7B Business Solutions

AI Best Practices
+ Getting Started

The framework · Three power moves · Six habits · Five prompts to try this week

What this is: A pocket-sized field guide pulled from the talk you just heard. Start with the 4-part prompt framework — that alone will make every AI tool 3× more useful. Then steal the three power moves most people miss, the six habits that separate "AI works for me" from "AI doesn't work for me," and the five prompts you can paste into Claude (free at claude.ai) this week.

The 4-Part Prompt:
Role · Context · Task · Output

Every good prompt has these four pieces. Most weak prompts are missing one of them. Memorize this — it's the single biggest unlock between "AI gave me garbage" and "AI just saved me an hour."

Role
Tell the AI who to be. "You are an experienced bookkeeper for small construction companies." Sets the lens, the voice, the depth of expertise. Without a role, you get a generic answer from a generic assistant.
Context
Tell it the situation. Who you are, what business, who the customer is, what's already been tried. The 2-3 sentences a smart consultant would want before starting. Context is the difference between an answer for "small business owners" and an answer for your small business.
Task
Be specific about what you want. Not "help me with marketing." Try: "Write three versions of a 90-word follow-up email to a homeowner who got a quote two weeks ago and went silent." Vague request, average result. Specific request, useful result.
Output
Define what "done" looks like. Word count, format, tone, what to avoid. "Three versions, under 90 words each, no exclamation points, end with a question, no 'just circling back.'" Constraints make AI better, not worse.
All 4 parts in one prompt

What it looks like when you string them together

Weak (no framework) Write me a follow-up email for a customer who hasn't responded.
Strong (Role · Context · Task · Output) You are a sales coach who helps trade-business owners write follow-ups that don't sound desperate. I'm a roofing contractor in North Idaho. I gave a homeowner an estimate two weeks ago for $24,000 — they were warm in the meeting, then went silent. I've already sent one "checking in" text. Write me three follow-up email versions: warm-and-casual, professional-and-direct, and value-add (leading with a useful tip). All under 90 words, no exclamation points, end each one with a single question.

Three Power Moves
Most People Miss

If you only remember three things from the talk, remember these. Each one takes the framework above and multiplies it. None of them require you to be a "prompt engineer" — they require you to be willing to ask differently.

Power Move 01

When in doubt, make AI interview you.

The hard part of prompting isn't the words — it's knowing what to include. You're stuck because you don't know what context the AI needs. Fix: make it ask you. Paste this:

I want to accomplish [GOAL]. Before you do anything, ask me 5 questions, one at a time, so I can give you the context you need to do this well. After I answer all 5, write the final output.

Then answer each question as it comes. The AI will arrive at a far better output than you would have prompted into being yourself.

Why it works: AI is great at extracting context from you. Stop trying to write the perfect first prompt. Let the model do the digging.

Power Move 02

Ask AI to be the prompt engineer.

You don't have to be a prompt engineer. Tell the AI it is. Paste this:

You are an expert prompt engineer. Write me a prompt I can use in [Claude / ChatGPT / Gemini] to accomplish [DESCRIBE THE TASK IN PLAIN ENGLISH]. The prompt should include role, context, task, and output format. After you write it, briefly explain why you structured it the way you did.

The output is a polished prompt you can save, reuse, and modify forever. Build a personal prompt library this way — let the AI write the first draft of every prompt you'll ever use.

Why it works: AI knows more about how to talk to AI than you do. Use that asymmetry.

Power Move 03

Show one example, not zero.

"Few-shot prompting" sounds technical. It just means: paste one example of what good looks like before you ask for more. Like this:

Write me a follow-up text in the same voice as this one: "Hey John — couple of guys cancelled this week so I've got an opening Thursday morning if you want to grab it. Otherwise no rush, just let me know." New context: client is Sarah, she got a roofing estimate 10 days ago, hasn't responded.

Even one example shifts the AI's output from "generic AI voice" to "sounds like me." Two examples is even better.

Why it works: AI can't read your mind, but it can pattern-match better than any human. One good example does more work than three paragraphs of instruction.

Seven Habits That Separate
Useful from "Meh"

The framework gets you started. The power moves multiply your output. These habits keep you from blowing your reputation, your time, or your customer's trust.

Habit 01

Give context before you give the task.

The first sentence of every prompt should explain who you are, what business you're in, and who your customer is. AI without context produces generic output. AI with context produces useful output.

Why it matters: Most "AI doesn't work for my business" complaints trace back to skipping this step.

Habit 02

Be specific about what "good" looks like.

Don't ask for "a marketing email." Ask for "a 120-word email to past clients, friendly but direct, no exclamation points, ending with a single question that invites a reply." Constraints make AI better, not worse.

Why it matters: Vague requests get average results. Specific requests get tailored results.

Habit 03

Iterate. The first answer is rarely the best one.

If the output isn't right, don't start over — refine. Say "shorter," "more direct," "rewrite without the corporate language," "give me three more variations." Two or three rounds of refinement beat any single perfect prompt.

Why it matters: AI is a conversation, not a vending machine.

Habit 04

For complex problems, tell it to think step by step.

For anything involving logic, math, comparing options, or multi-step reasoning, add this line: "Think through this step by step before giving me your final answer." The output gets noticeably more accurate. This is the single highest-ROI sentence in prompt engineering.

Why it matters: AI rushes to an answer by default. "Step by step" forces it to slow down — and reduces hallucinations on hard problems.

Habit 05

Never trust facts, names, numbers, or laws without verifying.

AI fabricates with confidence. If the answer includes a stat, a citation, a person's name, a legal claim, or a specific dollar figure — verify it before you use it. Use AI for drafts, structure, and language. Use Google and your own knowledge for facts.

Why it matters: The biggest reputational risk with AI is publishing something that sounds right but isn't.

Habit 06

Keep the human voice. Edit before you send.

AI output sounds like AI by default — too smooth, too symmetrical, too many em-dashes. Read it out loud. Strip the words that don't sound like you. Add one specific detail only you would know. That's the difference between AI content and your content.

Why it matters: People can tell. They engage less with content that feels machine-made.

Habit 07

Use AI for what's repetitive, not what's strategic.

AI is best at the work you've done a hundred times and want to do faster — listing descriptions, follow-up emails, social posts, meeting recaps, summarizing documents. AI is worst at the work that requires judgment, taste, or relationship context. Don't let it replace your thinking; let it replace your typing.

Why it matters: Pointed at the right tasks, AI gives you back 5–10 hours a week. Pointed at the wrong ones, it costs you trust.

Bonus · Bookmarkable Phrase

"Thanks. Now do that again, but ___." The single most underused sentence in prompting. The first answer is the rough draft. The fifth answer is the keeper. Don't accept version 1 if you wouldn't accept it from a human writer.

Five Prompts to Try This Week

Copy. Replace the bracketed sections. Paste into Claude (free at claude.ai) or ChatGPT. Spend ten minutes per prompt. Notice the time savings.

Prompt 01 · Email Triage

Catch up on 30 emails in 90 seconds

Use when: You're behind on email and need a fast read on what actually needs a reply.

You're my executive assistant. I'm going to paste 10-30 emails below. For each one, give me:
1. Sender + one-line summary
2. Does it need a reply? (yes / no / FYI only)
3. If yes — a one-sentence draft reply I can send as-is

Skip newsletters and obvious noise. Be brutal about what doesn't need a response.

EMAILS:
[PASTE EMAILS HERE]
Prompt 02 · Meeting Prep

Show up to a meeting actually prepared

Use when: You have a meeting with someone you don't know well, and want to walk in informed.

I have a meeting with [NAME] from [COMPANY] tomorrow at [TIME] about [TOPIC]. Here's the email thread leading up to it:

[PASTE EMAIL THREAD]

Help me prep:
1. What's the actual goal of this meeting from their side?
2. What three questions should I ask?
3. What's one thing I should know going in that I might be missing?
4. What's a graceful way to push toward [SPECIFIC OUTCOME I WANT]?
Prompt 03 · Client Follow-Up

Send the follow-up you've been putting off

Use when: You owe someone a follow-up email and have been avoiding it for 3+ days.

I need to follow up with [NAME] about [SITUATION]. Last time we talked, [LAST CONTEXT]. It's been [TIME PASSED] since we connected.

Write me three versions of a follow-up email:
1. Warm and casual — like I'm checking in with a friend
2. Professional and direct — like a respectful nudge
3. Value-add — leading with something useful for them (a thought, a resource, a question), not a request

All three should be under 100 words. No "I hope this finds you well." No "Just circling back."
Prompt 04 · Decision Help

Stop spinning on a decision

Use when: You've been turning a decision over for days and can't see it clearly anymore.

I'm trying to decide [THE DECISION]. Here's the context:

[2-4 SENTENCES OF BACKGROUND]

Here's what I'm considering:
- Option A: [DESCRIBE]
- Option B: [DESCRIBE]

Help me think through this step by step:
1. What am I likely underweighting?
2. What's the cost of being wrong on each option?
3. What would I want to know in 6 months that I might not be thinking about today?
4. If a friend in my position asked you, what would you actually tell them?
Prompt 05 · Content Repurposing

Turn one idea into a week of content

Use when: You have one good idea (a story, a lesson, an observation) and need to fill your content calendar.

Here's an idea I want to turn into content:

[ONE PARAGRAPH — your story, lesson, observation, or insight]

Help me turn this into:
1. A 90-second video script (talking-head, single take, no graphics needed)
2. A LinkedIn post (under 1,200 characters, no emojis, ends with one question)
3. A Facebook post (warmer tone, story-driven, can be longer)
4. An email subject line + 3-sentence opener for a newsletter
5. A 1-sentence quote I can put on an image

Keep my voice direct and personal. No corporate words. No "in today's world."
Want help building this into your business?

The 7B Operations Sprint installs one quick-win automation in 14 days.

We find the leak. We fix it. You keep it forever. One quick-win automation, live in your business, in 14 days — backed by a full refund guarantee.

Book a free 15-min intro call →