AI for the Rest of Us

Breakfast Club
Spartanburg AI Day

March 12, 2026

Kyser Thompson · AI for the Rest of Us


The Four Principles

These came up throughout the day. They’re the foundation for how you work with AI going forward.

01

Stop Googling, Start Talking.

AI isn’t a search engine. Talk to it like a smart colleague who happens to know a lot. When you type a Google-style query into Claude, you get a Google-style answer — generic and surface-level. But when you give it context about who you are, what your business does, and what you actually need, the output changes completely. Give it context. Ask follow-ups. Push back on answers you don’t like. The conversation is the product.

02

Talk Back.

Never accept the first answer. The first answer is a first draft. Always. Think of it as a loop: you describe what you want, you look at what you get, you sharpen your description, you go again. “Make it shorter” is fine, but “cut the first two paragraphs and make the conclusion more action-oriented” is way better. Claude gets better the more specific your feedback is. Every round gets you closer to what you actually want.

03

Don’t Outsource Your Voice.

AI writes like AI unless you teach it YOUR voice. Generic. Polished. Soulless. Your sales emails sound different from your competitor’s. Your customer communications have a tone that’s uniquely yours. That’s what Skills are for — you teach Claude your voice, your format, your standards. Without a Skill, you get generic AI output. With a Skill, you get output that sounds like you. Night and day.

04

Tool, Not Crutch.

Use AI to do things you couldn’t do before. Don’t use it to avoid learning things you should know. Analyze 500 customer reviews. Build a dashboard from a spreadsheet. Organize 300 files. Those are things AI makes possible that weren’t practical before. But don’t use it to avoid thinking, to skip learning, or to replace the judgment that makes your business yours. The goal is to make you better at what you do — not to do it for you.


Talking to AI

Three Levels of Asking

How you ask determines what you get.

Level 01
Vague
“Help me with my marketing.” You’ll get a vague answer.
Level 02
Structured
Tell it who you are, what you do, and what you need.
Level 03
Meta
Ask Claude to ask YOU questions first. This is the cheat code.

WHO / WHAT / WHAT Framework

  1. WHO you are — your role, background, industry
  2. WHAT you do — your business context, daily work
  3. WHAT you need — the specific output or deliverable
“I’m a general contractor who runs a commercial construction firm with 75 employees in the Upstate SC market. I need to draft a response to a subcontractor who’s asking for a 15% price increase on a project we’re already committed to.”

That simple structure gets you 80% of the way to a great response. Most people skip all three and wonder why the answer is generic.

Six Ways to Level Up Your Prompts

Your prompting cheat sheet. You don’t need all six every time — even one or two will make a real difference.

# Technique What It Means
1 Context Give background about you and the situation. “I run a 16-location oil change business in the Upstate SC market.”
2 Examples Show what good looks like. “Here are 3 emails I’ve sent. Write the next one like this.”
3 Constraints Set boundaries on length, format, and tone. “Keep it under 200 words. No bullet points.”
4 Step-by-step Ask it to walk through its reasoning before giving a conclusion. “Walk me through this decision step by step.”
5 Think first Ask it to figure out what it needs before answering. “Before you answer, think about what info you’d need.”
6 Tone/role Give it a persona. “You are a commercial estimator with 20 years of experience in SC.”
Key Concept

Model Choice & Reasoning

Claude has multiple models — think of it like choosing the right tool for the job. Haiku is fast and light, good for simple tasks. Sonnet is the default — balanced speed and quality, good for 90% of what you do. Opus is the most powerful — slower, uses more of your daily quota, but save it for complex analysis or when quality really matters.

Extended Thinking: When you turn this on, Claude thinks step-by-step before answering. You can see its reasoning. Use it for complex math, multi-step analysis, or anything where you want to verify HOW it got to its answer. Don’t use it for “write me a quick email.”


Why Claude

We spent the day on Claude because it goes from chatbot to agent in one place.

The Suite of Products

Chat
Cowork
Claude Code
Projects
Skills
Artifacts
Plugins
Connectors

Most platforms do one or two of these. Claude does all of them.


New Feature

Claude in Excel & PowerPoint

Excel: Claude reads your entire workbook — formulas, dependencies, multiple tabs. Ask questions and it answers with cell-level citations.

PowerPoint: Reads your slide master, fonts, and colors. Generates and edits slides that match your brand. Available on all paid plans.

What AI Can Do Right Now

As of March 2026

01

Build Software

Generate working apps, dashboards, and tools.

02

Write & Edit

Drafts, emails, proposals, content at your level.

03

Analyze Data

Patterns, summaries, and insights from your files.

04

Research

Deep dives, comparisons, market analysis.

05

Organize

Sort files, extract info, build spreadsheets.

06

Automate

Recurring tasks, scheduled reports, workflows.


The Bridge

Three things you need to know before going from chatbot to agent.

01

Everything is a text file

Skills, custom instructions, configuration — all plain text you can read and edit. No databases, no code, no black boxes.

02

Markdown is the new Word doc

Simple formatting with # for headings, ** for bold, - for bullets. Every AI tool speaks it natively.

03

Folder structure matters

An organized folder gets better results than a junk drawer. Think filing cabinet, not desktop pile.

Key Concept

These Tools Have Quirks

The mindset that works: treat it like a brilliant new hire in their first week. Incredibly capable, occasionally confused, gets better every day.


Working with Agents

You went from worker to manager. Here’s how to manage well.

Agent Playbook

01

Sandwich your tasks — Start a task, do something else while it works. Come back to the results.

02

Scope the sandbox — Only give access to the folder it needs. Never your whole system.

03

Check the plan — Read what Cowork proposes before it executes. This is your management moment.

04

Log what worked — Save your best task descriptions. Over time, these become Skills.

What Skills Are

Reusable instruction packages. Teach Claude once, use it forever.

01

Brand Voice

Write like you, every time. No more generic AI output.

02

Report Format

Standard templates for proposals, weekly updates, invoices.

03

Analysis Framework

Consistent methodology for reviewing bids, costs, data.

Go to Customize → Skills → Create with Claude, and Claude will interview you and build the skill. No files to write by hand. If you find yourself explaining the same thing repeatedly, that’s a skill waiting to be built.

Projects vs. Skills vs. Custom Instructions

01

Projects store knowledge

Reference materials, context docs, past work. They give Claude background.

02

Skills teach how to work

Steps, methodology, voice, process. Your standards, encoded.

03

Custom Instructions = identity

What Claude always knows about you. Carries forward across every conversation.

The three work together: Projects give context, Skills give procedure, Custom Instructions give identity.


The Five Key Concepts

The mental models that make everything else click.

01

Message in a Bottle

Every prompt is a message to someone you can’t see. The better the message, the better the result.

When you write a prompt, there’s no body language, no tone of voice, no shared context. Just your words. This isn’t about magic words or secret formulas — it’s about clear communication. And that’s a skill everyone in this room already has. You communicate with clients, employees, and vendors every day. Same skill, new medium.

02

Hallucinations Are Real

AI will confidently tell you things that are completely wrong. Verify everything.

This is the honest truth that a lot of AI hype skips over. AI will cite studies that don’t exist. It’ll invent statistics. It’ll give you a contract clause that sounds legitimate but isn’t. This isn’t a bug they’re going to fix next month — it’s how the technology works. Start with low-stakes tasks where mistakes don’t matter much. Build trust over time. And never send a client something Claude wrote without reading it first.

03

Everyone Becomes a Manager

AI doesn’t eliminate your job. It makes you the manager of AI workers.

You still need to know what good output looks like. You still set the standards. You still review the work. Think about how you onboard an employee — you give them instructions, you spell out the job, you define what “good” looks like and what to avoid. That’s exactly what you do with Claude. The better the instructions, the better the output.

04

Talkers vs. Doers

Chatbots give you info and you act. Agents research, write, execute, and you review.

This is the big shift. Chatbots are conversations — you ask, it answers, you do the work. Agents are task runners — you describe what you need, AI does the research, the writing, the building, and you review the output. The gap between “I use AI sometimes” and “AI is part of how my business runs” lives right here.

05

Taste Is the Scarce Resource

Generation is free. Judgment is what matters.

Now that anyone can produce text, code, images, and dashboards, what becomes scarce? Knowing what’s good. Knowing what serves your customer. Knowing what fits your brand. AI handles volume. You handle quality. You set the standard, Claude does the production. Your taste — your knowledge of your business, your customers, your standards — that’s something AI can’t replace.