A company wants a customer support assistant that answers questions from its own product manuals, which change every month. The assistant must cite the manual it used and must not require retraining when the manuals change. Which approach meets these requirements with the least effort?

Use retrieval-augmented generation (RAG) over the manuals with a foundation model on Amazon Bedrock

Fine-tune a foundation model on the manuals each month

Train a new model from scratch on the manuals

Increase the model's temperature so it recalls more detail

A generative AI application confidently returns a policy number and a date that do not exist. What is this behaviour called, and which control most directly reduces it?

Hallucination; ground the model's answers in retrieved, verified documents

Overfitting; add more training data

Bias; rebalance the training set

Drift; retrain the model on recent data

How hard is the AWS AI Practitioner exam?

It is a foundational exam and the easiest of the AWS AI credentials, but it is not a giveaway: the questions are scenario-based and test whether you can pick the right service or technique for a described need. People with some AWS background typically pass after four to eight weeks of part-time study.

Do I need to know how to code?

No. There is no coding on the exam. You need the concepts and the AWS service landscape, not implementation.

How long should I study?

Plan for 30 to 50 hours if you are new to AWS, 15 to 25 if you already hold the Cloud Practitioner. Sit a diagnostic first and let the result set the plan.

How much does it cost and how long is it valid?

About US$100 at the time of writing, with regional pricing, and the credential is valid for three years. Check the AWS certification page for the current fee.

The AWS Certified Machine Learning Engineer, Associate, if you want to build; or the Solutions Architect, Associate, if you want the broader cloud role. The AI Practitioner counts for nothing on either, but the shared vocabulary makes both faster.

All certifications

Amazon Web Services · AIF-C01 · Foundational

AWS Certified AI Practitioner

The entry-level AWS credential for AI, ML and generative AI on AWS. Vendor-neutral concepts plus the AWS services that implement them.

Free practice: 269 questions written to the official exam guide, with the diagnostic, practice papers, timed mocks, a study plan and an explanation on every answer.

Official exam page

Find out where you stand on AIF-C01

20 questions drawn in the exam’s own domain weights. You get a report straight away: your score against the pass mark, where the points went, and what to study first.

Questions: 20
Time: about 20 minutes
Cost: free

Your email

Your answers save as you go, and the report stays in your record. A new email needs no code.

· your answers stay on this device and you can save the report at the end.

The course

The diagnostic shows where you stand. Then you learn the syllabus chapter by chapter, answer that chapter’s questions, and move on. When every chapter is mastered, you sit full timed mocks in the real exam’s shape.

15 chapters, about 11.5 hours in all. Each is a 45-minute study session with quick checks, then a 10-question quiz on that chapter.

Fundamentals of AI and ML20% of the exam

AI, ML and the words the exam uses45 min
Where AI fits, and the AWS AI services45 min
The AI/ML lifecycle, MLOps and metrics45 min

Fundamentals of Generative AI24% of the exam

How generative AI works45 min
Agentic AI: agents, tools, memory and MCP45 min
What GenAI is good and bad at, and how to choose a model45 min
AWS services for building GenAI, and what they cost45 min

Applications of Foundation Models28% of the exam

Designing FM applications: selection, parameters, RAG and customisation45 min
Prompt engineering and Prompt Management45 min
Training and fine-tuning foundation models45 min
Evaluating foundation models and FM applications45 min

Guidelines for Responsible AI14% of the exam

Building AI responsibly45 min
Transparent and explainable models45 min

Security, Compliance and Governance for AI14% of the exam

Securing AI systems45 min
Governance and compliance for AI45 min

About the AWS Certified AI Practitioner (AIF-C01) exam

The AWS Certified AI Practitioner is the entry-level AWS credential for people who work with AI, machine learning and generative AI on AWS without necessarily building models themselves: product managers, analysts, sales engineers, consultants and developers moving into AI. It tests whether you understand what the technology does, which AWS service does it, and how to use it responsibly and securely.

It is a foundational exam. You do not need to write code, tune a model or know the mathematics. You do need to know the vocabulary well enough to tell a foundation model from a fine-tuned one, retrieval-augmented generation from prompt engineering, and Amazon Bedrock from Amazon SageMaker, and to pick the right one for a described situation.

Most people who pass have spent four to eight weeks with the material. The exam guide's 'recommended six months of exposure to AI on AWS' is a description of the typical candidate, not a requirement.

The exam at a glance

Questions: 65 questions (50 scored, 15 unscored), multiple choice, multiple response, ordering, matching and case study
Time: 90 minutes
Pass mark: 700 out of 1,000 (scaled)
Exam fee: About US$100
Valid for: 3 years
Where you sit it: Pearson VUE test centre or online proctored
Prerequisites: None. AWS suggests six months of exposure to AI or ML on AWS

As published in the vendor’s exam guide at the time of writing. Fees vary by country and change; the vendor’s page is the source of truth.

The fee and how to sit it in your country: Sri Lanka · India · Pakistan · Bangladesh · Nepal · Australia · New Zealand · Philippines · Indonesia · Vietnam

What is on the exam, domain by domain

Fundamentals of AI and ML
Definitions and distinctions: AI, ML and deep learning; supervised, unsupervised and reinforcement learning; the ML lifecycle from data to monitoring; when ML is the wrong tool.
20% of the paper
Fundamentals of Generative AI
How foundation models work at the level of tokens, embeddings, prompts and context windows; what they are good and bad at; the AWS services that serve them.
24% of the paper
Applications of Foundation Models
The largest domain. Choosing a model, prompt engineering, retrieval-augmented generation, fine-tuning versus training, agents, and evaluating output. Expect scenarios.
28% of the paper
Guidelines for Responsible AI
Bias, fairness, explainability, hallucination, model cards, human oversight, and the AWS tools that support them.
14% of the paper
Security, Compliance and Governance for AI
IAM, encryption, data residency, guardrails, audit and the shared responsibility model as they apply to AI workloads.
14% of the paper

How to prepare

Read the official exam guide first and write the five domains and their weights on one page. Applications of Foundation Models is 28% of the paper: give it 28% of your time.
Learn the services by what problem they solve, not by their feature lists: Bedrock (use a foundation model), SageMaker (build and train), Q (assistant), Comprehend, Rekognition, Transcribe, Translate, Polly, Lex and Kendra (one job each).
Do the free AWS Skill Builder course for the exam, then spend the rest of your time on practice questions with explanations. This exam rewards recognising the right service for a scenario, and that is a practice skill.
Sit a timed diagnostic before you study, so you know which domains are already fine and which are not. Then study the gaps, not the whole syllabus.
Book the real exam when you clear your target on two practice papers in a row. One good paper is luck; two is readiness.

Sample questions

Three questions in the exam’s style, written by Square 1 against the published domains. They are ours, not the vendor's. The answer and the reasoning are under each one; the practice papers on this page draw from a bank written the same way, with an explanation on every question.

Question 1 · Applications of Foundation Models
A company wants a customer support assistant that answers questions from its own product manuals, which change every month. The assistant must cite the manual it used and must not require retraining when the manuals change. Which approach meets these requirements with the least effort?
1. Fine-tune a foundation model on the manuals each month
2. Use retrieval-augmented generation (RAG) over the manuals with a foundation model on Amazon Bedrock
3. Train a new model from scratch on the manuals
4. Increase the model's temperature so it recalls more detail
Answer: B. RAG retrieves the relevant passages at question time and hands them to the model, so updated manuals are used immediately and the answer can cite its source. Fine-tuning and training bake knowledge into weights and would need repeating every month; temperature controls randomness, not knowledge.
Question 2 · Fundamentals of AI and ML
A retailer has years of labelled transaction records marked 'fraud' or 'not fraud' and wants a model that flags new transactions. Which type of machine learning is this?
1. Unsupervised learning
2. Reinforcement learning
3. Supervised learning
4. Generative AI
Answer: C. The training data carries the answer (the label) for every example, and the goal is to predict that label for new examples. That is supervised learning, specifically classification. Unsupervised learning has no labels; reinforcement learning learns from rewards; generative AI produces new content.
Question 3 · Guidelines for Responsible AI
A generative AI application confidently returns a policy number and a date that do not exist. What is this behaviour called, and which control most directly reduces it?
1. Overfitting; add more training data
2. Hallucination; ground the model's answers in retrieved, verified documents
3. Bias; rebalance the training set
4. Drift; retrain the model on recent data
Answer: B. A fluent, confident, false answer is a hallucination. Grounding (retrieval-augmented generation with citations) gives the model verified text to answer from and lets the output be checked. Overfitting, bias and drift are real problems, but they describe different failures.

Questions people ask before they book

How hard is the AWS AI Practitioner exam?: It is a foundational exam and the easiest of the AWS AI credentials, but it is not a giveaway: the questions are scenario-based and test whether you can pick the right service or technique for a described need. People with some AWS background typically pass after four to eight weeks of part-time study.
Do I need to know how to code?: No. There is no coding on the exam. You need the concepts and the AWS service landscape, not implementation.
How long should I study?: Plan for 30 to 50 hours if you are new to AWS, 15 to 25 if you already hold the Cloud Practitioner. Sit a diagnostic first and let the result set the plan.
How much does it cost and how long is it valid?: About US$100 at the time of writing, with regional pricing, and the credential is valid for three years. Check the AWS certification page for the current fee.
What comes after it?: The AWS Certified Machine Learning Engineer, Associate, if you want to build; or the Solutions Architect, Associate, if you want the broader cloud role. The AI Practitioner counts for nothing on either, but the shared vocabulary makes both faster.