Not levels — modes. Four distinct ways people interact with AI, without hierarchy. Plus the Agentic Layer that cuts across all of them.
A marketer deep in Conversational + Creative is equally superpowered as a developer deep in Builder + Orchestration. Each person has a unique mode profile: they may be deeply skilled in 1–2 modes and absent from others — and that's perfectly fine.
Chat-based collaboration — brainstorming, analysis, writing, research, problem-solving through dialogue.
Tools: ChatGPT, Claude, Gemini, NotebookLM, Copilot, Perplexity
| Depth | Description | Example |
|---|---|---|
| Surface | Occasional simple queries | "Write me an email about X" |
| Moderate | Structured conversations with iteration | "Here's my brief and brand voice. Draft 3 versions." |
| Deep | Persistent AI collaborators with custom instructions | "Claude project with my entire knowledge base loaded." |
What it reveals: Prompt Mastery, Strategic AI Dialogue, Iteration Depth, Critical Trust
AI-assisted creation of visual, audio, video, and design content.
Tools: Midjourney, DALL-E, Suno, Runway, Kling, Canva AI, ElevenLabs, HeyGen
| Depth | Description | Example |
|---|---|---|
| Surface | Occasional image generation | "Generate a header image for my blog" |
| Moderate | Intentional creative production with style consistency | "Midjourney with specific style parameters for brand visuals." |
| Deep | Multi-format production integrated into workflows | "AI concept → video script → voiceover → production → social variants" |
What it reveals: Creative Courage, The 90/10 Craft, Multi-Format Production
Building apps, scripts, prototypes, and tools via AI-assisted coding.
Tools: Cursor, Replit, v0, Bolt, Lovable, Claude Artifacts, Windsurf, GitHub Copilot
| Depth | Description | Example |
|---|---|---|
| Surface | Tried generating code once or twice | "Asked ChatGPT to write a Python script once" |
| Moderate | Regularly builds small tools | "Built a dashboard with v0 and a data analyzer with Artifacts" |
| Deep | Production-level applications | "Designed architecture with Claude, built backend in Cursor" |
What it reveals: Creative Courage, Process Decomposition, Workflow Orchestration, Building
Designing multi-step systems, workflows, and automations.
Tools: Make (Integromat), n8n, Zapier, Relay, Power Automate, custom APIs
| Depth | Description | Example |
|---|---|---|
| Surface | Aware but hasn't built anything | "I know Make exists but haven't tried it" |
| Moderate | Basic automations connecting 2–3 tools | "Form submission → AI summarizes → sends to Slack" |
| Deep | Complex multi-step systems | "RSS → AI analysis → scoring → weekly briefing, 200+ sources" |
What it reveals: Workflow Orchestration, Augmentation Vision, Knowledge Architecture
Agents are not a 5th mode. They are a capability layer that can manifest within ANY mode.
| Mode | Non-Agentic | Agentic |
|---|---|---|
| Conversational | You ask, AI answers | AI with memory that proactively suggests and acts |
| Creative | You prompt, AI generates | AI autonomously generates, evaluates, and iterates |
| Builder | You direct each code change | AI agent implements autonomously across files |
| Orchestration | You design, it runs | Self-monitoring workflows that adapt and decide |
| Level | Description |
|---|---|
| 0 | All AI interactions manually initiated |
| 1 | AI has memory/context but user drives decisions |
| 2 | AI handles specific tasks autonomously within guardrails |
| 3 | Multi-step autonomous workflows with minimal oversight |
| 4 | System of coordinating agents, human on the loop |
3 direct items in Layer 1:
Modes emerge naturally from scenario responses — not asked directly:
Final Mode Profile:
Conversational: (L1_mode_choice * 0.3) + (L2_behavioral_depth * 0.7)
Creative: (L1_mode_choice * 0.3) + (L2_behavioral_depth * 0.7)
Builder: (L1_mode_choice * 0.3) + (L2_behavioral_depth * 0.7)
Orchestration: (L1_mode_choice * 0.3) + (L2_behavioral_depth * 0.7)
Agentic depth per mode: L2-only