I’m going to admit something: I’m super impatient when it comes to prompting. I want refined, substantive responses from AI, and I also just want the tools to read my mind and get on with it.
Of course I know that effective human-AI collaboration doesn’t work like this (yet). And at Bodine & Co., we train all of our employees on crafting effective prompts. But still, I’m impatient.
It reminds me of the pushback I often get about human-centered design:
- Client: “But doesn’t all the customer research and journey mapping and process slow us down?”
- Me: “Yes. And it dramatically increases the odds that you’ll wind up with something that moves the needle for both your customers and your business.”
Prompting works the same way. But did I mention I’m impatient? 😅
So yes, I cut corners here and there with my prompts, usually to my dismay. That’s why, when I had the seed of a strategic initiative that I wanted AI’s help with, I pinged my colleague Jess. She possesses a deep well of patience — and, not coincidentally, mad prompting skills.
And we tried something new: We created a prompt together.
When Prompting Becomes a Team Sport
Co-prompting with Jess was exactly what I expected. She tempered my impatience. She brought yes-and energy, elevating my initial idea. She pushed for more clarity when I was ready to “just see what it says.”
We debated the context.
Clarified the role we wanted the AI to play.
Expanded the list of actions we wanted it to take — a list that got longer and better as our discussion unfolded.
The AI output was stronger because of our collaboration.
But our session also surfaced a deeper tension that I’ve been sitting with.
The 1-to-AI Problem in AI Collaboration
Today’s AI tools are built around a 1-to-AI interaction model: one person, one interface, one thread of thought. Microsoft is furthest along in embedding AI into team workflows, and the other giants are all working on it in their own ways. But what’s still missing is shared prompting as a first-order design principle — multiple humans actively shaping one AI interaction together, in real time, without workarounds.
I was able to share my ChatGPT project with Jess, but our conversations immediately branched. I controlled the original thread, and Jess controlled the branch that started from where I shared it with her.
Now, I understand that branching is a useful construct and software feature. But in a collaborative group setting, branching can quietly reinforce silos — and Jess and I experienced this firsthand. When we stayed in a single shared prompt, we were building on a common artifact. When we branched, we drifted into parallel exploration. Productive, yes. But separate. Instead of co-creating, we were independently iterating.
And that wasn’t the only friction.
ChatGPT did a terrible job of supporting our collaboration. Here’s what we actually needed to make the session work:
- Zoom for conversation and screen sharing
- A shared doc to co-create our prompt (because it’s much easier than typing in ChatGPT’s tiny text box and accidentally hitting enter too soon)
- Another shared doc for pasting the output into (because we each wanted separate control over reading and digesting the response — screen sharing didn’t cut it)
- Our own self-directed orchestration of the session
The overall experience felt clunky and NOT AT ALL designed with collaboration in mind. 🤦♀️
From 1-to-AI to Many-to-AI
What if prompting became a collective practice, like whiteboarding or journey mapping? Co-prompting showed me how powerful that could be. It also exposed how poorly our tools support it today.
If the future of work is teams collaborating with AI, we need to move beyond 1-to-AI models toward a many-to-AI approach. Not one human working with AI as a personal assistant, but multiple humans working together with AI as a shared canvas.
Every AI platform is racing to become our operating system of choice. Collaboration features will come. (Probably much faster than mind reading.)
In the meantime, though, we have choices to make:
- Are we satisfied with adapting our team workflows to tools optimized for individuals?
- Do we want AI products built with human collaboration at the center?
- And perhaps more importantly: As we design the future of AI at work, are we designing for isolated efficiency — or collective intelligence?
I’d love to hear how you’re collaborating (or at least trying to!) with humans and AI.
