There’s still a lot I get to think through when it comes to focusing on the why of my work in the agentic AI era.
It bears repeating and reviewing. I don’t think there’ll be a more important topic to revisit. It isn’t so much a distinction line between humans and AI as much as it is more like a horizon sphere that expands and contracts with each evolution of human-AI interaction.
So.
I asked Claude AI for clear, actionable guidance regarding people and AI labor and management.
Here’s part of what it said:
The question behind the question. The most powerful move in human and AI interaction is the same: ask what’s really being asked. With AI, that means prompting it to reflect your question back before answering. With humans, it means slowing down enough to say “is this the right question?”
Specificity is empathy. Specific, contextual questions signal you’ve actually thought about the other party. This is true whether you’re talking to a colleague or AI.
Make your network smarter, not just faster. The real leverage is designing loops where humans and AI keep making each other’s questions sharper. Prompt libraries, fail logs, and “what would make this delightful?” rounds are all mechanisms for that.
Joy is a design constraint, not a bonus. Energized, curious people ask better questions than exhausted ones. Building play into question design is practical, not frivolous.
Claude referenced interactive cards as part of this output, but that part of the output failed. So I tried again:
Questions as infrastructure. The most powerful thing you can do is design contexts (meetings, tools, rituals, prompts) where better questions are what naturally emerge. That’s the shift from practitioner to leader.
The “both” tag matters. Attend carefully to what kind of answer you actually need, and create the conditions for it to surface.
Joy is a legitimate design criterion. It tends to get treated as decoration or a nice-to-have. Ask what would make something delightful to surface functional improvements that purely functional questions miss.
Claude failed to generate the interactive part of the output again, so I tried a third time:
The meta-question habit. The single highest-leverage practice across everything above is asking “what question should I be asking?” before asking any object-level question. It works on humans (“what do you wish I’d asked?”), on AI (“what context am I failing to give you?”), and on yourself (“what am I avoiding?”).
Stakes before facts. Most questions get answered at the surface — what happened, what’s needed, what’s next. The richer layer is always why it matters. Leading with stakes (“this is consequential because…”) changes what humans volunteer and what AI calibrates toward.
Adversarial use of AI is underexploited. Most people use AI to produce. Using it to stress-test, resist, and poke holes — “what’s the strongest case against what I just said?” — is where it generates the most distinctive value for anyone with diagnostic sensibility.
Joy is key. Questions asked from genuine curiosity are structurally different from questions asked from obligation. Building rituals and language that keep curiosity alive — in yourself, your network, and your AI interactions — is the long-run compounding asset.
Claude then successfully generated a robust, interactive system of 7 subject cards of concrete, actionable guidance with opportunities to delve deeper into each subject—none of which I asked for yet remains intriguing for future consideration. I’ll give each subject it’s own blog post in the future. For now, a comment on today’s human-AI interaction with Claude.
Claude AI’s incognito mode is becoming quite useless.
While I can’t confirm the veracity of the X post I talked about, it is showing signs of at least some truth. Something’s going on with Claude specifically and perhaps Anthropic broadly. I’ve felt it over the past month.
I plan to create another “Month 2 summary as deliverable” of I release the first one (hopefully by next week). Until then, I can only mentioned what I’m experiencing intuitively. Anyway.
For now, here are the 7 subjects Claude produced cards for (in part, admittedly, at my behest):
- AI literacy – Keeping current with AI isn’t about knowing everything; it’s about knowing what changes what.
- Collab tools – Tools shape the questions that get asked. The best ones create shared context; the worst create silos.
- Empathy – the substrate of good questions — you can’t ask what matters to someone without modeling their inner world.
- Human-machine basics – The quality of your question determines the quality of the machine’s answer. Prompt design is question design.
- Joy in work – Joy isn’t decoration — it’s signal. It marks where energy is alive and questions are genuine rather than performative.
- Mental health – Good questions protect cognitive bandwidth. Better questions come from people — and machines — with clear, calm minds.
- Network leadership – Good network leaders move ideas and questions through human webs — they’re brokers, not just connectors.