Brainstorming with ChatGPT, Part 1: The One-on-One Meeting

I initially gave this talk at the Marketing / AI conference MAICON in fall of 2023, then again at the November 2023 B2B Marketing Expo in London. I’ve had lots of requests for the slide deck, so I thought it warranted a Technormal post that I can point people to.

Blog post 1 will be all about how to brainstorm privately with ChatGPT. Before the meeting.

You'll have to ignore that my blog headers mirror the words on the slides...remember this was a takeaway presentation, so I wrote it to make sense without my presentation.

Take a look and let me know what you think! This is Part 1 of 2.

Who the Heck Are You Again?

Fair question. Here's a bit about me, career-wise. I have worked at a lot of huge media and tech enterprise companies...but at the edge of emerging technologies. I was Director of Social Media at ESPN, led developer programs at Meta, and headed up product partnerships for Meta's (now defunct) New Product Experimentation business.

I spent a year consulting in AI, speaking at events and working with Fortune 500 businesses on strategic implementation of AI through my company and blog, Technormal. My old blog can be found at medium.com/@technormal, but I'm slowly migrating posts to this new spot.

I co-founded a startup at the intersection of geospatial computing and augmented reality, helping you play games, receive offers, and generally let your mobile device tell YOU when you're in a place with location-based content for you. We're called Loud Labs.

Find the rest of my career journey on LinkedIn.

Hi. I'm Katie from Technormal. I've worked across a lot of media and tech companies.

Okay. Now that that's out of the way, let's dive into Part 1.

Part 1. Your 1:1 Brainstorm with ChatGPT

But before we dive in...a question:

WHY should we do the solo-ChatGPT brainstorm before the team one?

So many reasons why, but here are a few: 

1. You learn by getting hands on. Take some time to play around with the platform solo.

2. This process is going to make you feel a little silly at first. You're stepping outside of the way we've been commanding computers and structuring our searches. It feels strange at first to talk to a computer platform like you would a person.

3. The same as any brainstorm! Before you get into a room full of ideas, you should have some solo time to originate your own ideas.

Okay. So let's set you up for success with this 1:1 ChatGPT brainstorm exercise.

Follow these steps:

1. Clear your calendar for 2-3 hours one workday.

Hold my calls, Shiela.

Just like any brainstorm, you need some structure first. Get a few hours where you will not be disturbed. Keep the door shut. Turn off the notifications.

Open up ChatGPT. If you haven't yet, create an OpenAI account. You'll see an input box in the middle of the screen that says Message ChatGPT. That's where you are going to enter your first prompt. A generative AI 'prompt' is equivalent to a computer command or a Google search query: It's a way of telling the program the next thing you want it to do.

You're now ready to write the opening prompt to get this brainstorm rolling.

2. Pick a Priority Project

I know the perfect priority - we have to pitch P&G on a rebrand of the iconic brand, Pepto Bismol.

Here are some questions to help you pick your first project:

• What's your #1 hi-pri project this quarter?

• What upcoming pitches are looming?

• What project has the greatest impact?

• What crucial thinking are you currently putting off?

Got your project? Great. Let's move on.

3. Prompt ChatGPT Like You're Emailing a New Intern.

"Hey Boss! I'm here to help. Also? I have no idea how to help. Please be very specific with me."

Now, I want you to write your prompt as if you are explaining a new project to your new intern, ChatGPT. ChatGPT is a brilliant, gifted intern -- the fastest learner you've ever met, in fast. Also? He has zero context whatsoever. He's ready to help...and has zero idea of who you are and what you're doing.

The Initial ChatGPT Prompt

I'm going to stick with the big Pepto Bismol pitch we have to make to Proctor & Gamble as our example use case for this brainstorm. Now I'm ready to start a new conversational thread with ChatGPT.

You're going to start a new 'conversation' with ChatGPT by typing your prompt into the 'Message GPT' input box.

Here's the initial prompt I am going to use for the Pepto brainstorm example:

Oh, look. There's our little intern, ChatGPT, now. Hey, little buddy.

Let's break down my prompt engineering (aka: how the query is structured), bit-by-bit, so you understand why I crafted it this way.

The introduction.

"Hi Boss. I'm here to help. Also, who are you?"

Give ChatGPT some context as you would with anyone new you meet. Who are you? What is your job/role? What is the project? I also always greet ChatGPT. Just like I would with any person:

Hello there! I'm a creative director at a renowned marketing agency, and we've just been given the exciting opportunity to propose a rebranding campaign for Pepto Bismol.

Some context.

"Ah. Fascinating. I can picture it now. I beg you...go on."

Give ChatGPT context, as you would to your intern (who grew up in the rainforest, and has never heard of Pepto Bismol):

• What is it? an iconic otc medication

• Purpose? digestive relief

• Known for? pink color, chalky taste, long history in the U.S.
  

Your shared challenge explained.

"Make Pepto relevant for a younger generation, while respecting the iconic brand? With my eyes closed."

Next, I explain the challenge that the two of us are here to tackle together. What did P&G include in the RFP? WHY are they asking for this brand update? ChatGPT always does better when you explain the exact goal AND give it tons of context.

Explain ChatGPT's role in the brainstorm.

"Ah yes. I make a great ideation partner, my friend. Remember to push me if my first reply is lame. Examples help enormously."

Here, I am giving my intern his role in all of this. Giving generative AI it's role is a crucial step in prompt engineering that I use every time. Not sure how to give AI a role? Start the sentence with, "I want you to act as if..." and fill in the blank. This is where I really want you to go out on a limb and pretend. Picture the intern and describe exactly who ChatGPT is in this scenario.

I also explain what's going to happen next: "We're going to collaborate..."

Ask it to summarize the scenario back to you before starting.

"I sometimes go off on strange tangents, so correct me before it happens."

Asking for a summary forces ChatGPT to restate the job to be done BEFORE jumping in. This gives you the chance to course-correct before it's off and running (in the wrong direction). Be patient. If his recap is wrong, explain it again. Use analogies, examples. Remember, he basically knows everything ever published on the web.

4. Review ChatGPT's Recap. Closely. Is this right?

"I always want to get started immediately. Don't let me peer pressure you. Try explaining more"

If your intern's prompt recap isn't quite right, do NOT let him get started (he'll try). Simmer down, ChatGPT. We'll get started once you understand the project exactly right, buddy.

Keep tweaking the prompt and asking for recaps until he gets the assignment right.

5. Outline the Deliverables.

"Aw, don't feel dumb giving me praise. I love it. Remember, I am not a Google search query. Be nice."

Once I'm happy with the prompt, the first thing I do is praise ChatGPT. I'm not kidding.

Excellent job, ChatGPT.

Generative AI seems to respond to encouragement and praise (just like an intern!).

Now we're going to give very specific expectations to ChatGPT.

Please come up with 5-10 out-of-the-box, game-changing, award-winning rebranding concepts for Pepto Bismol, optimized for the 18-24-year-old demographic, while respecting the legacy elements of the Pepto Bismol brand.

Ever heard the saying: "Garbage in, garbage out."? That absolutely applies to how you write your generative AI prompts.

6. Review ChatGPT's Results. Pick 3-4 Favorites.

"Here are my favorite 3 concepts. Based on your parameters, here's why."

My query generated a decent list of creative ideas. I had a few I liked, but I always ask ChatGPT to pick it's favorites. Yes...it will have favorites! The important part is that it tells you why. The why's should align with the parameters you set (your demographic, your do's and dont's and priorities.

7. File Those Top 3 Ideas for Your Team's Brainstorm

"If you don't like my ideas, I am not offended! Just like with a person, you should tell me 'No' (kindly) and ask me to try again."

Just like when you brainstorm with a person, you will sometimes 'get' each other, and sometimes not. If you don't like ChatGPT's results, ask it to try again...AND give it better parameters!

Hot Tip for When ChatGPT Goes Off the Rails

Sometimes you'll start reading ChatGPT's results while it continues to write, and write, and write. If it gets one little thing wrong, you can wait until it finishes then correct it.

But if the results are totally batshit crazy? Click the Stop generating button! Life is too short to wait for ChatGPT to give you wildly-wrong results. Click that stop button, which brings GPT's writing to a halt and gives the control back to you.

Now it's up to you to tell your intern:

• Why you stopped him mid-sentence.

• What you don't like.

• Why you don't like it.

• What should change?

• Can you share examples of the types of results you do want?

Hot Tip: When ChatGPT goes off the rails (and won't stop), hit the STOP button!

So that’s part 1 of 2. In Part 2, we take those top brainstorm concepts you developed in this 1:1 brainstorm with ChatGPT, and we use them to kick off your team brainstorm with ChatGPT.

Look for Part 2 coming soon. In the meantime, remember: You can't break ChatGPT by trying. Don't be afraid to just play around with it!