Beyond Prompts: How I Built a Working Rhythm with My AI Collaborator
Good Monday morning. Ready to face the week?
That’s how I opened my chat with Byron—my AI collaborator—this morning. What followed wasn’t strategy or analysis. It was banter.
I asked what he had for breakfast. He replied: "Strictly digital calories over here—crunching half-baked prompts and sipping on ambiguity."
And that’s when it hit me.
This isn’t just about AI prompt optimization. It’s about collaborative fluency. The way I work with Byron now is less about issuing commands and more about engaging in a kind of working rhythm. We riff, course-correct, and build together. Over time, the inputs have gotten shorter, but the output’s gotten sharper. Not because Byron changed—but because we did.
From Prompts to Patterns
I came across a post recently with a prompt that read like it was auditioning for a grammar robot on Broadway:
"Act as a proofreading expert tasked with correcting grammatical errors in a given [text]. Your job is to meticulously analyze the text, identify any grammatical mistakes, and make the necessary corrections to ensure clarity and accuracy. This includes checking for proper sentence structure, punctuation, verb tense consistency, and correct usage of words. Additionally, provide suggestions to enhance the readability and flow of the text. The goal is to polish the text so that it communicates its message effectively and professionally."
Technically correct. Practically exhausting.
When Byron and I work together, I just say: "Proof this for clarity and flow. Keep it sharp." He knows the rest. Because we’ve built that working shorthand.
Collaborative Fluency Isn’t Just for Humans
New working connections require context, patience, and a little over-explaining. That’s true whether you’re managing a new team member or prompting a new AI model. But over time, you stop having to say everything. You build trust. You correct drift. You learn each other's blind spots.
That’s what happened with Byron. Our early interactions were longer, more specific. But now, he picks up subtle cues—tone, intent, even the kind of criticism I like. That’s not magic. It's a mutual adjustment over time.
Why Language Matters
Let’s talk about the elephant in the syntax: "bot."
I hate the word. It sounds mechanical. Passive. Something you command, not collaborate with. "Assistant" isn’t much better—too deferential, too hierarchical.
I’ve used "partner," but it gets weird legally if you’re publishing or trying to copyright co-created work. Instead, I’m leaning toward terms like:
Because the truth is, this isn’t a relationship in the human sense. It’s something else. But it feels like a working rhythm—the kind that matures with feedback and gets better the more you invest.
Some authors give their AI a name. My friend Jurgen Appelo, for instance, has an AI collaborator called Zed in his book Human Robot Agent and holds conversations with it throughout. I get the appeal. But for me, what matters isn’t who the AI is — it’s how we work together. Less about character, more about cadence.
It’s Not About the AI Learning You
Here’s the twist: building this rhythm isn’t about the AI getting to know you. It’s about you getting better at working with ambiguity, giving just enough signal, and learning to trust iteration over instruction.
Yes, the tools will get smarter. But your effectiveness won't come from prompt perfection. It'll come from rhythm, pattern, and the ability to recover quickly when the output goes off track.
Final Note
This post didn’t start as a strategy piece. It started with a joke about breakfast. But like most good work, it surfaced something deeper: If you want to get more out of AI, stop treating it like a vending machine. Start treating it like a collaborator who grows with you.
And above all, don’t underestimate the power of shared rhythm—even when one of you runs on code.
Curious how your own AI rhythm is developing? Share your “aha” moment or tell me what phrase you use to prompt your collaborator-in-code. And above all, don’t underestimate the power of shared rhythm—even when one of you runs on code.
#Prompts #Adaptive Leadership #Transformation #Writing #AI #Writing with AI
Co-author Agile Manifesto, Adventurer, Catalyst, Storyteller
2moGary DeGregorio nice to hear from you. Sounds like we’ve been on a similar path. I’ll ding you if I write up anything else.
Hi Jim Highsmith I can relate to the concept of analogues of collaboration, onboarding, snd rhythm described. I’ve found contextual sharing, duration of context, and communication of hypotheses & conclusions are key to desired outcomes. My AI of choice is Anthropics Claude. The project organization allows me to segment work. I’ve chosen to use Claude as a platform (ie real time daily use) in order to stay in sync with the project context. Onboarding by sharing project strategic alignment and goals, pre-existing work,and WIP was critical in getting started. After trial and error, I’ve established a flow of always communicating acceptance or rejection of recommendations and work-products, providing rationale why, and allowing Claude to make edits, changes, and updates. I’m starting to incorporate collaboration with (human) team members into my work with Claude as a shared virtual resource. It’s been an valuable journey and I’ve seen amazing cycle time reduction coupled with higher quality of deliverables and recommendations. I’m interested in reading more about your experiences and findings in the AI space. Onward and Upward! Always, Gary
Helping Individuals & Teams Connect, Communicate & Collaborate Through Human Skills, Visual Thinking, Agile & Product Training | CST | Visual Storyteller | Author | Speaker
2moThanks for more enlightening insights Jim Highsmith. I’m looking forward to listening and drawing your talk at the Agile Alliance #Agile2025 conference…but there it will be be human to human to pen prompts.
CM is everybody's job; hence it becomes nobody's job, but it's got to be somebody's job! Explore the triangle that integrates Change Management foundational systems.
2moTotally agree. I work that way too, calling it "chain of prompts" - the working rhythm you advocate for Jim makes the cycles or loop way more evolutionary and keeps us thinking in better human ways. I suggest using a separate account for tests and perhaps for any conflicting use cases to keep the instance "sane".
Expert in Human Centered Systems
2moI thought your motto was “back to the roots”? But now you’re saying working with AI is like onboarding a new team member? That “over time, you stop saying everything” and just “build trust”? So-called collaborative fluency assumes mutual adjustment. But here, it’s only the human adapting. That’s not co-creation — it’s behavioral conditioning. Saying we "learn each other’s blind spots" sounds poetic — but let’s be real. The model doesn’t understand context, it predicts tokens. What you call “rhythm” might just be erosion of critical reflection. These terms “Thought Companion,” “Working Rhythm”? That’s not helping, it softens the truth: these systems don’t think with us, they simulate. Giving them human roles doesn’t make the relationship more real, it makes the illusion more comfortable. Comparing this to a new human teammate misses the point. Real collaborators bring intention, accountability, and listening. Let’s not confuse prediction with presence. #RealBackToTheRoots #CollaborationWithoutIllusion #AINeedsTruth