
The four moves behind every answer in this test (Verify, Match, Gate, Escalate) and the trap version of each.
Questions 1 to 5 on Domain 1, Prompting and Task Execution, with every answer option explained.
Questions 6 to 13 on Domain 2, Output Evaluation and Validation, with every answer option explained.
Questions 14 to 17 on Domain 3, Product and Model Selection, with every answer option explained.
Questions 18 to 23 on Domain 4, Workflow Integration and Solution Design, with every answer option explained.
Questions 24 to 27 on Domain 5, Configuration and Knowledge Management, with every answer option explained.
Questions 28 to 32 on Domain 6, Governance, Risk, and Responsible Use, with every answer option explained.
Questions 33 to 36 on Domain 7, Troubleshooting and Optimization, with every answer option explained.
A recap of the four moves on one screen, and how to read your score: which move you missed and in which domain.
The CCAO-F does not test what Claude can do. It tests what you do next.
Every question is a small workplace decision. Someone hands you an output, or a deadline, or a customer spreadsheet, and the exam asks which move you make. There are four moves. Verify a claim against its source. Match the tool or the model to the task. Gate the data before it moves. Escalate when the call is not yours. That pattern comes out of Anthropic's own published sample questions, and once you can see it, the answer options stop looking equally reasonable.
The course follows one person through one work week. Maya runs marketing ops. Her manager wants speed. Compliance wants a signature before anything goes out. Every objective in the blueprint turns up as something she has to decide before lunch. Concepts get named after you have watched them happen, not before, because the exam is written entirely in scenarios and the teaching should match.
Twelve lectures, a little under two hours in total, in blueprint order, covering all seven scored domains:
Prompting and Task Execution, 14 percent of the exam
Output Evaluation and Validation, 21 percent, split across two lectures because it is the heaviest domain
Product and Model Selection, 12 percent
Workflow Integration and Solution Design, 16 percent
Configuration and Knowledge Management, 12 percent
Governance, Risk, and Responsible Use, 15 percent
Troubleshooting and Optimization, 10 percent
Three more lectures sit around those seven. A question lab that works eight mixed scenarios drawn from every domain. An exam-day lecture on the multiple-response mechanic, two-pass pacing, and what a fixed pass line means for how you answer. A capstone that runs one launch day through all seven domains in a single continuous scenario.
What the exam itself looks like: 60 items in 120 minutes, multiple-choice and multiple-response, with every item stating how many responses to select. Scoring is scaled 100 to 1000 and 720 passes. The fee is 99 USD and the certification is valid for 12 months.
There is no practice test in this course. Questions are worked on screen instead, with each wrong option struck and the reason given, and every correct answer stamped with the move it used. You are never asked to pause and guess.
This is the associate track, for people who use Claude in their job. Ops, marketing, project management, communications. If you build against the API, the developer and architect tracks are the ones that test your work.
Registration for this exam currently runs through Anthropic's partner network, so check the current registration page before you schedule anything.