on the far side of that line in section 5 (something that has to repeat, connect to your systems, run unattended, or where errors matter), that’s the work I do: building AI into your actual business workflows so it runs reliably without you babysitting it.

Trey Overton, Heatherstone. AI workflow automation. Kingsport, TN.

Reach out: trey@heatherstone.com · heatherstone.com

Last updated: June 2026. This page grows; new techniques and examples get added over time.