AI framework skills - Jeff Su Channel
https://www.youtube.com/watch?v=s_765enJBy8 Here is a markdown summary of the video “4 Skills to Get Ahead” by Jeff Su.
4 AI Skills to Get Ahead (Beyond Basic Prompts)
“Using AI” is no longer a differentiator on a resume; it is a baseline expectation, similar to Microsoft Word proficiency. To actually get ahead, you need to build four specific skills on top of that baseline.
1. The Cockpit Rule
This is a mental model for deciding when to delegate to AI, when to collaborate, and when to avoid it entirely. Think of it like a pilot: use autopilot for cruising, but take manual control for takeoff/landing or emergencies.
The Three Modes
- Autopilot Mode: Hand off the task with clear instructions. Trust the output with minimal review.
- Collaboration Mode: You and AI iterate together through multiple rounds. Neither could produce the result alone.
- Manual Mode: Do the work yourself because the risk of error is too high or AI lacks the context.
The Decision Framework (Agentic Cost-Benefit)
To decide which mode to use, weigh these three factors:
- Human Baseline Time: How long would it take you to do it manually?
- Probability of Success: How likely is the AI to get it right?
- AI Process Time: How long does it take to prompt, wait, and check the output?
Rule of Thumb: Delegate tasks where Human Time is high, Probability of Success is high, and AI Process/Checking Time is low.
2. Build the Rails
Because AI is so capable, your competitive advantage is no longer doing the work, but designing the process so AI can do it for you.
The Concept
Like laying tracks for a bullet train, the upfront work is tedious, but once the “rails” (workflow) are built, the work glides with almost no friction.
- The “Pion”: Uses AI with no structure (performed 19% worse in studies).
- The “Centaur”: Clear handoffs between human and AI tasks.
- The “Cyborg”: deeply integrates AI into every step.
How to Redesign Your Workflow
- Break It Down: Decompose a deliverable (e.g., a newsletter) into component steps.
- Classify Each Step: Apply the Cockpit Rule to each step.
- Prioritize Autopilot: Redesign the “Autopilot” steps first for the highest ROI. Example: Instead of one prompt for a whole newsletter, create separate, optimized prompts for the subject line and the body content.
3. The Storytelling Moat
In an AI world, information is a commodity. The real skill is turning that information into meaning. If you just pass along data, you are replaceable. If you turn data into a story that moves people, you are safe.
Framework A: The “ABT” (Randy Olson)
- AND: Sets the stage (Here is where we are).
- BUT: Introduces the conflict (Something is wrong/needs attention).
- THEREFORE: Delivers the resolution (Here is the next step/solution).
Framework B: The “SCQA” (McKinsey/Bain)
- Situation: Here is where we are.
- Complication: Here is the obstacle/conflict.
- Question: What do we need to answer to move forward?
- Answer: Here is the resolution.
Key Takeaway: Both frameworks introduce conflict and then resolve it. This is what makes stakeholders care about your data.
4. Manual Override
Intentionally choosing not to use AI for certain tasks to prevent Cognitive Atrophy.
The Danger
If AI writes every email and summarizes every meeting, your “stabilizer muscles” (critical thinking) weaken. Studies show that over-reliance on AI leads knowledge workers to stop questioning assumptions and checking sources, making them less prepared for edge cases.
Two Habits to Protect Your Brain
- Think First, Prompt Second: For analytical tasks, spend a few minutes forming your own opinion or outline before asking AI.
- Interrogate the Output: Do not accept the AI’s answer immediately. Debate it. Ask: “What are the tradeoffs here?” or “How would I verify this?”
“Your brain is safe. Your thinking, however, is up to you.” — Ethan Mollick