# AWS AI Practitioner (AIF-C01)

Independent preparation for AWS Certified AI Practitioner — the blueprint, 70 study points across four note pages, the AWS AI service map, and how this credential differs from the AI Business Strategist exam.

import { Card, CardGrid, Badge } from '@prosefly/astro-components';

<Badge color="accent" variant="soft">AIF-C01</Badge> <Badge color="info" variant="soft">65 items · 90 min</Badge> <Badge color="warning" variant="soft">Pass 700/1000</Badge> <Badge color="success" variant="soft">$100</Badge> <Badge color="neutral" variant="soft">Foundational</Badge>

**AI vocabulary plus the AWS AI service catalogue.** AIF-C01 validates a foundational understanding of AI, ML and generative AI *and* the ability to name the AWS service that fits a use case. You are not expected to code a model, tune hyperparameters or build a pipeline — but you are expected to know what Bedrock is for, what SageMaker AI is for, which pre-trained service handles which modality, and what responsible AI requires of you.

:::caution[Unofficial preparation]
Independent study material, not affiliated with or endorsed by AWS, containing no official exam questions. Verify every figure on this page against the current [AIF-C01 exam guide](https://docs.aws.amazon.com/aws-certification/latest/ai-practitioner-01/ai-practitioner-01.html) before booking.
:::

## The blueprint

| # | Domain | Weight | Where it is taught here |
| --- | --- | --- | --- |
| 1 | Fundamentals of AI and ML | 20% | [AI & ML foundations](/aws/ai-practitioner/notes/01-ai-and-ml-foundations/) |
| 2 | Fundamentals of GenAI | 24% | [Generative AI](/aws/ai-practitioner/notes/02-generative-ai/) |
| 3 | Applications of Foundation Models | **28%** | [Generative AI](/aws/ai-practitioner/notes/02-generative-ai/), [AWS AI services](/aws/ai-practitioner/notes/03-aws-ai-services/) |
| 4 | Guidelines for Responsible AI | 14% | [Responsible AI & security](/aws/ai-practitioner/notes/04-responsible-ai-and-security/) |
| 5 | Security, Compliance, and Governance for AI Solutions | 14% | [Responsible AI & security](/aws/ai-practitioner/notes/04-responsible-ai-and-security/) |

Generative AI and its applications are **52%** of the paper. Prompting, inference parameters, model customisation, RAG and agents are where the marks concentrate — not in classical ML theory.

## Exam mechanics

- **65 items, 90 minutes.** **50 scored**, 15 unscored pretest items mixed in unmarked.
- **Scaled 100–1,000, pass at 700**, compensatory across domains.
- **Four item types**, and this is the only exam on this site with the last two:
  - Multiple choice — one of four.
  - Multiple response — two or more of five-plus, all correct answers required.
  - **Ordering** — arrange 3–5 responses in the correct sequence; the whole sequence must be right.
  - **Matching** — pair a list of responses against 3–7 prompts; every pair must be right.
- Ordering and matching are all-or-nothing, so practise lifecycle sequences (data → training → evaluation → deployment → monitoring) until the order is automatic.
- Unanswered items score as incorrect; there is no guessing penalty.

## The four note pages

| # | Page | Study points | Covers |
| --- | --- | --- | --- |
| 1 | [AI & ML foundations](/aws/ai-practitioner/notes/01-ai-and-ml-foundations/) | 20 | Learning paradigms, deterministic vs probabilistic models, train/validation/test splits, bias and variance, evaluation metrics, embeddings, transfer learning |
| 2 | [Generative AI](/aws/ai-practitioner/notes/02-generative-ai/) | 21 | Foundation models and LLMs, inference parameters, prompt engineering, model customisation, RAG, agents |
| 3 | [AWS AI services](/aws/ai-practitioner/notes/03-aws-ai-services/) | 14 | The pre-trained Amazon AI services, SageMaker AI, Bedrock, SageMaker Clarify and explainability |
| 4 | [Responsible AI & security](/aws/ai-practitioner/notes/04-responsible-ai-and-security/) | 15 | Dimensions of responsible AI, interpretability vs explainability, AI security threats, audit, compliance and governance |

**70 study points in total**, split evenly between the concept pages and the service pages — which is how the exam splits too.

## The service map

The fastest way to lose marks here is to know the concept and not the service. Know which box each item belongs in:

| Need | Service |
| --- | --- |
| Foundation models from multiple providers, managed, no infrastructure | **Amazon Bedrock** |
| Build, train, tune and deploy your own models | **Amazon SageMaker AI** |
| Bias detection and model explainability | **SageMaker Clarify** |
| Grounding a model in your own documents, managed | **Bedrock Knowledge Bases** (managed RAG) |
| Content filters, denied topics, PII filtering, grounding checks | **Bedrock Guardrails** |
| Text extraction from documents and forms | **Amazon Textract** |
| Image and video analysis | **Amazon Rekognition** |
| Speech to text / text to speech | **Amazon Transcribe** / **Amazon Polly** |
| Translation, sentiment and entity extraction | **Amazon Translate** / **Amazon Comprehend** |
| Conversational bots | **Amazon Lex** |
| Personalised recommendations | **Amazon Personalize** |

## Choosing between prompting, RAG and fine-tuning

This trade-off is asked repeatedly, in several disguises:

| Approach | Use when | Cost and effort |
| --- | --- | --- |
| **Prompt engineering** | The model already knows enough; you need format, tone or reasoning structure | Lowest — no training, instant iteration |
| **RAG** | The answer depends on your own or on fresh data | Moderate — retrieval infrastructure, no training; content stays current |
| **Fine-tuning** | You need consistent style, domain vocabulary or behaviour the prompt cannot reach | Highest — labelled data, training cost, and it goes stale as the domain moves |

The exam's default preference runs left to right: try the cheapest approach that meets the requirement, and reach for fine-tuning only when the stem rules the others out.

## AIF-C01 versus AIB-C01

Both are AI credentials and they are routinely confused. The dividing line is whether AWS services are assessed.

| | **AI Practitioner (AIF-C01)** | **[AI Business Strategist (AIB-C01)](/aws/aib-c01/)** |
| --- | --- | --- |
| Category | Foundational | Business |
| Assesses AWS services? | **Yes** | **No** — strategic familiarity only |
| Validates | AI/ML and GenAI concepts applied on AWS | Investment judgment, business cases, governance, scaling |
| Items · time · pass | 65 · 90 min · 700 | 85 · 170 min · 700 (beta) |
| Price | $100 | $50 during beta |

If your question is "which service", this exam. If it is "should we fund this and how do we govern it", [AIB-C01](/aws/aib-c01/).

## A two-week plan

<CardGrid>
  <Card title="Days 1–4 · Concepts" icon="lucide:brain-circuit">
    [AI & ML foundations](/aws/ai-practitioner/notes/01-ai-and-ml-foundations/). Be able to place any scenario into supervised, unsupervised, reinforcement or generative without hesitating.
  </Card>
  <Card title="Days 5–8 · Generative AI" icon="lucide:sparkles">
    [Generative AI](/aws/ai-practitioner/notes/02-generative-ai/) twice — it feeds the two heaviest domains. Learn the inference parameters by what they do to output, not by name.
  </Card>
  <Card title="Days 9–11 · Services" icon="lucide:boxes">
    [AWS AI services](/aws/ai-practitioner/notes/03-aws-ai-services/) plus the service map above. Drill it as flashcards, both directions.
  </Card>
  <Card title="Days 12–14 · Responsibility" icon="lucide:scale">
    [Responsible AI & security](/aws/ai-practitioner/notes/04-responsible-ai-and-security/), then the [AIB-C01 mock exam](/aws/aib-c01/practice-exam/) for the governance items, which overlap well.
  </Card>
</CardGrid>

## Where to go next

- Compare all five tracks on the [certification tracks page](/aws/certifications/).
- [Cloud Practitioner](/aws/cloud-practitioner/) if the AWS-service half of this exam was the hard part.
- The [AWS appendix](/appendix/aws/frameworks/) carries the responsible-AI frameworks, the governance toolkit and the service landscape.