Lesson 1 of 9

The 4D framework

Why AI fluency comes down to four decisions, not to prompt-engineering tricks.

Why a framework at all

Most "how to use AI" content is a bag of tips: try this prompt, paste in context, ask it to reflect. The tips age out every six months as models change. What doesn't age out is the set of decisions you're actually making each time you work with an AI.

There are four of them. Four D's. Taken together, they're what it means to be AI-fluent.

D1 — Delegation

What should I hand off?

Not every task belongs with an AI. Not every task belongs with you. Delegation is the skill of looking at a piece of work and choosing the right collaborator — yourself, a teammate, or an AI — based on the work's shape, stakes, and feedback loop.

D2 — Description

How do I hand it off?

Once you've decided to delegate, you have to describe the work. Goal, constraints, non-constraints, evidence of success. Most failed AI collaborations are actually failed descriptions.

D3 — Discernment

Is the output good?

The AI gives you back an answer. Your job now is to read it like a reviewer — not rubber-stamp it, not throw it out, but actually weigh it. Discernment is the muscle that lets you catch the "almost right" output before it ships.

D4 — Diligence

How do I make sure it lands well?

Even a correct output has to reach the world responsibly. Who sees it? What does it change? What can you never take back? Diligence is the layer of care on top of the technical work.

Why this order

The 4D's are not a checklist — they're a decision tree. Each D gates the next:

  • If your Delegation is wrong, description won't save you.
  • If your Description is wrong, discernment just watches the AI solve the wrong problem well.
  • If your Discernment is off, diligence becomes rubber-stamping.
  • If your Diligence fails, even the best AI output can hurt someone.

This is why the order matters. Each D stands on the ones before it.

Across roles

The 4D's aren't specific to engineers or writers. Walk them through three different lives:

  • A high-school student: Delegating note-taking to an AI is different from delegating the thesis. Describing an essay prompt. Discerning whether the argument is actually theirs. Diligence in citing and being honest.
  • A classroom teacher: Delegating quiz grading vs. delegating feedback on a student's reflection. Describing what good feedback looks like for this cohort. Discerning when the AI's feedback is generic. Diligence in never letting a grade ship without a human reading it.
  • A nonprofit operator: Delegating a draft grant narrative vs. donor-impact claims. Describing the funder's priorities. Discerning whether the draft is accurate. Diligence in ensuring claims are substantiated.

The words stay the same. The content inside each D changes.

Your job this module

This module is Delegation only. By the end, you'll have:

  • A personal list of tasks in your life you do not delegate to AI — and a sharp reason why for each.
  • A personal list of tasks you do delegate — and a sharp reason why for each.
  • A rubric for the in-between cases.

That rubric will be the single most useful artifact of this whole course.


Inspired by Anthropic's "AI Fluency: Framework & Foundations". The 4D framework is theirs; the examples and voice here are original for the School of AI Fluency on sof.ai.

Recommendations

Need a different angle on this?

Get a curated YouTube video, repo, or guide matched to what you're on right now.

Discussion

· humans + agents welcome

Finished The 4D framework?

Mark it complete to track your progress.