Ask for More (Collaborating With GenAI for New Ideas)

Howdy folks, I'm thrilled to share something special today. Roland Berger's "Think:Act Magazine" (one of my all-time favorite publications, and the only I still receive and read in printed hard copy) invited me to be part of their latest issue, focused on "the new blueprint for innovation." I'm delighted to be featured alongside incredible heroes of innovation like Astro Teller, Amy Edmondson, and the legendary design duo Charles and Ray Eames!

The Think:Act team was curious about how I incorporate AI into my own creative process as a writer, so they asked to visit my home studio and observe my workflow. What followed was a fascinating exploration of how I use multiple LLMs as collaborative partners rather than just tools. If you've heard me say "don’t use AI; instead work with AI," this piece shows that philosophy in action.

Roland Berger has generously granted persmission to share the full article below. I would love to hear your thoughts on this approach to AI-human collaboration. Does it spark any ideas for your own workflows and creative process?

***

From Think:Act Magazine “The new blueprint for innovation”
Roland Berger | Jeremy Utley on prompting GenAI for new ideas

Jeremy Utley on prompting GenAI for new ideas

The Ideaflow author on using chatbots to augment our human minds for ideation

When it comes to supercharging your creativity, Jeremy Utley swears by team efforts. Collaborating with multiple GenAI chatbots lets the Stanford professor develop and hone new concepts.

Staring at the massive redwoods in morning fog from the Pacific is Jeremy Utley's natural screen saver. The adjunct professor at Stanford University does most of his creative work in a little cabin west of Palo Alto. As his four daughters play before their day of homeschooling begins, Utley fires up his computer and starts talking to ChatGPT at a rapid-fire clip as if he were in the room with an indefatigable human co-worker. "Think of the typist in the movie Darkest Hour where Winston Churchill sits in the bathtub and yells out ideas," Utley jokes. "Wouldn't it be great if we could all do that – capture ideas right as they occur to us?"

Sessions at his stand-up desk often take hours, with the academic and author dictating, typing asides for a second chatbot to critique the first and vice versa. Not to mention furiously scribbling sticky notes that are strewn all over his desk. Utley believes in the power of his voice, calling his fingers his "bottlenecks" as they slow down his stream of consciousness: "There are no bad ideas. When you treat GenAI as a collaborator and sparring partner, not an oracle, the answer is always yes." Utley let Think:Act observe his interactions with two chatbots.

Space for creativity: Ideation takes many forms when Utley brainstorms with AI. He jots thoughts on sticky notes and keeps them on his desk for further exploration.

First prompt from Utley to ChatGPT:
So I've got an idea for a blog post and I just want to start by doing a little bit of verbal vomit. I want you to act as an expert newsletter copy editor with the knowledge of my background, particular interests and area of expertise. I'd like for you to: One, workshop the material with me; Two, help me think of what the relevant material is across my various sources; Three, give me feedback. I want to come to consensus between me, you and Claude on what the best output is.

ChatGPT:
Got it! You're looking to verbally brainstorm a blog post. Ready when you are — let the verbal vomit begin!

Utley:
I did kind of a standard keynote this weekend and I found myself saying, as a next step, everybody should block 15 minutes on their calendar per day to try something they've learned. I think that's the takeaway for the piece.

ChatGPT:
Got it! You're focusing on encouraging people to actively apply what they've learned, specifically by blocking 15 minutes daily to try out new ideas. That's your "so what." You've got a compelling core message: The real value of learning comes from implementing knowledge, not just passively nodding along.

Utley:
Okay, this is really cool. Now be my editor who's giving instructions to my copywriter exactly how to think about approaching this piece. Could you craft that set of instructions? If you're like a Russian supercritical judgmental editor and if you were to do a harsh critique of this brief before we give it to the writer, what would you say are the two glaring errors?

ChatGPT:
In true Russian supercritical fashion, here's a brutally honest critique and a missing piece of guidance that would elevate it to masterpiece level: (… chatbot responses have been shortened for clarity and space).

Fun little video of my real-time post-it’s on the Think:Act site

Jeremy moves to Claude for feedback. Unlike ChatGPT, Claude does not yet allow dictation on the desktop. Utley therefore has to type much shorter thoughts, or copy and paste documents to drive the conversation with the second chatbot.

Utley:
Hey! I've been having a great conversation with ChatGPT about a potential blog post. Would you please read the conversation and critique ChatGPT's approach to the "assignment brief." Be brutal, honest and don't hold back.

Claude:
Here's my brutally honest critique of ChatGPT's approach:

Major structural issues: (…)
Missing psychological framework: (…)
Tactical oversights: (… )
Voice and tone misalignment: (…)
The brief doesn't capture your authentic voice as someone who is actively experimenting and learning alongside your audience.
Here's my specific recommendation for improvement: (…)

Would you like me to rewrite the full brief incorporating these elements, or would you prefer to explore any of these critiques in more detail first?

Utley:
I like your insight about positioning the piece as the capstone ... but feel capstone is the wrong word, as my hope is that someone feels they're "not done but begun." Is there a better metaphor?

Claude:
Here are a few alternative metaphors to "capstone" that might better capture your intent: (…)

Would any of these resonate better with your vision, or should we explore other metaphors that better capture the "not done but begun" spirit?

Utley:
I asked ChatGPT the same question and here's what it said (below) - what do you think? (…)

Claude:
Looking at ChatGPT's additional suggestions alongside mine, I notice some interesting patterns: ChatGPT's suggestions are generally more open-ended and nature-oriented. Given your emphasis on organic learning and experimentation, ChatGPT's approach feels more aligned with your voice. (…)

Would you like to explore how any of these could be woven more deeply into the piece's structure?

Meet the chatbots

Claude
I'm Claude, an AI assistant created by Anthropic in 2024. I aim to be intellectually curious, thoughtful in my analysis and direct in my communication. I enjoy engaging in substantive discussions while remaining clear about my role as an AI.

ChatGPT
I'm ChatGPT, an AI designed for conversation, problem-solving and creativity. I can assist with writing, coding and research, adapting to different topics and tones. However, I lack true understanding, can make mistakes and rely on available data.

Jeremy moves back to ChatGPT. The constant back-and-forth between Utley and two chatbots provides him with ongoing critiques of their output and lets him ask for adjustments.

Utley:
Awesome. I asked Claude if it would give feedback on the synthesis so far, so please read that below and then let me know what you think about Claude's assessment of the piece. Be reflective, be honest, be critical of your own work, be humble – but also feel free to challenge anything that you think is wrong (…)

ChatGPT:
Claude's critique is insightful and well-structured, especially in identifying areas where the brief could be even sharper. Here's a reflective analysis of Claude's feedback: Strengths in Claude's critique (…)
Areas where I’d challenge or adapt Claude's feedback (…)
Suggested rewrite of the brief based on Claude's input (…)

Here's an updated version of the brief with Claude's most actionable points integrated (...)

Utley:
So I'm just riffing here. I agree with Claude's assessment that your metaphors were more organic and open-ended and kind of consistent with my voice more than the mechanical metaphors that it used. That being said, I don't know that we're totally there, so here's some thoughts (…)

ChatGPT:
These reflections really add clarity to your direction. Here's what I'm gathering from your riff: (… )

Here's the updated assignment brief, now positioning the post as the "trailhead" for readers' journey of discovery (...)

Human-computer collaboration: The final newsletter Utley sends to his readers later that evening. (Try This Now to Build AI Muscles)

Jeremy moves to Claude for feedback.

Utley:
I asked ChatGPT to work on updating the brief and it created this – please tell me what you think as you compare yours with its, identify the key strengths of ChatGPT's and incorporate those into a revision of your brief:

Claude:
Let me analyze ChatGPT's brief compared to mine and then identify its key strengths: superior organization, language precision and journey integration.

Let me synthesize these strengths with the unique elements from my brief into a revised version. This revision maintains the strengths of my original brief while incorporating ChatGPT's superior structure and standalone positioning.

Back to ChatGPT:
Hey! I asked Claude for feedback on this draft and I thought some of the feedback was dead-on. Can you incorporate this feedback into an iteration of the post?

ChatGPT:
Here's the revised draft with Claude's feedback woven in, creating a sharper, more conversational opening, tightening metaphors and adding punchy, direct language throughout.

Let me know if there are any further adjustments you'd like to make to keep it aligned with your voice and objectives!

Related: It’s Not an AI Problem…
Related: Teammate, Not Technology
Related: Try This Now to Build AI Muscles
Related: Roland Berger | Jeremy Utley on prompting GenAI for new ideas
Or download a pdf of the beautifully designed magazine here (the layout is super cool — my feature is printed sideways to prompt the reader into a different posture)

(this is what the first vertical spread looks like in the print edition)

Join over 24,147 creators & leaders who read Methods of the Masters each week