The AI Fluency Framework names four competencies people use when they work with AI. This explainer follows one illustrative ticket, a database migration, and adds a mark to it for each competency.
Figure 1
One ticket, four competencies
Illustration, not a real ticket. Definitions: The AI Fluency Framework, © 2025 Rick Dakan, Joseph Feller and Anthropic, CC BY-NC-SA 4.0.
- 01 · Delegation
The owner decides that an agent may draft this migration and that a person reviews the draft. The ticket says so before anyone writes a prompt.
- 02 · Description
Four lines give the agent the goal, the limit, what done looks like and what to leave alone. The reviewer checks the draft against the same four lines.
- 03 · Discernment
The draft passes all four lines and still fails review: nobody wrote down a way back. The owner adds a rollback line and sends the agent back.
- 04 · Diligence
The owner signs the change and answers for the rollout. The agent does not sign.
Delegation: decide whether an agent drafts at all
The framework defines Delegation as “Setting goals and deciding whether, when and how to engage with AI.” On a platform team that decision comes before any prompt: may an agent draft this migration, and who reviews what it produces? The ticket records the answer: the agent drafts and the owner reviews.
What does the agent get to read?
Description: four lines the reviewer can check later
“Effectively describing goals to prompt useful AI behaviors and outputs.” The description is the spec the agent works from: the goal, the limit, what done looks like and what to leave alone. Each line is something the reviewer can hold the draft against.
The draft comes back. Is it any good?
Discernment: check the draft against the description
“Accurately assessing the usefulness of AI outputs and behaviours.” The draft passes all four lines and still fails review, because it has no way back and nobody asked for one. The owner adds a rollback line to the description and sends the agent back.
Who answers for it in six months?
Diligence: a person signs and answers for it
“Taking responsibility for what we do with AI and how we do it.” The owner signs the change, owns the rollout and answers for it later. The agent does not sign.
The same order holds on any task an agent drafts: a person delegates, the agent drafts, a reviewer checks, the owner signs.
Figure 2
Hand-off: the owner delegates and signs, the agent drafts
Concept, not measurement. Characters are roles (Open Peeps, CC0).
- 01 · Delegate
The owner hands the agent a goal and its limits.
- 02 · Draft
The agent returns a draft, not a decision.
- 03 · Review
A reviewer checks the draft against the goal.
- 04 · Sign
The owner signs and answers for the result.
Drafted with AI assistance; checked and edited by Second Author.
Keep reading
People-led AI combines employee leadership with shared direction
People closest to the work lead improvements, supported by shared priorities, connected tools and governance.
Faster agents do not shorten a review queue
If changes arrive faster than your reviewers clear them, the queue grows no matter how fast the agent wrote them. Changing who reviews what shortens it.