AI Fundamentals and Literacy
Ask whether a written rule settles it before choosing any kind of AI
Can you write the rule down completely? AWS's Responsible AI Lens asks that before any commitment to artificial intelligence (AI): a problem you can solve 'by writing down a clear and compact set of rules' may not need machine learning, and AI is the choice only if it provides 'clear, substantial benefits' over competing solutions. Where AI is justified, the next question is whether the machine learning (ML) model must create new content: ML is the subset of AI focused on learning from data without explicit programming, and ML models perform tasks like classifying documents, labeling images, or predicting the maintenance schedule of factory equipment, while generative AI (GenAI) is a type of AI that can create new content and ideas and works by using machine learning models, so both answers use ML models. Two opposite mistakes follow: treating content generation as the test for AI, though a classifier that generates nothing is still AI because all machine learning is AI, and reaching for machine learning when a compact set of rules may make it unnecessary.
Inside generative AI, ask which gap shows
Prompt engineering and Retrieval Augmented Generation (RAG) change the inputs sent to the model, while fine-tuning changes the model parameters (the values the model learned during training), so the gap that shows in the response decides the technique. A wrong task or format calls for prompt engineering: a prompt is a specific set of inputs, provided by the user, that guide a large language model (LLM) to generate an appropriate response, and adding a few example inputs with their desired outputs is few-shot prompting, also known as in-context learning. Missing private or changing facts call for RAG, which augments an LLM with external data, such as a company's internal documents, and AWS recommends starting from a RAG-based approach for a question-answering solution over your custom documents. A persistent behavior gap (a format or tone the model should always follow that prompts did not fix) calls for fine-tuning, in which the model parameters are adjusted from labeled examples (supervised fine-tuning), but it can take a few hours to days, so it may not suit frequently changing documents.
Domain 1 task objectives
Task 1.1 of the AWS AIB-C01 exam guide asks you to distinguish AI, ML, and generative AI, explain fundamental AI concepts in business contexts such as algorithms, models, training, and inference (a trained model generating educated predictions or outputs from new data), distinguish structured from unstructured data, describe why data quality matters for AI outcomes, and know global frameworks for a unified AI vocabulary such as ISO/IEC 23053 and ISO/IEC 42001. Task 1.2 asks when to use rule-based automation and when to use AI, how AI agents (software that can use the data it collects to perform self-directed tasks that meet predetermined goals) differ from other AI solutions, why AI solutions need ongoing monitoring and updates because models degrade over time due to changes in the real world, such as data drift and concept drift, and how classifying AI tools as approved, blocked, or under evaluation mitigates shadow AI, the unsanctioned use that AWS ties to unmanaged and uncontrolled data risks. Task 1.3 asks you to apply basic prompt engineering, identify when token limits or context window constraints affect performance, and recognize how model adaptation techniques such as RAG and fine-tuning improve responses for specific business needs, although only fine-tuning changes the model parameters; RAG changes the inputs sent to the model. A token is a sequence of characters that a model can interpret or predict as a single unit of meaning, and the context window is the maximum amount of text, measured in tokens, that the model can consider at one time while generating a response.
Domain 1 tasks and the decision-tree question each one supplies
| Task | Its decision-tree question | Drill into |
|---|---|---|
| Task 1.1: Describe core AI concepts and define terminology | Tree question 2: Must the ML model create new content (generative AI), or classify and predict? | Describe core AI concepts and terminology |
| Task 1.2: Identify and select appropriate AI solution types | Tree question 1: Can you write the rule down completely (rule-based automation), or is AI needed? | Identify and select appropriate AI solutions |
| Task 1.3: Apply GenAI concepts and techniques | Tree question 3: Which gap shows: wrong task or format, missing private or changing facts, or a persistent behavior gap? | Apply generative AI concepts and techniques |