ChatGPT Deep Dive for Business Owners: Models, Prompts, File Uploads, Images, Custom GPTs & Projects

Worksheet

The worksheet mention in the video can be found here: https://tinyurl.com/GPTDeepDive.

ChatGPT as a Business Tool

The goal of this workshop was simple:

  1. Help you see ChatGPT as a useful business tool

  2. Give you reusable prompt frameworks that produce better output

  3. Make you familiar with the features that matter for business owners

1) Know What You’re Using: Models Matter

Most people treat ChatGPT like it’s one thing. It’s not. It’s a menu.

My practical breakdown:

  • Latest / default models (fast): great for everyday work like emails, brainstorming, outlines, and summaries.

  • Thinking modes (more careful): better when the work is complex and mistakes are expensive.

  • Reasoning model (o3): use when you want more methodical problem-solving.

  • Deep Research: when you want a sourced report (think: market scan, competitor table, pricing research).

  • Codex: for code work (aka “vibe coding”). Most business owners won’t need this day one.

Example: I ran a Deep Research prompt on the wedding market for a Grass Valley winery. It took real time and returned a full report—competitors, pricing, positioning, and demand patterns. That’s not a “chat.” That’s a junior analyst doing work.

My suggestion:
Use the default model for 80% of tasks. Switch to “thinking” or o3 when accuracy and reasoning matter. Use Deep Research when you need citations and market intel.

2) Personalization: Get Output That Sounds Like You

ChatGPT has a problem: it tends to sound like… ChatGPT.

Personalization helps fix that.

Here’s what I taught:

  1. Collect 3–5 writing samples (emails, LinkedIn posts, blog intros—whatever is you)

  2. Ask ChatGPT to analyze them and generate a short personalization block

  3. Paste that into ChatGPT’s personalization settings

But here’s the catch: personalization “leaks.”
Meaning: if your default tone is snarky or casual, it might show up in places you want to sound more formal.

Fix: Start a new thread with a “clean slate” instruction like:

New task. Ignore prior personalization and context.

Or keep your personalization neutral/professional and use Custom GPTs for specialized voices.

3) Prompt Frameworks: Stop Winging It

Most bad outputs come from bad inputs. The fix isn’t “more prompting.” It’s better structure.

Framework #1: CREATE+

This one is easy to teach because it’s basically a mini-brief.

  1. C — Context: what’s happening

  2. R — Role: who the AI should act as (fractional CMO, ops manager, analyst)

  3. E — Expectation: what “good” looks like

  4. A — Audience: who it’s written for

  5. T — Task: what to produce

  6. E — Examples: examples if you have them

  • a built-in fail-safe: ask up to 3 clarifying questions if needed

Example: I used CREATE+ to generate five LinkedIn posts promoting an upcoming AI workshop. It asked smart follow-up questions like:

  • Is the event free?

  • Are you offering food/drink?

  • What’s the CTA?

That’s the point: good frameworks create good follow-ups.

Framework #2: Socratic Method (One question at a time)

This is my favorite for business owners who don’t want to “learn prompting.”

You tell ChatGPT:

  • Ask one question at a time

  • Summarize my answer

  • Keep going until you have enough

  • Then produce the deliverable

It’s basically a guided intake form—without the form.

4) Custom GPTs: Make It Repeatable

If you have a workflow you do over and over, you don’t want to rebuild the prompt every time.

Custom GPTs are reusable assistants you configure once with:

  • a consistent role

  • a workflow

  • guardrails

  • (optional) knowledge files

My real example: I created a “WWII Poster Maker” GPT to generate 5x7 WWII-style recruiting posters from an uploaded veteran photo. It became a speaker-gift machine: upload photo → pick branch → generate 3 options → select → produce final.

Key lesson: Custom GPTs are best when you have:

  1. Repeatable flow

  2. Consistent output format

  3. A specific audience

  4. Guardrails (don’t invent facts, don’t share personal info, etc.)

Bonus: Actions

Actions let a GPT connect to external tools via APIs (CRM, scheduling, ticketing, etc.).
For most business owners: good to know it exists; not required for your first week.

5) File Uploads + Data Analysis: The Hidden Superpower

This is where business owners start leaning forward.

What you can upload

  • PDFs, Word docs, text files

  • Spreadsheets (XLS/CSV)

  • Images (JPG/PNG/WebP)

Big note: audio/video files aren’t reliably supported, but there’s a workaround:

  • upload video to YouTube (unlisted)

  • grab transcript

  • paste transcript into ChatGPT

What to do with documents

Here are the highest ROI uses:

  1. Summarize a document

  2. Extract requirements into a table (great for projects and construction scopes)

  3. Simplify legal-ish language (NDAs, contracts)

  4. Turn notes into checklists/SOPs

Example: I used ChatGPT to review an NDA. It flagged a key clause: the other party reserved the right to build competing products. That kind of review would cost real money with a lawyer—so even as a first-pass tool, it’s valuable.

What to do with spreadsheets

I demoed this prompt:

  • 5 key trends (with numbers)

  • 3 anomalies

  • 3 actions in the next 7 days

  • 5 follow-up questions

That’s how business owners think: “What’s happening and what do I do next?”

6) Image Generation: Structure Beats Creativity

You don’t need to be an artist. You need a repeatable brief.

I used a simple framework (PICTURE) to keep image prompts consistent:

  • Purpose

  • Identity

  • Copy

  • Treatment (style)

  • User view (camera)

  • Requirements

  • Experiments

Example: generate multiple logo concepts for a winery wedding brand.
The win here isn’t the first image—it’s the variations.

7) Projects: Organize Your Work (With Limits)

Projects let you keep related chats, files, and instructions together.

It’s a nice workflow improvement when you have:

  • a marketing project

  • a product launch

  • an event

  • a client engagement

But there’s a limitation I don’t love:

you can’t add Custom GPT threads into Projects.

So I use Projects to organize standard chats because I rely heavily on Custom GPTs for repeatable workflows.

The Business Owner Rule: You Still Own the Output

This is the part I’m going to keep repeating:

  1. ChatGPT can save you time

  2. It can improve your thinking

  3. It can draft content and structure work

  4. But you must review before publishing

Because your output represents your brand.

(We’ve all seen the examples where someone didn’t proofread and published nonsense. Don’t be that person.)

Takeaway + Next Step

ChatGPT is most valuable when you stop using it like a chatbot and start using it like a tool.

My recommendation: pick one workflow this week:

  • summarize a document

  • extract requirements

  • analyze a spreadsheet

  • write 5 posts using CREATE+

  • build one Custom GPT for a repeatable task

Do one. Then stack the next.

That’s how you get real ROI from AI.

Next
Next

The “Minimum Viable Please” Episode: What Founders Still Get Wrong About MVPs