These five rules form the foundation of my approach to working with AI. They represent lessons learned from extensive use and help ensure consistent, high-quality results. Each rule addresses a common mistake that limits AI effectiveness.
AIFM: AI First Mindset
Start every task by asking "How can AI help?"
The most successful AI users don't treat AI as a special tool for special occasions. They default to AI assistance and opt out when needed, not the reverse. This mindset shift is the single biggest predictor of AI effectiveness.
The Principle in Practice
Before AIFM: "I need to write this report. Maybe I'll use AI if I get stuck."
After AIFM: "I need to write this report. Let me start with AI to outline the structure, draft the sections, and refine the language. Then I'll add my expertise and judgment."
Why It Matters
Most professionals dramatically underutilize AI because they only think of it for certain tasks. An AI-first mindset means:
- Brainstorming: AI before blank page anxiety
- Research: AI to synthesize before deep-diving
- Writing: AI for first drafts, you for final voice
- Analysis: AI to find patterns, you to interpret meaning
- Communication: AI for structure, you for relationship
The Opt-Out Framework
AIFM doesn't mean using AI for everything. It means consciously choosing. Ask:
- Can AI help with this task? (Usually yes)
- Should AI help with this task? (Consider sensitivity, relationships, learning value)
- What part of this task benefits most from AI? (Often the production, not the thinking)
For educators: Think "How can AI help me with this?" rather than "Is it appropriate to use AI here?" The tool is available. The question is how to use it well, not whether to use it.
AI Is Smart: Trust the Model
Don't over-specify or micromanage. Give context, not step-by-step instructions.
Top-tier AI models are remarkably capable. They perform at PhD-level on many reasoning tasks. Treating AI like a simple command-line tool limits what it can do for you. Treat it like a capable colleague instead.
The Principle in Practice
Over-specified prompt: "Write an email. First, include a greeting. Then write a sentence about the field trip. Then write a sentence about the date. Then write a sentence about what to bring. Then write a closing. Then add my signature."
Context-driven prompt: "Write a warm, informative email to 3rd grade parents about our field trip to the science museum next Friday. Include the key details: date, cost ($15), permission slip due Wednesday, pack a lunch. Match my communication style: friendly but professional."
Why It Matters
When you micromanage AI:
- You do most of the thinking anyway
- The output feels mechanical and choppy
- You miss AI's ability to synthesize and improve
When you provide context and trust:
- AI brings its knowledge and patterns to bear
- Outputs feel natural and well-structured
- You get suggestions you wouldn't have thought of
The Colleague Test
Would you give these instructions to a capable colleague? If your prompt reads like assembly instructions, you're micromanaging. If it reads like a briefing, you're collaborating.
For educators: Instead of "Write an email to parents" (too vague) or step-by-step instructions (too controlling), try "Write an email to parents of 3rd graders about our upcoming field trip. Tone should be warm but informative. Include date, cost, permission slip deadline, and what to pack."
Be Curious: Experiment Constantly
Try different approaches. Learn from failures. Stay current.
AI is evolving rapidly. What didn't work six months ago might work brilliantly today. What works with one model might fail with another. Curiosity and experimentation are essential skills.
The Principle in Practice
Fixed approach: "I always use ChatGPT for writing. It's good enough."
Curious approach: "Let me try this task with Claude and GPT to see which handles it better. I'll experiment with different prompt structures. I'll try the new model that just released."
Why It Matters
The difference between AI tiers is substantial:
| Tier | Models | Capability | | ---- | ----------------------------------------- | ------------------------------- | | A- | Opus 4.5, GPT-5.2 Pro, Gemini 3 Pro | PhD-level reasoning (~90% GPQA) | | B- | Sonnet 4.5, Gemini 3 Thinking | Senior professional (~83% GPQA) | | C- | Haiku 4.5, GPT-5.2 Instant, Gemini 3 Fast | Junior professional (~73% GPQA) | | F | GPT-4o, Sonnet 3 | Intern level (~53% GPQA) |
Using the wrong model is like sending an intern to do a partner's job. Model selection alone can double your output quality.
The Experimentation Habit
- Weekly: Try one new prompt technique
- Monthly: Test a model you haven't used recently
- Quarterly: Review whether your tool choices still make sense
For educators: Use premium models when drafting board presentations, analyzing policy implications, or making recommendations. Use fast models for quick questions or simple formatting tasks. Match the model to the stakes.
Builder Mindset: Create Systems, Not Just Outputs
Build tools, workflows, and content. Don't just consume AI's outputs.
The most powerful AI use involves moving beyond single prompts to building systems. Instead of asking AI to do a task once, build a system for handling that category of tasks.
The Principle in Practice
Consumer approach: "Write me a parent email about the field trip."
Builder approach: "Help me create a parent email template system. I want consistent structure, adjustable components for different event types, and a checklist of required information."
Why It Matters
Single prompts give you one output. Systems give you:
- Consistency: Same high quality every time
- Efficiency: Faster execution of common tasks
- Scalability: Handle more work with less effort
- Improvement: Refine the system over time
What to Build
| Category | One-time Prompt | System to Build | | -------- | ---------------------- | ------------------------------------------------- | | Emails | Write this email | Email template library by type | | Meetings | Summarize this meeting | Meeting summary format with action tracking | | Reports | Write this section | Report structure with reusable components | | Lessons | Create this lesson | Lesson planning workflow with standards alignment |
The Investment Mindset
Building systems takes more time upfront. But the return is enormous. A one-hour investment in a good system saves hundreds of hours over time.
For educators: Rather than writing one parent email, build a template system. Rather than summarizing one meeting, create a consistent format for all meeting summaries. The investment pays dividends.
Delta X (Validation): Verify What Matters
Always verify outputs. AI is confident but not infallible. Human judgment remains essential.
AI doesn't know when it's wrong. It generates plausible text with equal confidence whether the content is accurate or hallucinated. Your job is to verify what matters.
The Principle in Practice
No validation: Accept AI output, use immediately.
Calibrated validation: Match verification effort to stakes.
| Criticality | Validation Level | Example | | ----------- | ---------------------------------- | ------------------------------------------------------- | | High | Full review, fact-check all claims | Board presentations, legal documents, public statements | | Medium | Review for accuracy and tone | Internal communications, draft policies | | Low | Quick scan for obvious errors | Brainstorming, internal notes, first drafts |
Why It Matters
AI fails in predictable ways:
- Hallucinations: Confidently stating false information
- Outdated information: Knowledge cutoff limitations
- Context blindness: Missing local or specific factors
- Bias: Reflecting patterns in training data
None of these failures announce themselves. The AI sounds equally confident when right and wrong.
The Verification Checklist
For high-stakes content, verify:
- [ ] Statistics and numbers
- [ ] Quotes and attributions
- [ ] Legal or policy references
- [ ] Historical claims
- [ ] Current events (AI knowledge has cutoffs)
- [ ] Local context accuracy
The Expert Rule
Don't try to be an expert where you're not an expert.
AI can make you more effective in your areas of expertise. It cannot substitute for expertise you don't have. If you're not qualified to verify an output, get someone who is.
For educators: A brainstorming session for professional development topics needs less validation than a letter to parents about safety incidents. Calibrate accordingly. Use AI to amplify your educational expertise. For legal questions, consult legal counsel. For medical questions, consult healthcare professionals.
Putting It All Together
These five rules work as a system:
- AIFM (AI First Mindset) gets you using AI consistently
- AI Is Smart ensures you use it effectively
- Be Curious keeps you learning and improving
- Builder Mindset multiplies your impact
- Delta X (Validation) maintains quality and trust
Master these principles, and your AI-assisted work will be genuinely indistinguishable from your best human-only efforts, but accomplished in a fraction of the time.
Action Items
This week:
- Apply AIFM to three tasks you wouldn't normally use AI for
- Try one task with a premium model instead of your usual choice
This month:
- Build one reusable system (template, workflow, or prompt library)
- Develop your calibrated validation habit
This quarter:
- Review your AI toolkit and update model choices
- Share your best systems with colleagues