.jpg)
Raise your hands if you believe in AI’s creative potential...
But brainstorming with AI is kinda underwhelming.
Sure, ChatGPT spits out hundreds of ideas.
But they are dull and not quite usable.
Last week, I sat down with Jeremy Utley.
Jeremy is an AI researcher, adjunct professor at Stanford University, and co-author of the book Ideaflow.
He completely changed how I use AI in my ideation process and I instantly used his techniques to work on some ad creative for Motion.
There was so much gold in our conversation, so I’ve spun up our interview into a mini course on AI and ad ideation.
You’re going to learn:
If you follow the exercises, I promise you will better integrate AI into your ideation & creative problem-solving workflow.
Jeremy’s techniques are incredible and I can’t wait to share them with you.
🐇 Dive deeper into Jeremy’s AI research here with his FIXIT methodology for improving AI & human collaboration: https://howtofixit.ai/
Jeremy also publishes research and AI techniques over at his website: https://www.jeremyutley.design/
How do you push past AI’s tendency toward mediocre ideas and get to something great?
Try these techniques:
The potential for gen AI in problem-solving is real.
But without the right training, AI can limit your creativity and leave you with bland, predictable ideas.
This is what Jeremy discovered in a research study he conducted along with Kian Gohar, CEO of the leadership and development firm GeoLab.
Kian and Jeremy collaborated with four companies—two in Europe and two in the U.S.—for a study involving up to 60 employees at each firm.
These employees formed small teams to tackle specific business challenges, such as developing internal training or expanding B2B sales.
Some teams worked without AI (control group), while others used an open-source version of ChatGPT (experimental group).
You’d think the teams using ChatGPT would blow the other team out of the water, right?
Contrary to expectations, the teams using ChatGPT only produced 8% more ideas on average than the control group, with a slight shift in the quality of those ideas.
“The teams using ChatGPT had fewer bad ideas,” Jeremy told me. “But also, they had fewer A+ ideas than the teams not using AI.”
As the Harvard Business Review noted in their coverage of Jeremy’s research, “misconceptions about generative AI, problem-solving, and the creative process are causing workers and their managers to use the tools improperly, sometimes leaving them worse off than if they’d proceeded without AI input.”
As Jeremy discovered, the problem is the psychological principle of cognitive closure.
“Arie Kruglanski is a Russian psychologist who has studied this,” Jeremy told me.
“What he found is that it is distressing for something to be unresolved. So as humans, we seek cognitive closure. We love closing things and saying it is done.”
In the context of brainstorming, superficially using ChatGPT can actually limit your creativity.
“People will ask ChatGPT to solve the problem, get a few okay ideas, and then sort of settle and they’ll stop ruminating on the problem because they want cognitive closure.”
“Less bot. More chat.”
The solution is to treat AI as a problem-solving partner.
If you don’t have a back-and-forth conversation and continue to work on the problem, you’ll get a lot of okay ideas. But nothing great.
You need to keep pushing deeper, increasing the surface area to let better ideas emerge.
So how do you start to treat AI as a conversation partner instead of an oracle?
That’s the focus of today’s mini-assignment.
In our interview, I asked Jeremy the best way to start getting serious about AI and going beyond the usual shallow interactions.
His advice?
Don’t start by asking ChatGPT to solve a work problem. Instead, start by using AI to help you think through a personal problem.
“This will help you understand what conversing with AI can be like. You can then apply this new knowledge to your business conversations.”
Here’s Jeremy explaining the prompt and how to properly converse with AI. Watch the clip and then follow the instructions below.
Follow these steps during your Chat GPT prompt exercise:
After you get the first response back from ChatGPT, use the tips below from Jeremy to help you further refine and work through the problem together with your AI partner.
“ChatGPT needs inputs. It's not enough to just plug in a prompt, get some ideas back, and then go with those first ideas," says Jeremy.
If you don’t have a back-and-forth conversation with ChatGPT, you’ll get a lot of okay ideas. But nothing great.
You need to keep pushing deeper, asking ChatGPT to revise, explore different angles, feed it more information, and keep chipping away until better ideas emerge.
Watch this clip of Jeremy explaining why human collaboration with AI is broken and how to use the FIXIT method to dramatically increase your idea quality.
.webp)
Here is how to use Jeremy’s FIXIT method.
F - is for having a focused problem statement.
General problems get generic solutions. Instead, you need to define what you’re trying to solve.
A bad problem statement:
“Hey ChatGPT, give me some new ad ideas for an electronics brand.”
A good problem statement:
“We sell a niche earphone product that has noise cancelling features. The problem is, people often think this is a product for listening to music and we struggle to communicate the true use cases which are improving sleep quality, helping people with noise sensitivity, and reducing anxiety while traveling.
We’ve tried to communicate these benefits but it is hard to communicate them quickly in an ad and people are still often confused. Or the ads become too boring and filled with features, making people ignore them.
We need to figure out how to both communicate the unique use cases of our product, while also getting attention on social feeds. How can we show people these are not music-based earphones while also hooking attention and keeping things emotional?”
I - is to ideate individually first.
You need to think through the creative problem yourself before going to AI or going to your team.
This prevents you from reaching “cognitive closure” too fast by accepting B+ ideas from ChatGPT or your team brainstorm.
Spend some time thinking, hit a wall, and then bring your new understanding of the problem back to ChatGPT and your team.
X - is to provide sufficient context or background information to AI.
“You need to supply a minimum of 400 characters,” says Jeremy. “The model is only as good as the input and you are the only input AI has.”
I - is to interact iteratively with the AI.
“Have a conversation. Get ChatGPT to ask you five questions back so that it can gather the input it needs,” says Jeremy. "Then, criticize the ideas it gives you, ask it to regenerate, and don’t settle for the first ideas you find.”
T - is for bringing it back to your team.
The ideas are not going to come out perfect. Take the ideas you get from AI and apply your creative judgment. Review them with your team. Be critical. Look for ways to polish.
With the FIXIT model, you'll better use ChatGPT as an input source into your own creativity. It's about interacting with the AI and treating it as a conversational partner rather being passive and waiting for the oracle to spit out finished ideas.
As a bonus, use a few of Jeremy's prompts below to get new ideas flowing back and forth between your team and Gen AI tools.
According to Jeremy’s research on innovative companies, the teams that win achieve what he calls “idea flow.”
When this state is achieved, teams orientate away from searching for the elusive few good ideas. Instead, they use daily creative practices to develop a high volume of ideas.
“Every business problem is an idea problem,” says Jeremy. “And for teams to succeed, leaders need to create an environment where there is a free flow of ideas.”
So how can you help your teams achieve idea flow?
Jeremy had three actions for me.
“Many folks get creativity and innovation all wrong,” Jeremy told me.
“If you look empirically, having a good idea is purely a numbers game. Everyone is fixated on having ‘good ideas.’ But the best way to have a good idea is to have a lot of ideas.”
Similarly, Frans Johannsen describes this need for idea volume in his book, The Medici Effect.
He found that the best predictor for when scientists produce exceptional contributions is when they produce the most research papers. But there's a catch. During those prolific periods, scientists also published a lot of their worst papers.
“Bad ideas are the price of good ideas,” Jeremy told me. “And leaders need to encourage more bad ideas if they want to get to the genius ones.”
“The great irony is in cutting off the left-hand side of the distribution (ie "no 'bad' ideas, please"), you unwittingly eliminate the right-hand side where the genius lies,” says Jeremey.
In her recent acceptance speech for iHeart Radio’s “Innovator Award,” Taylor Swift echoed this idea. According to Swift, you need the courage to embarrass yourself on the way to greatness.
“I really, really want everyone to know, especially young people, that the hundreds or thousands of dumb ideas that I’ve had are what led me to my good ideas,” she said.
Creativity takes courage. Get comfortable embarrassing yourself. Be the first in a meeting to say a bad idea.
And encourage other people to say their bad ideas. Say something like, "this is a dumb idea, but I want to share it in case it sparks something else."
This lowers the quality bar for your team and encourages more lighthearted idea generation versus performative brainstorming.
Linus Pauling was a biochemist who studied what happens in the brain when someone enters creative flow. He took master musicians and put them in a fMRI scanner to study how the brain operates.
Pauling discovered that when master musicians enter a flow state, the mind stops judging.
“The first thing that is going to kill creativity is judgment,” Jeremy told me. "To be more creative and enter a flow state, you need to turn off judgment.”
As Jeremy points out, suspending judgment is a foreign skill in most corporate environments. We aren’t given much training in our careers about how to judge less. It’s the opposite.
Most of the work we do is reviewing, offering feedback, and learning in business school how to apply critical analytical thinking.
While it’s easy to judge, it’s not helpful in the creation process.
But if you want to stand out as a leader, learn how to give your team space to generate ideas in a judgment-free zone.
In the corporate world, it’s often admired to be critical, direct, and quickly interrogate new ideas. There’s a place for this. But not when ideating.
Learn to separate the judgement/critical phase of creative reviews from the creation.
Creatives are sensitive beings. If there is a culture of judgment, you’ll cause them to be too cautious and block your company from real breakthroughs.
If you’re hitting a wall during your ad brainstorming session, you can likely remedy this by increasing your inputs.
“Your imagination is stimulated by unexpected input,” says Jeremy. “It especially likes novel information and new experiences."
You need to get the brain going by absorbing raw information. Review a Swipe File of ads together. Look at new customer reviews. Look at industries far away from your own for inspiration.
It’s here where you can use AI as a brainstorming partner. Instead of using it as a prompt in, idea out machine, think of it as an extension of your creativity.
How can you feed it information and get it to help you think through the creative problem in a new way?
What raw data (such as customer chat logs) can you upload to ChatGPT to kickstart new ways of looking at the problem?
How can you smash together two unrelated information sources in ChatGPT—such as uploading video transcripts of your top-performing ads and a romantic comedy script you find online—and then ask ChatGPT to think of some unexpected ad ideas?
Keep on adding input until the ideas start flowing.
Encourage unstructured research as a part of the ideation process. Ask the team to follow their fascinations, research unrelated topics, and overall bring new raw substance into their thinking.
Subscribe to Thumbstop for the latest insights, templates, and techniques to ship more winning ads on Meta, TikTok, and Youtube.
By entering your info, you'll join 70k+ media buyers and creative minds reading Thumbstop for a weekly dive into everything creative strategy. (Unsub anytime in a click)
It originally appeared in Thumbstop—a free weekly newsletter filled with tips to help you ship winning Meta, TikTok, and YouTube ads.
By entering your info, you'll join 25k+ media buyers and creative minds reading Thumbstop for a weekly dive into everything creative strategy. (Unsub anytime in a click)
