If the first lesson made the case for fluency, this one hands you the map. Anthropic's framework breaks working with AI into four habits, the four D's, and almost every problem you will hit with a team sits inside one of them. Watch the two short videos, then I will put the four into plain terms for a working team.
The first video makes the case. The second introduces the framework the rest of the course is built on.
The four D's, in practice
Here is how I would translate each one for a team that actually has to deliver with these tools:
- Delegation. Deciding what to hand to AI and what to keep. Most teams either give it nothing or give it everything. Both cause problems.
- Description. Telling the AI clearly what you want, with enough context to get something useful back. This is the skill that improves fastest with practice.
- Discernment. Judging the output. Knowing when it is good enough, when it is subtly wrong, and when to start again.
- Diligence. Owning the result. Checking facts, being honest about what AI helped with, and staying responsible for what goes out under your name.
They build on each other. Weak delegation makes description harder, and weak discernment quietly lets mistakes through.
Try this
Pick a recent piece of AI-assisted work and walk it through the four D's with your team. You will usually find one that nobody is really doing. Start there.
Common questions about the AI fluency framework
Why does AI fluency matter for managers?
Having access to AI is not the same as knowing how to use it well. Most of us now have a capable assistant a click away, but that does not stop us getting an unexpected response, struggling to explain what we actually need, or wondering whether the data we just pasted in is safe. Fluency closes that gap so a team uses AI in a way that is effective, efficient, ethical and safe, rather than relying on a handful of prompt tricks that go out of date.
What goes wrong when a team is not AI fluent?
Almost every problem a team hits with AI sits inside one of the four D's. Someone hands the AI the wrong task, briefs it vaguely, fails to catch a weak answer, or sends something out without checking it. Naming which D broke, rather than blaming the tool, is how you fix it.
Do the four D's of AI fluency build on each other?
Yes. They are not four separate boxes to tick. Weak delegation makes description harder, because you are briefing the wrong task in the first place, and weak discernment quietly lets mistakes slip through to diligence. That is why the weakest of the four is usually where a team's AI problems are really coming from, and the most useful place to start.