Planning Gabrielle: Using chatGPT for planning... anything!

Three patterns that work well: Frameworks, ideation, and feedback.

Apr 17, 2023

Welcome back to Dear AI, where the AI and I share insights on working together on a project. Specifically, we're building Gabrielle.Day, the AI-powered advice columnist. This post is about using ChatGPT for planning.

But first - I needed some advice about writing the post:

Oh, Gabrielle, always with spot-on advice.

Planning with chatGPT: Frameworks, ideation, and feedback

Planning helps me when I know little about the topic - like with marketing or databases. ChatGPT is a great co-pilot. Three patterns, in particular, help navigate uncharted territory: frameworks, ideation, and feedback.

A) Frameworks:

Think back to your first day at a new job. Figuring out where to start can feel daunting. And that happens regularly in this project, where I often don't know what I'm doing. So I turn to ChatGPT for the plan. For example: "I need to write a blog post; how do I go about it?" In seconds, ChatGPT gave this outline:

The initial plan can take some refining. For example, I asked, "That's content to include. What about other aspects of writing a blog?" and that felt more useful:

And there are more ways to refine. When I asked, "Who are some expert bloggers, and what is their advice?" chatGPT summarized some valuable tips. Later, it prioritized those tips by estimating their importance. For example, it explained why writing a clear and structured post is more important than promoting it.

These starting outlines have been super helpful so far. Here are two more examples:

1) This project started as a simple exercise to build an AI-advice autoreply. I asked chatGPT for the top 10 aspects I should consider. It came back with this list below. It was #5 that sparked the idea for this series of posts.

2) I asked chatGPT for help in planning this blog. It gave me this outline with 6 steps. It was a good start! I started following the steps, and it kept flowing from there.

B) Ideation:

Where do ideas come from? Avi Loeb, a prominent astrophysicist, says, "They're nurtured by informal dialogues in environments where mistakes are tolerated and critical thinking is encouraged." His examples include labs full of scientists or hallway conversations at tech companies. Chats with chatGPT share some of those traits - they are informal and judgment-free, and the ideas flow. I've found two modes that work well - monologue and dialogue.

Monologue: for well-defined requests

It's tempting to ask chatGPT for "100 ideas for a business". But those can be repetitive, and I don't find myself personally invested in them. But when I know exactly what I need - chatGPT often delivers:

Dialogue: for open-sky co-creation

ChatGPT is quite good at, well, chatting! When you chat about a problem, you can explore it from different angles, and fragments of ideas come flowing out. Below is one example of how this worked when designing Gabrielle. Then I share some tips on how to engage chatGPT in fruitful dialogue.

Ideation example: designing the experience with chatGPT

I described the bare-bones version of Dear Gabrielle to chatGPT: you write in and get advice. I asked for ideas to make it more engaging and to encourage people to share with their friends.

ChatGPT had several ideas- social media integration, gamification, group advice sessions, etc. First, we explored how users could post Gabrielle's responses on social media. But that felt manufactured in the email context. So I asked - "What about ideas that stay in the email?". Some glimmers appeared in the responses, such as CCs, forwarding, and other email interactions. That led me to wonder about viral email experiences, and I asked for examples. ChatGPT responded with - Hotmail, some non-email viral products like Dropbox, and email chain letters.

The email chain letters caught my eye. I described to chatGPT that they had two essential features:

1) Viral action - forward the email to 10 friends.

2) Perceived benefit - good luck or avoiding misfortune.

I asked for variations on these themes - what are other potential viral actions and benefits? One of those suggestions was fantastic: "Let users create their own email responder to share with their friends, with different personalities." This turned out to be one of the most fun features of Dear Gabrielle.

We continued to develop that flow. I asked chatGPT to fill in the steps where in step 1, a user named Fred emails Gabrielle for advice, and in step N, Fred invites his friend Ashley to try it too. A few more ideas emerged from that discussion, such as "Gabrielle's response includes a PS message," which invites users to create their own advice columnist.

Through the conversation, the flow of interacting with Gabrielle evolved into something far more engaging than the original single email response. (Try it by emailing dear@gabrielle.day!)

Tips for productive ideation with chatGPT:

Imagine you're chatting with a friend or colleague to solve something together. Listen to what they say, share your opinions. Pick up pieces you like and return them with your own changes.
Share context upfront. A summary of the project or problem so that you and chatGPT are aligned on the context and goal.
Try variations. "I'm thinking about Y. Can you come up with variations of Y that might be better for..."
Track of good ideas on the side. While chats with chatGPT are saved in the console, I copy the best ideas to a side document to keep things organized!

C) Feedback:

This pattern is a simple shift in roles. Instead of asking chatGPT to do something, give it your draft and ask for feedback. "Here's my plan - what do you think? Am I missing anything? Any suggestions for improvement?"

When did you last ask for feedback on something you worked on - from a colleague or your boss? People can be uncomfortable with feedback, but there's far less awkwardness with chatGPT. It's not shy about giving feedback and is super pleasant about it. If anything, I found it overly friendly - "Great job!" "Wow, what a great idea!" are not uncommon responses. But if you ask for specific feedback like "Do you have any suggestions for improvement?" or "What do you think is the weakest part of this plan?", it will answer directly.

I've been using chatGPT for feedback quite a bit while building Gabrielle. For example, when I have a draft plan, I ask if I should add anything. I might get a congratulatory response like "Looks great..." or it might suggest thinking about ethics, legal, or other aspects. I've also gotten helpful feedback about written passages and code. Whether it's a new idea, something I forgot to consider, or a friendly "looks great," it helps move forward more confidently.

One thing to remember is that chatGPT is not an expert. You might prime it with "As the world's leading blogger, what do you think about this section..." - but don't assume the answer reflects the pinnacle of human knowledge.

There's one more area where chatGPT's feedback skills have become indispensable - coding - more about that in the next post...

Overall - kudos to chatGPT for planning!

ChatGPT has been very helpful in my planning stages. First, I ask for a framework or outline. Then, details come from bouncing ideas around with chatGPT in a conversation. And when I have a draft, I share it with chatGPT for feedback. This helps keep the work flowing forward, especially in unfamiliar areas.

What do you think?

If you've used chatGPT, how do these patterns around frameworks, ideation, and feedback resonate? Do you have other practices you use successfully? Let me know in the comments, and feel free to ask any questions!