What you will be able to do
- Explain Delegation as an outer-loop decision that starts by breaking a problem into steps
- Sort the steps of an existing process into ones Claude can take on and ones that stay with a person
- Write concrete, gradeable statements of what "done" means before handing any step to Claude
Key concept
Delegation — Delegation is the AI Fluency competency of breaking a piece of work into its steps and then deciding, step by step, which ones Claude takes on and which stay with you. You make that call before any prompt is written.
1.Delegation is an outer-loop decision
Using Claude to improve a process starts with a question about the process itself, not with a prompt. The AI Fluency 4D Framework names four competencies: Delegation, Description, Discernment and Diligence. It groups them into two loops. The inner loop is Description and Discernment, which guide each conversation with Claude. The outer loop is Delegation and Diligence, which guide the bigger decisions about when and how to use AI at all.
Delegation is the entry point. In the framework's words, Delegation means deciding what AI handles and what stays with you. Diligence closes the outer loop: you own the final result fully, however much of it Claude produced. Taken together, the two set the boundary for process work. Claude can take over steps, but accountability for the outcome stays with a person.
The builders' course sharpens this into a method. Delegation is not "should I use AI here?" It is a decomposition: take the real problem, break it into steps, then assign a role for Claude at each step. A process you want to optimise is already a set of steps. The work is to look at each one separately instead of handing over "the process" as a single block.
Checkpoint 1 of 4· Check yourself
An operations manager wants to "use Claude for our onboarding process." Under the 4D Framework, what should come first?
Delegation means decomposing the problem first and then deciding what AI handles at each step. Prompting (Description) and evaluating output (Discernment) come after that decision.
“Delegation means decomposing the problem first, then deciding what AI handles at each step.”Source: academy.claude.com
2.Which steps to hand off, and which to keep
The AI Fluency course for creative work splits Delegation into three ordered moves. Problem Awareness comes first: be clear about the problem you are actually solving before you reach for a tool. Platform Awareness comes next, because you cannot decide what to give a tool you do not understand. Only then does Task Delegation assign the work. The same course names a common mistake: reaching for AI when you do not know what you want.
Checkpoint 2 of 4· Put it in order
Put the three Delegation moves in the order the AI Fluency course gives them.
- 1.Task Delegation: assign the work
- 2.Platform Awareness: understand the tool you might hand work to
- 3.Problem Awareness: be clear on the problem you are actually solving
The course says Problem Awareness comes first, Platform Awareness comes next, and then Task Delegation assigns the work.
“Platform Awareness comes next: you cannot decide what to hand to a tool you do not understand.”Source: academy.claude.com
Which steps are good candidates? The builders' course draws the main line: delegating implementation is safe, and delegating judgment is not. Anthropic's guidance on Claude Cowork adds a practical checklist for spotting work worth handing off. More than one thing goes in (several files, a folder, connected apps). A file comes out. You will do it again, since recurring tasks are the sweet spot. You already know what good looks like. And the middle is the boring part. The thinking happens at the start (deciding what you want) and at the end (deciding whether it is right). The extract, compile, reconcile and reformat work in between is what you hand off.
| Signal | Points toward handing it off | Points toward keeping it |
|---|---|---|
| Inputs | More than one thing goes in: multiple files, a folder, connectors | A single input that chat handles fine |
| Frequency | Recurring: you'll do it again | A true one-off |
| Knowing good | You can tell quickly whether output is right, wrong, or 70% there | You can't yet describe what a right answer looks like |
| Where the thinking is | The middle: extract, compile, reconcile, reformat | The start (deciding what you want) and the end (deciding if it's right) |
Checkpoint 3 of 4· Exam question
A market researcher needs Claude to incorporate current, real-world information published this week into a competitive-landscape summary, since this data postdates the model's training. Which capability should the researcher enable?
Correct answer: C — The web search tool, which augments Claude's knowledge with current data from across the web
- A. Prompt caching lowers cost and latency for repeated context; it does not fetch new external information.
- B. The memory tool persists information across sessions but does not retrieve fresh external web content.
- C. Correct. The web search tool augments Claude's comprehensive knowledge with current, real-world data from across the web, which is needed for this week's information.
- D. Structured outputs controls response formatting, not access to current external information.
3.Define "done" and the collaboration mode before you delegate
The builders' course asks for acceptance criteria before any work starts, because they give you and Claude a shared definition of done. Its example of a good one is "A patient with a basic smartphone can find the current wait time in under 30 seconds without creating an account." Its example of a bad one is "The system is easy to use." The same habit appears in the 4D exercise: before opening AI, write down what "good enough" looks like for the output. The educators' capstone asks the same before any task: note what done looks like, and which parts AI can help with and which stay yours.
The same planning step assigns each delegated step a collaboration mode. Automation means AI does it and you check. Augmentation means you and AI work on it together. Agency means AI operates with latitude inside boundaries you set. A step with a sharp definition of done and an easy check suits Automation. A step that needs your judgment partway through suits Augmentation. Agency only makes sense once you have defined the boundaries.
Checkpoint 4 of 4· Match them up
Match each AI collaboration mode to how the work is shared.
Tap a term, then the definition that fits it.
These are the three modes the builders' course uses when mapping each part of the work to a role for AI.
“Automation (AI does it, you check), Augmentation (you and AI work it together), or Agency (AI operates with latitude inside boundaries you set)”Source: academy.claude.com
Exam traps
Each one states something that sounds right. Open it to see what is actually true.
1.If Claude can do a step well, the judgment about whether that step should exist can be delegated too.Why is that wrong?
Handing off the mechanical work is safe. Handing off the judgment is not, so decisions about changing a process stay with the people who own it.
Covered in Which steps to hand off, and which to keep
2.You can decide what a good result looks like after you see Claude's first output.Why is that wrong?
Concrete definitions of done come before the work, so you and Claude share the same target from the start.
Covered in Define "done" and the collaboration mode before you delegate
Sources
Every claim above is drawn from one of these pages, quoted as it was written on the date shown.
- 1.
“The outer loop (Delegation and Diligence) guides bigger decisions about when and how to use AI.”
↩︎ Delegation is an outer-loop decision“Delegation means deciding what AI handles and what stays with you.”
↩︎ Delegation is an outer-loop decision“Diligence means owning the final result fully”
↩︎ Delegation is an outer-loop decision“Discernment means evaluating what AI gives back before acting on it.”
↩︎ Which steps to hand off, and which to keep - 2.https://academy.claude.com/courses/ai-fluency-for-builders/delegation-the-builder-s-toolkitOfficial docs
“Delegation means decomposing the problem first, then deciding what AI handles at each step.”
↩︎ Delegation is an outer-loop decision“Explain why delegating implementation is safe and delegating judgment is not”
↩︎ Which steps to hand off, and which to keep“Write acceptance tests before code. They give you and AI a shared definition of done.”
↩︎ Define "done" and the collaboration mode before you delegate“Automation (AI does it, you check), Augmentation (you and AI work it together), or Agency (AI operates with latitude inside boundaries you set)”
↩︎ Define "done" and the collaboration mode before you delegate“Delegation means decomposing the problem first, then deciding what AI handles at each step.”
↩︎ Key concept“Explain why delegating implementation is safe and delegating judgment is not”
↩︎ Exam trap 1“Write acceptance tests before code. They give you and AI a shared definition of done.”
↩︎ Exam trap 2“Each must be concrete enough that a stranger could tell you whether the finished product passes or fails.”
↩︎ Prediction - 3.https://academy.claude.com/courses/ai-fluency-for-creative-work/delegation-and-diligenceOfficial docs
“Problem Awareness comes first: be clear on the problem you are actually solving before reaching for a tool.”
↩︎ Which steps to hand off, and which to keep“reaching for AI when you do not know what you want is the immediate one.”
↩︎ Which steps to hand off, and which to keep“Platform Awareness comes next: you cannot decide what to hand to a tool you do not understand.”
↩︎ Checkpoint - 4.
“Everything in between (extract, compile, reconcile, and reformat) is what you hand off.”
↩︎ Which steps to hand off, and which to keep“One-offs are fine, but recurring tasks are the sweet spot.”
↩︎ Which steps to hand off, and which to keep - 5.
“which parts AI can help with and which stay yours”
↩︎ Define "done" and the collaboration mode before you delegate