Podcast Summary:

The ChatGPT Experiment – Ep 96
Title: Real Example: Coaching a Team to Use ChatGPT for Scalable Research
Host: Cary Weston
Date: January 14, 2026
Episode Length: ~31 minutes

Episode Overview

In this episode, Cary Weston demonstrates, step-by-step, how he coached a team from an industry trade association to use ChatGPT for automating and scaling their ongoing research tasks. The real-world example focuses on leveraging ChatGPT’s "project" feature to transform a repetitive, time-consuming manual process into an efficient automated workflow. Cary emphasizes his signature four-part framework for communicating with AI, explores why most people get frustrated with ChatGPT's responses, and shows how treating AI as your strategic partner unlocks greater value.

Key Discussion Points & Insights

1. The Power of Focused Use Cases & Mindset Shift

• Don’t feel you need to “learn all of ChatGPT”; start with one thing that takes too much of your time and is definable and scalable.
  • “The best way to get comfortable and to start seeing value is to find just one thing.” (02:00)
• Treat ChatGPT as a partner, not merely a tool.
• AI as “The Amazing Intern”: Communicate as if instructing a human assistant, not just firing off generic queries.

2. Cary’s Four-Part Framework for Communicating with ChatGPT

• “I always keep this four-part framework in the back of my mind…” (04:22)
  • What are we doing?
  • Why are we doing it?
  • What does success look like?
  • Do you have questions for me to help you do your best work?
• Quoting Cary:
  • “When I say what does success look like, it actually starts thinking about how to do its work to achieve what it is that I'm looking to do.” (05:54)
  • “When I close with ‘do you have questions for me’, it gives it permission to have a conversation, a dialogue, so that it can pull out the details I didn’t share.” (06:35)

3. Real Example: Automating Repetitive Research for an Advocacy Group

• Scenario: Industry group needs to monitor regulatory and legislative changes across 50 states. Their manual method—checking multiple sources—was time-consuming and inconsistent.
• Team was “trying but not finding success” with ChatGPT; defaulted back to manual.
• Cary’s Intervention: Recorded video call. Gave homework:
  • List regularly-checked URLs and define an ideal report (“what would success look like?”).
• Critical Step: Frame the AI’s Role Before Jumping to Action.
  • Cary’s prompt (paraphrased):
    • “I have a big task… I need you to serve as my strategic advisor to help me build out a step-by-step plan to achieve a repetitive goal...”
    • Attach transcript and example information; ask ChatGPT to review and then engage in a dialogue (not just produce output).

4. Why Dialogue Matters: Don’t Rush to Output, Co-Create Solutions with AI

• Cary avoids saying “go fix this;” instead invites ChatGPT into a dialogue:

  • “I want ChatGPT to think with me so that we can end up with a solution that will help ChatGPT do its best work… I'm skipping the step of saying, go fix this problem. And I want ChatGPT… to be my problem solving partner.” (18:25)

• ChatGPT's response (as interpreted by Cary):

  • Identified the problem (“busy middle”) — repeated manual work, filtering, and re-explaining.
  • Outlined steps:
    • Define job description
    • Specify trusted sources
    • Articulate what “good output” is
    • Agree frequency/reporting rhythm

• Key reflective question from ChatGPT:

  • “If you had to choose one primary outcome, do you want a recurring summarized tool or do you want a deeper analysis?” (22:37)

5. Building a Modulated, Flexible Project in ChatGPT

• Why Projects, not one-off prompts:

  • Projects have memory and can be updated/upgraded; custom GPTs do not.
  • Cary:
    • “Think of a ChatGPT project as a tool with a memory so it can remember and build upon itself…” (25:35)
    • “I don't want to laminate anything, right? Because the minute you laminate something, you can't change it.” (26:38)

• Design Elements:

  • The “brain” (Project Instructions) — fixed purpose, big-picture overview, and consistent identity.
  • 4 modular external resources/documents:
    1. Sites to Monitor: Trusted URLs (easy to update/replace without changing instructions)
    2. Material Change Criteria: What counts as material change versus noise
    3. Time and Look Back Rules: Defines reporting intervals
       • “ChatGPT is really poor at knowing what the date is… so define it explicitly in the file.” (28:08)
    4. Output Template: Explicit format for reports (users simply update the template as needs evolve)
  • “The reason I'm doing this in a modulated way… is setting up a project that has a fixed purpose, a single reason for existing. Giving it the brain and the instructions, giving it the big picture one, and then allowing it to look at its resources, its documents, its external tools...” (29:33)

6. Maintenance & Change Management

• Only update the central instructions when resource documents are added/removed.
• External resources can evolve; just update the relevant files, not the whole system.

7. Closing Advice & Philosophy

• Most ChatGPT frustration comes from rushing for an answer rather than properly “inviting it in” to co-create/process.
• “Let ChatGPT do the busy middle (the hard work).”
  • Use transcripts, emails, and upfront work; focus your expertise only at the beginning (setup) and end (editing/perfecting).
• Final Encouragement:
  • “The biggest attribute that will give you the most success is your own curiosity. So dig in and be curious.” (30:25)

Notable Quotes & Memorable Moments

• On Framing When Communicating with ChatGPT:

  • “If you can explain it and if it can be defined, then more than likely ChatGPT can help.” (03:15)

• Why Co-Creation with AI Matters:

  • “Nothing understands ChatGPT like ChatGPT itself.” (16:12)
  • “I'm practicing my own philosophy: I'm getting out of the busy middle, I'm letting ChatGPT do the busy middle.” (22:51)

• On AI Project Design Principles:

  • “The project's brain is in the instructions... The instructions know about the resources, and the resources can change as often as needed.” (27:05)

Timestamps: Key Segments

• [02:00] – Why the best way to start with ChatGPT is to find one time-consuming, repetitive, definable task
• [04:22] – Cary’s four-part framework for working with AI
• [09:50] – Real-world use case: advocacy group research challenge
• [16:12] – Crafting an effective, dialog-driven prompt with transcript and homework
• [18:25] – The importance of engaging ChatGPT as a strategic problem-solving partner
• [22:37] – ChatGPT’s key diagnostic question (summary output vs. deeper analysis)
• [25:35] – How projects work, and the importance of memory in AI workflows
• [26:38] – The "don’t laminate" principle: keep systems flexible and modulated
• [28:08] – Handling date and time issues in ChatGPT projects
• [29:33] – Modular project structure: “brain” + flexible resource files for easy updates
• [30:25] – Closing encouragement: “Dig in and be curious.”

Takeaways for Listeners

• Start with a real, repetitive challenge—don’t overthink “mastering AI.”
• Treat ChatGPT as a smart partner, not just a search bar.
• Invest in framing the problem—the more context and intent, the better the results.
• Use a four-part communication framework to guide your AI interactions.
• Design AI projects for flexibility: separate the “brain” from changeable resource files.
• Let ChatGPT do the “busy middle”—your value is at the beginning (setup) and end (final edits).
• Stay curious, experiment, and iterate—AI effectiveness grows alongside your own curiosity.