What this course is, and what it is not
This course prepares you for the AWS Certified AI Practitioner (AIF-C01) exam. Read these three points before you start:
- Independent. This course is not produced, endorsed or reviewed by AWS.
- Version. It follows the AIF-C01 exam guide version 1.1, published April 30, 2026. AWS says guide updates are published roughly one month before they show up on the exam.
- No pass promise. The course teaches to the guide. It cannot guarantee a result.
Do this now: open the official exam guide on docs.aws.amazon.com, find the Revisions page, and confirm 1.1 is still the latest version.
Step 1: Learn the exam facts
| Item | Fact |
|---|---|
| Total questions | 65 |
| Scored questions | 50 |
| Unscored questions | 15 (not identified on the exam) |
| Time | 90 minutes |
| Score range | Scaled, 100–1,000 |
| Minimum passing score | 700 |
| Result | Pass or fail |
Three scoring rules to act on:
- Answer everything. Unanswered questions are scored as incorrect, and there is no penalty for guessing.
- Treat every question as scored. You cannot tell which 15 are unscored.
- You pass on the overall score. Scoring is compensatory: you do not need a passing score in each section.
A scaled score is not a percentage. 700 does not mean 70% correct; scaling equates exam forms of slightly different difficulty.
Tip (practitioner advice, not from the guide): 90 minutes across 65 questions is about 83 seconds each. Practise at that pace.
Step 2: Know the four question types
- Multiple choice: one correct response and three distractors.
- Multiple response: two or more correct responses out of five or more options. You must select all the correct ones to get credit.
- Ordering: a list of 3–5 responses to complete a task. You select the correct responses and place them in the correct order.
- Matching: a list of responses to match with 3–7 prompts. All pairs must be correct.
No type gives partial credit as the guide describes them. One wrong pair or one missed selection loses the question.
Step 3: Weight your study by domain
| Domain | Share of scored content |
|---|---|
| 1. Fundamentals of AI and ML | 20% |
| 2. Fundamentals of GenAI | 24% |
| 3. Applications of Foundation Models | 28% |
| 4. Guidelines for Responsible AI | 14% |
| 5. Security, Compliance, and Governance for AI Solutions | 14% |
Do this: split your study hours in these proportions. Domains 2 and 3 together make up 52% of scored content.
Step 4: Note what is out of scope
The target candidate uses AI/ML solutions on AWS but does not necessarily build them. The guide lists these job tasks as out of scope (the list is non-exhaustive):
- Developing or coding AI/ML models or algorithms
- Implementing data engineering or feature engineering techniques
- Performing hyperparameter tuning or model optimization
- Building and deploying AI/ML pipelines or infrastructure
- Conducting mathematical or statistical analysis of AI/ML models
- Implementing security or compliance protocols for AI/ML systems
- Developing and implementing governance frameworks and policies for AI/ML solutions
The pattern: you may be asked to describe or identify these things, but not to do them. If you catch yourself studying code or formulas, stop.
Step 5: Place the five big terms
Task statement 1.1 asks you to explain how AI, ML, deep learning, GenAI and agentic AI relate. Think of nested scopes, like an AWS Organization, an account inside it, and a VPC inside that.
- Artificial intelligence (AI): the broadest field. Systems that perform tasks normally needing human intelligence, such as understanding language or recognising images. It includes rule-based systems that learn nothing.
- Machine learning (ML): a subset of AI. Systems learn patterns from data instead of following hand-written rules.
- Deep learning: a subset of ML. It uses neural networks with many layers. A neural network is a set of connected nodes arranged in layers, loosely modelled on neurons in the brain.
- Generative AI (GenAI): a subset of deep learning. It creates new content (text, images, audio, code) rather than only classifying or predicting.
- Agentic AI: an extension of GenAI, not a deeper subset. An agent uses a model to plan, call tools and take multi-step actions toward a goal with limited human direction.
Example using one support desk:
- A fixed rule routes tickets containing "refund" to billing: AI in the broad sense, no ML.
- A model trained on past tickets predicts the category: ML.
- A model drafts a reply to the customer: GenAI.
- A system reads the ticket, looks up the order, issues the refund and sends the reply: agentic AI.
Two application fields also appear in the guide's list:
- Computer vision: AI that interprets images and video.
- Natural language processing (NLP): AI that interprets and produces human language.
Step 6: Match the core terms
Cover the right column and recite each definition. This is the format of a matching question.
| Term | Definition |
|---|---|
| Algorithm | The procedure used to learn patterns from data |
| Model | The output of training: what you get when an algorithm has learned from data, and what you use to make predictions |
| Training | The process of running an algorithm on data to produce a model |
| Inferencing | Using a trained model to produce a prediction or output from new input |
| Bias | Systematic error that skews a model's results, often unfairly toward or against a group, commonly from unrepresentative training data |
| Fairness | The goal that a model treats individuals and groups equitably |
| Fit | How well a model captures the patterns in its data. Overfit: memorises training data and performs poorly on new data. Underfit: too simple to capture the pattern at all |
| Large language model (LLM) | A very large deep learning model trained on huge amounts of text to understand and generate language |
An analogy you already know: the algorithm is the build process, the training data is the source, and the model is the built artifact. Inferencing is invoking the deployed artifact. Calling a model in Amazon Bedrock is inferencing; you are not training anything.
What the exam asks about this
- Expect scenario stems that describe a system and ask which term fits best. Choose the most specific correct term: a system that writes text is AI, but the best answer is GenAI.
- Expect training and inferencing to be confused deliberately. Ask: is a model being produced or used?
- Expect bias (the problem) and fairness (the goal) as paired distractors.
- The exam facts in Steps 1–4 are for your planning. The exam tests the domain content, not its own blueprint.
Key takeaways
- 65 questions, 50 scored and 15 unscored, 90 minutes, scaled passing score 700 on a 100–1,000 scale.
- Four question types: multiple choice, multiple response, ordering, matching. Always answer; guessing is not penalised.
- Domain weights: 20%, 24%, 28%, 14%, 14%. Applications of Foundation Models is the largest.
- Coding models, feature engineering, hyperparameter tuning, building pipelines and statistical analysis are out of scope.
- AI contains ML, which contains deep learning, which contains GenAI. Agentic AI extends GenAI with planning and action.
- An algorithm plus data, through training, produces a model. Inferencing uses the model.
- This course is independent of AWS, follows exam guide version 1.1, and promises no pass.