Claude Certified Associate - Foundations course
8 modules, one per exam domain. 48 lessons, each opening with its key concepts; 6 timed practice tests.
One purchase covers the lessons and the tests.
What does the CCAO‑F course cover?
Every module is one of the 8 exam domains Anthropic publishes, in exam order, with the domain's share of the exam as its bar. Output Evaluation and Validation and Workflow Integration and Solution Design carry the most at 21% and 16%. The exam runs 120 minutes for 60 questions (all scored). Open a module for its lessons; the practice tests close the course.
Write a request Claude can act on, supply the context only you hold, and correct a draft in turns instead of starting again.
14% of the exam · Exam domain: Prompting & Task Execution
Writing a Task Claude Can Act On
Everything you leave unsaid becomes a guess, so name the action, the reader, the shape and the boundary.
- 1.1
What a Weak Prompt Leaves Out
A weak prompt is unspecified rather than short · Name the action, the reader, the format and the boundary · State anything two sensible people would answer differently · Say what to do when the material runs short · Precision costs a line, not a page
20 minHigh priority
- 1.2
Give Claude the Role and the Context It Lacks
A role sets tone, structure and priorities at once · A role cannot supply facts Claude has never seen · Supply what a colleague starting today would ask · Keep the background block apart from the request · Paste the clause rather than the whole handbook
20 minHigh priority
- 1.3
Show an Example of the Output You Want
One example fixes shape, length and tone together · Two examples separate the pattern from the wording · Include the awkward case with a missing value · Examples must obey the rule you wrote · A repeated format belongs in the project
20 minCore topic
Working in Turns
Treat the first answer as a draft, place long material where the request stays visible, and know when the thread itself is the problem.
- 1.4
Iteration Beats the Perfect First Prompt
The conversation already holds the material and the request · Name the fault and the target, not the feeling · Say which part to preserve · Ask Claude to critique its own answer first · Change one thing per turn while the draft still needs work
20 minHigh priority
- 1.5
Long Inputs: Where to Put the Document
Material first, request last, on anything over a page · Say to use only the document above · Ask for the quote that supports each point · One-off material in the chat, shared material in the project · A paid project moves to search mode as it fills
20 minHigh priority
- 1.6
When to Start a New Chat Instead
An instruction from turn two still applies at turn thirty · Correct answers inside the thread, change subjects outside it · A new chat in the project keeps knowledge and instructions · Memory topics carry context between chats · An incognito chat leaves nothing behind and cannot reopen
20 minCore topic
Module assessment
Drawn from the lessons above, a different set each attempt.
Read an answer critically, verify the claims that carry risk, and decide what standard the work has to meet before it ships.
21% of the exam · Exam domain: Output Evaluation & Validation
Reading an Answer Critically
Fluency carries no accuracy signal, so pick what to check by what a sentence asserts.
- 2.1
Fluent and Wrong Look the Same
A verified fact and an invented one read identically · Hedging reflects style rather than certainty · Asking Claude if it is sure adds nothing checkable · The most specific sentence carries the most risk · Rank claims by damage before you start checking
20 minHigh priority
- 2.2
Claims That Need a Source
Numbers, names, quotes and rules carry the risk · Plausibility is what an invented figure already has · Check first where cost is high and effort is low · A citation names a source without proving it · Most sentences restate common ground and need nothing
20 minHigh priority
- 2.3
Checking Numbers and Arithmetic
Add one column and recompute one percentage · A percentage without its base can mislead while correct · Units that shift mid column produce plausible totals · Ask for the working, then read it · A self-check repeats the process that made the error
20 minHigh priority
Verifying Against the Real World
Open the source, understand what research mode changes, and sample when the volume makes full checking impossible.
- 2.4
Open the Source Before You Trust the Citation
A citation is a claim about a source · It may not exist, may not mention it, or may say otherwise · Context around a quote can reverse its meaning · Ask for the supporting sentence to make checking cheap · Existence is the second of four verification steps
20 minHigh priority
- 2.5
What Research Mode Changes About Checking
Research searches the web and connected context, with links · Links prove pages exist, not that a claim matches · One citation rarely covers a whole paragraph · Judge the publisher before you count citations · Research runs on Pro, Max, Team and Enterprise
20 minHigh priority
- 2.6
Sampling: How Much Do You Check
Full checks for the few, samples for the many · An unfinishable policy becomes no policy · Sample across the batch, not just the first items · A fault in a sample is evidence about the batch · Let the observed error rate set the rate
20 minCore topic
Deciding What Ships
Match the standard to the consequence, name who signs off, and record the check so it survives a challenge.
- 2.7
The Accuracy the Audience Requires
Disposable, internal, external and binding · The reader and the consequence set the standard · One standard for everything gets abandoned · Polish reflects the prompt, never the accuracy · Re-check whenever work crosses from internal to external
20 minHigh priority
- 2.8
Where a Human Has to Sign Off
Verification and accountability are separate gates · People, money, safety and the irreversible need a name · A team name means nobody approved it · Give the approver the claims and the sources · A gate that covers everything protects nothing
20 minHigh priority
- 2.9
Recording the Check So It Can Be Repeated
One line per claim: claim, source, verdict, checker, date · Sources change, so a verdict has a shelf life · Record only the claims you actually checked · Keep the register beside the work · A register answers a challenge in minutes
20 minCore topic
Module assessment
Drawn from the lessons above, a different set each attempt.
Choose the model by what a wrong answer costs, then send the work to the part of the product that suits it.
12% of the exam · Exam domain: Product & Model Selection
Choosing the Model
The picker changes every few months, so learn the trade it presents rather than the names on the list.
- 3.1
Choose Speed or Depth by What the Answer Is Worth
The cost of a wrong answer decides how much depth to buy · Fast turns suit work you were going to read anyway · Long or tangled source material rewards a deeper model · Usage tracks length, complexity, features and model · Switch as the task changes rather than once a week
20 minHigh priority
- 3.2
Read the Model Picker Without Memorising the List
Heavier options sit above lighter ones in every picker · The description line is written for people outside model releases · A version number ranks releases inside one family only · Compare your current entry against its neighbour · Check the release notes when the lineup changes
20 minCore topic
Choosing the Surface
Chat, project, research, artifact, plan tier and device each solve a problem the others handle badly.
- 3.3
Send the Work to the Right Place in the Product
A project repays you on the second visit, never the first · Research reads live sources and hands back the links · An artifact is a thing you can edit, keep and share · Retrieval turns on as a project nears its knowledge limit · Free accounts hold up to five projects
20 minHigh priority
- 3.4
Know What a Paid Plan Buys You
Five tiers: Free, Pro, Max, Team and Enterprise · A five hour session limit and a weekly limit run together · Research, chat search, retrieval and Cowork need a paid plan · Settings and then Usage shows both progress bars · Usage credits extend Pro and Max beyond the included allowance
20 minHigh priority
- 3.5
Pick the Surface by What Claude Needs to Touch
Sessions follow the account across web, desktop and mobile · Only the desktop app reaches local files and its own browser · Web and mobile borrow that reach while desktop stays open · Cowork carries out a job with many steps · Cowork runs on Pro, Max, Team and enabled Enterprise
20 minCore topic
Module assessment
Drawn from the lessons above, a different set each attempt.
Find the steps in a real workflow worth moving, design the joints between people and Claude, then prove the saving on a sample.
16% of the exam · Exam domain: Workflow Integration & Solution Design
Finding the Work Worth Moving
A job is a chain of steps, and only some of them qualify. One of them has to stay with a person.
- 4.1
Find the Steps in a Workflow That Suit Claude
Move steps rather than whole jobs · The best candidate is heavy on text and fast to check · Checking time counts against the saving · Volume decides whether the design work repays you · Undocumented knowledge means the step stays put
20 minHigh priority
- 4.2
Name the Step That Has to Stay Human
Ask who answers for the outcome · Decisions that bind the organisation stay with a person · Anything affecting somebody materially needs an appeal route · Claude prepares the decision and never makes it · Name a role rather than a department
20 minHigh priority
Designing the Flow
Split the work into checkable steps, learn what a connector really reaches, and design both handovers.
- 4.3
Split the Work Into Steps You Can Check
A step you cannot check is not a step · Splitting makes a failure locatable · Steps become reusable, giant prompts do not · One verb per step · Stop splitting once each step is checkable
20 minHigh priority
- 4.4
Know What a Connector Actually Reaches
Verified means quality and compatibility, never a security audit · Web connectors reach services, desktop extensions reach your machine · Gmail exposes attachment metadata and not the content · Drive extracts text and skips embedded images · Team and Enterprise owners enable a connector first
20 minCore topic
- 4.5
Design the Handover in Both Directions
A handover carries what happened, what did not and what to decide · Standing rules belong in project instructions · State the gaps, because silence looks like completeness · Attach evidence so the receiver checks one line · End with the question rather than the conclusion
20 minCore topic
Proving It Works
Pilot on a sample loaded with awkward cases, then report a saving that survives a finance review.
- 4.6
Pilot on a Sample That Includes the Awkward Cases
Write the pass condition before the first case runs · Load the sample with cases that already went wrong · Keep it small enough to read every result · Run the old way on the same cases · Log every intervention as part of the cost
20 minHigh priority
- 4.7
Measure the Saving You Can Defend
Measure the loop rather than the step · Review time moved, it did not disappear · Exception cases carry the full old cost · Spread setup hours across a year of volume · Report the exception rate and the failure mode
20 minHigh priority
Module assessment
Drawn from the lessons above, a different set each attempt.
Curate what a project knows, write instructions that change the answer, and keep the whole workspace working after months of use.
12% of the exam · Exam domain: Configuration & Knowledge Management
Projects
Knowledge holds what is true and instructions hold how to behave. Both go wrong as a project grows.
- 5.1
Decide What Earns a Place in Project Knowledge
Stale knowledge is worse than missing knowledge · Keep one current version of each document · Single-use documents belong in the message · A document with no owner will go stale · Relevance matters more than the capacity limit
20 minHigh priority
- 5.2
Write Project Instructions That Actually Change the Answer
Instructions carry behaviour, knowledge carries facts · An instruction you cannot test is decoration · Name the reader concretely rather than asking for clarity · Say what to do when the knowledge runs short · Test one line at a time, in a fresh chat
20 minHigh priority
- 5.3
Recognise a Project That Has Outgrown Itself
Retrieval turns on near the limit and shows an indicator · Once it searches, wording and file names decide what it finds · Prune before restructuring · Split by audience or workflow rather than by size · Free accounts have no retrieval, so pruning is the only route
20 minCore topic
Memory and Personalization
Three mechanisms you author and one that authors itself, plus the review that stops a workspace rotting.
- 5.4
Tell Memory, Instructions and Skills Apart
Instructions for Claude reach every conversation on your account · Project instructions reach one project · Skills apply on demand for tone, format and expertise · Memory builds itself, so Settings lists topics you can edit · Memory is off by default on Team and Enterprise
20 minHigh priority
- 5.5
Keep a Workspace Working Over Months
Nothing forces a paid account to tidy up · Delete unused projects before pruning anything · Read the memory topics each quarter · Replace a document in the same action that adds its successor · Incognito suits the answer you do not want remembered
20 minCore topic
Module assessment
Drawn from the lessons above, a different set each attempt.
What happens to what you paste, which privacy control answers which question, and where preparation stops and a decision begins.
15% of the exam · Exam domain: Governance, Risk & Responsible Use
Data and Confidentiality
Retention, training and personal data, answered from the published policy rather than from rumour.
- 6.1
Know What Happens to What You Paste
Training is opt-in through Model Improvement · Standard consumer retention is thirty days · Opting in extends retention to as long as five years · Deleting clears history at once and the back end within thirty days · Claude for Work follows a separate commercial policy
20 minHigh priority
- 6.2
Pick the Right Privacy Control for the Job
Incognito leaves no history and cannot be reopened · Incognito is exempt from training whatever your settings say · Incognito does not work inside a project · Model Improvement off keeps the record and blocks training · Deletion acts on retention rather than on training
20 minHigh priority
- 6.3
Handle Personal Data the Way Your Rules Already Require
Obligations follow the data rather than the tool · Removing the name rarely de-identifies a record · Minimise before you paste · The usage policy is a floor, and local rules are usually stricter · Health and belief carry your strictest handling
20 minCore topic
Using the Output Responsibly
Disclosure duties in high-risk domains, the line between preparing and deciding, and a policy colleagues can apply.
- 6.4
Know When You Have to Say Claude Helped
High risk means legal, health, insurance, finance, employment, housing, testing, journalism · High-risk work needs disclosure at the start and qualified review · Consumer-facing chatbots disclose at the start of each session · Elsewhere, disclose when a reader would weigh it differently · Disclosure transfers no responsibility
20 minHigh priority
- 6.5
Separate What Claude May Prepare From What It May Decide
Fourteen prohibited categories, learned by shape · Four of them reach ordinary office work · Preparation is broad and decision is narrow · A ranking becomes a decision the moment you set a cut-off · Impersonating human-written content is a named prohibition
20 minHigh priority
- 6.6
Write a Team Policy People Actually Follow
Three lists beat ten principles · Fill the lists with tasks your team already performs · Name an owner and set a review date · Copying the vendor prohibitions answers nothing anybody asked · Test it by watching somebody apply it
20 minCore topic
Module assessment
Drawn from the lessons above, a different set each attempt.
Isolate whether the prompt, the context or the tool failed, read a refusal properly, and get more out of the plan you already pay for.
10% of the exam · Exam domain: Troubleshooting & Optimization
Diagnosing a Bad Result
Three causes account for nearly every disappointing answer, and one clean test separates two of them.
- 7.1
Find Out Whether the Prompt, the Context or the Tool Failed
A fresh chat is the cheapest single-variable test · Was right and now wrong points at a superseded document · Wrong format usually traces to an earlier instruction · Confident and out of date points at the tool · Change one thing per test
20 minHigh priority
- 7.2
Read a Refusal Instead of Fighting It
Rephrasing to evade is platform abuse · Legitimate professional work loses its context when compressed · Ambiguity gets the cautious reading · A capability limit is not a policy refusal · Some refusals stand, and stopping is a valid move
20 minCore topic
Getting More From the Same Plan
Most limit problems are habit problems, and a prompt that worked is worth keeping.
- 7.3
Get More Done Inside the Limits You Have
A five hour session window and a weekly limit run together · Length and complexity dominate usage · A project caches uploaded material, so you pay once · Batch related requests into one message · Fix habits before buying allowance
20 minHigh priority
- 7.4
Turn a Prompt That Worked Into One You Can Reuse
Copy it verbatim while it is still in front of you · Split the fixed instructions from the case specifics · Test it on a second case before anybody depends on it · Project instructions for every run, a skill for tone · A library nobody opens is another thing to maintain
20 minCore topic
Module assessment
Drawn from the lessons above, a different set each attempt.
Six walk-throughs in the shape the exam uses, each ending in the decision it asks for, with the tempting wrong answers explained.
0% of the exam
Worked Scenarios
One scenario per domain cluster. Read the situation, work it through, then compare your decision with the reasoning.
- 8.1
Scenario: The Announcement That Keeps Coming Back Wrong
Three drafts missing differently means the reader was never named · Tone adjectives no output could fail do no work · Correct in the same thread rather than starting again · Length, framing and action all follow from the reader · One extra sentence replaces three attempts
20 minHigh priority
- 8.2
Scenario: The Statistic Nobody Checked
A claim with no openable source is not a claim yet · Fluent writing borrows the credibility of its neighbours · Asking Claude to confirm produces confidence, never evidence · Softening a number hides it rather than fixing it · The pattern break is the signal worth hunting
20 minHigh priority
- 8.3
Scenario: Where the Monthly Competitor Review Should Live
Work with three needs cannot have one answer · Only research reaches pages that changed recently · A project holds what repeats every month · Single-use material belongs in the message · An upgrade does not fix a thread carrying the whole day
20 minHigh priority
- 8.4
Scenario: The Saving That Did Not Survive Finance
Measure the loop rather than the step · Review time moved and did not disappear · Exception cases carry the full old cost · A pilot on easy cases hides the categories that break · The honest figure was still worth reporting
20 minHigh priority
- 8.5
Scenario: The Project That Was Right Last Spring
Two symptoms can have two different causes · No instruction beats a superseded document that is present · A present document the search cannot reach needs a rename · Splitting answers a different problem from pruning · Projects rot because adding is easy and removing is nobody job
20 minHigh priority
- 8.6
Scenario: The Rejections That Went Out on Their Own
Accuracy does not answer who decided · A score wired to an outgoing letter is the cut-off · Disclosure moves no responsibility · Keep the preparation and restore the signature · Stop first, then unwind what already went
20 minHigh priority
Module assessment
Drawn from the lessons above, a different set each attempt.
Each test mirrors the real exam: 120 minutes, 60 questions, all domains in proportion. Learning mode shows the explanation after each answer; exam mode runs the clock and scores at the end. The study plan below schedules them across the weeks.
- 1
Practice test 1
Full CCAO-F exam simulation #1 covering all seven domains at their real blueprint weights, led by Output Evaluation & Validation (21%) and Workflow Integration (16%). Business-and-productivity scenarios in the style of the live exam: prompting, evaluating outputs, choosing product features, and responsible use.
60 questions · 120 min
- 2
Practice test 2
Full CCAO-F exam simulation #2 with proportional coverage of all seven domains. Emphasizes prompting and task decomposition, output evaluation and validation, and troubleshooting underperforming prompts in real business workflows.
60 questions · 120 min
- 3
Practice test 3
Full CCAO-F exam simulation #3 with proportional domain coverage. Emphasizes product and model selection (Projects, research mode, Artifacts; Haiku vs. Sonnet vs. Opus), workflow integration, and configuring Claude Projects with knowledge sources and connectors.
60 questions · 120 min
- 4
Practice test 4
Full CCAO-F exam simulation #4 with proportional domain coverage. Emphasizes governance, risk, and responsible use (data sensitivity, privacy, appropriate use cases, org AI policy) alongside output validation and workflow integration.
60 questions · 120 min
- 5
Practice test 5
Harder CCAO-F exam simulation #5 with proportional domain coverage and tougher distractors. Multi-step business scenarios spanning configuration, knowledge management, output evaluation, and responsible use that reward careful judgment.
60 questions · 120 min
- 6
Practice test 6
Harder CCAO-F exam simulation #6 with the most nuanced cross-domain business scenarios while staying at foundational scope. Distractors are plausible real approaches that a productivity user might reasonably consider.
60 questions · 120 min
Try 15 free questions on the certificate page before you buy.
How long does it take to prepare with this course?
About 16 hours of lessons, 48 of them at roughly 20 minutes each. The modules run in exam order, so the heaviest domains come first. Pick a pace and the plan lays itself out.
- Week 1Prompting and Task Execution · Output Evaluation and Validation5h of lessons5h
- Week 2Product and Model Selection · Workflow Integration and Solution Design · Configuration and Knowledge ManagementPractice tests 1, 2 and 35h 40m of lessons5h 40m
- Week 3Governance, Risk and Responsible Use · Troubleshooting and Optimization · Exam ScenariosPractice tests 4, 5 and 6 · Schedule the exam once you clear 75% on a fresh test5h 20m of lessons5h 20m
Ready to start the CCAO-F course?
The course and the practice tests come together in one purchase, or go Pro to add 7 build-and-submit projects for this cert plus every other track on Preporato.
Course + practice tests
Lifetime access: pay once, study forever
- 48 interactive lessons with checkpoints
- 6 full-length practice tests
- 360+ exam-style questions
- Detailed explanations for every answer
- Exam mode & learning mode
- ×No hands-on labs
Preporato Pro
Billed monthly, cancel anytime
- 7 build-and-submit projects for CCAO-F, graded instantly
- Every course and practice test across every certification
- GPU sandboxes and hosted environments
- Flashcards, study guides and articles
- Cancel anytime, no contract
CCAO-F course + practice tests
from$19.99course + tests