Using Generative AI as a "Sounding Board" Made Brainstorming New Ideas Surprisingly Easier

When you're trying to come up with a new project or improvement idea, have you ever noticed yourself circling the same handful of thoughts over and over just by thinking it through alone? In my work handling front-desk services and inter-departmental coordination, I often used to get stuck at the stage of drafting an initial proposal. Here's a concrete look at how brainstorming changed once I started using generative AI (ChatGPT and Claude) as a "sounding board" — failures included.

1<What Got Me Started>

It started when I noticed how inefficient it was to think through a proposal alone. Even locked away in a meeting room, I'd only ever end up generating ideas within patterns I already knew, left with nothing but the nagging doubt, "is this really the right approach?" At the same time, I felt awkward constantly running half-formed ideas past colleagues. Hitting that "limit of thinking alone" is what pushed me to try using generative AI as a sounding board.

2<What I Actually Tried>

The first thing I tried was explaining the background and goal of a proposal to the AI, then asking it to raise questions and pushback from both a "supportive" and an "opposing" perspective. I made a point of asking not just "what are the benefits of this idea" but also "what bottlenecks would come up in implementing it" and "if someone opposed this, what reasons might they give."

I also started listing out my ideas as bullet points and asking the AI to "sort these along two axes: ease of implementation and expected impact," using that as a rough draft for prioritizing them.

3<The Tweak That Worked>

The single biggest factor was giving the AI a clearly defined "role." Asking for opinions while specifying a stance — "you're a cautious manager" or "you're a frontline staff member" — produced concrete, role-appropriate concerns instead of generic platitudes.

It also helped a lot not to take the AI's first answer at face value, and instead always push for at least one more round — "can you be more specific about that reason," "now consider it from the opposing angle." Not stopping after a single exchange let me repeatedly surface points I wouldn't have noticed on my own.

4<What Went Wrong>

Early on, I sometimes took the concerns or objections the AI raised too much at face value, as if they were "the objectively correct answer." The AI is only ever presenting generic patterns — it doesn't actually know the realities of my specific workplace. I've made a point of remembering that the final judgment call is always mine to make.

I also came close to writing out internal information still under discussion, or the names of people involved, when explaining the background of a proposal. Since the brainstorming works fine even when the purpose is described in the abstract, I now make sure to leave out proper nouns when consulting the AI this way.

5<Results>

With these changes, the time it takes me to get from a blank slate to a rough draft proposal has roughly halved, by my own estimate. More than anything, the feeling of "going in circles alone" has disappeared — being able to stress-test an idea from multiple angles the moment it occurs to me has also eased the mental burden of the process itself.

6<Summary>

The tips for using generative AI as a sounding board come down to three things:

- Give the AI a concrete role (supporter, opponent, frontline staff, etc.) when asking for its opinion

- Don't stop at one answer — keep pushing with follow-up questions

- Treat the AI's opinions as a rough draft, and always make the final call yourself

It feels less like "having it think for me" and more like "having it surface perspectives I wouldn't have reached alone" — and that shift has eased my own reluctance around planning itself, at least a little. I hope this is useful to anyone else who feels stuck brainstorming alone.

#AIatWork #GenerativeAI #ChatGPT #Claude #Brainstorming #Ideation #Productivity