AI Weekly
Guides ·
The 4 habits that separate good AI users from everyone else
Delegation, Description, Discernment, Diligence — Anthropic's AI Fluency framework in plain English, with one thing to try for each habit today.
Two people can use the same chatbot and get wildly different results. The difference usually isn't the model — it's habits.
Anthropic and two university researchers, Rick Dakan and Joseph Feller, put those habits into a framework called AI Fluency, built around four skills they call the 4D Framework. Here's what each one means in practice, and one thing to try for each.
1. Delegation — deciding what AI should do at all
The framework defines it as setting goals and deciding whether, when and how to bring AI in.
Most people skip this step: they either hand AI everything or nothing. Good delegators split the job first — "AI drafts the outline, I write the opening, AI checks the formatting."
Try this: Before your next task, write two lines: what AI does and what I do. Keep anything you'll be held accountable for on your side.
2. Description — telling AI what you actually want
Defined as effectively describing goals to prompt useful AI behavior and output.
"Summarize this" leaves AI guessing about everything that matters. Three extra details fix most bad answers:
- Who will read it
- Where it's going (Slack message? board deck?)
- How long it should be
Try this: When an answer misses, don't start over. Say exactly what's off: "Too formal — rewrite for a teammate, under 100 words."
3. Discernment — knowing when AI is wrong
Defined as accurately assessing the usefulness of AI output and behavior.
This matters because confident wrong answers are built into how today's models are trained. OpenAI researchers argue that models "hallucinate" partly because training and evaluation reward guessing over admitting uncertainty.
Try this: Anthropic's own docs suggest two easy checks — give the model permission to say "I don't know," and ask it to back each claim with a source or a direct quote. No source? Treat the claim as unverified.
4. Diligence — owning what you do with AI
Defined as taking responsibility for what we do with AI and how we do it.
That covers what you put in (customer names, internal numbers, personal data) and what you put out (anything published under your name).
Try this: Before pasting, swap real names, companies and amounts for placeholders like [CLIENT] or [AMOUNT]. Before publishing, check every number, date and quote yourself.
Which habit is your weak spot?
Most people are strong in one or two of these and weak in another. We built a free, 3-minute test around the four habits: your strongest and weakest habit decide which of 15 AI types you are — like The Rubber Stamp (great at delegating, never checks) or The Lone Wolf (catches every error, then does it all alone).
Sources
- The AI Fluency Framework (PDF) — Anthropic, with Rick Dakan and Joseph Feller
- AI Fluency: Framework & Foundations course — Anthropic Academy
- Why Language Models Hallucinate — Kalai, Nachum, Vempala, Zhang (arXiv, 2025-09-04)
- Reduce hallucinations — Claude Docs
Written with AI (Claude) and checked against the linked sources.