AWS quietly revised the AIF-C01 exam guide, adding agentic AI and token pricing objectives

AWS published version 1.1 of the AWS Certified AI Practitioner (AIF-C01) exam guide on April 30, 2026, five weeks after version 1.0, and the guide itself says updates show up on the live exam roughly a month after publication. According to a line-by-line breakdown of the revision page published on dev.to, the update adds seven entirely new objectives, reshuffles which AWS services are in scope, and quietly swaps examples inside objectives that kept their numbering. Anyone studying from pre-May 2026 courses or question banks is now working from a partly outdated syllabus.

What is new in version 1.1

The exam format is unchanged: 65 questions of which 50 are scored, a pass mark of 700 out of 1,000, and five domains weighted at 20 percent for Fundamentals of AI and ML, 24 percent for Fundamentals of GenAI, 28 percent for Applications of Foundation Models, and 14 percent each for Guidelines for Responsible AI and Security, Compliance, and Governance for AI Solutions.

The objective list is where the movement is. According to the dev.to analysis, seven objectives in version 1.1 are new lines rather than rewordings:

• 1.2.6, when to use traditional ML models versus foundation models, for example because of regulatory concerns, explainability or operational constraints
• 2.1.4, token-based pricing and how it affects cost and performance for inference
• 2.1.5, the role of context engineering in foundation model applications
• 2.1.6, foundational agentic AI concepts, including multi-agent patterns, Model Context Protocol, agent communication, memory management, tool usage and workflow orchestration
• 3.2.5, prompt versioning and management with Amazon Bedrock Prompt Management
• 3.4.5, business alignment metrics such as task completion rate, user satisfaction and cost per interaction
• 5.1.5, hallucination detection and grounding, including RAG grounding, output validation and confidence scoring

Agentic AI is the thread running through the whole revision, the dev.to author notes. It now appears in the basic terms candidates must define (1.1.1), in the comparison between AI, ML, GenAI and deep learning (1.1.2), in real-world applications (1.2.4), in a dedicated new objective (2.1.6), and in the security domain through AgentCore Identity and Policy (5.1.1).

Services moved in and out

Seven services were added to the in-scope list: Amazon Aurora, Amazon Bedrock AgentCore, Kiro, Strands Agents, Amazon Q, Amazon SageMaker JumpStart and AWS Transform. Amazon MemoryDB left the list.

The post also flags an inconsistency worth knowing about before replanning your study schedule. The revision table lists Amazon Q as an addition, and it appears in the revision entry for objective 1.3.4, but neither the in-scope services page nor objective 1.3.4 as currently published mentions it. The author's advice is not to spend much time on Amazon Q until the guide settles.

Existing objectives with new examples

Older question banks go stale less through new objectives than through example swaps inside objectives that kept their numbers. The dev.to breakdown highlights several:

• Inference types (1.1.3) previously covered batch and real-time; asynchronous and serverless are now added.
• Model metrics (1.3.6) drop Area Under the Curve and now list accuracy, precision, recall and F1 score.
• Pipeline services (1.3.4) no longer point at SageMaker Data Wrangler, Feature Store and Model Monitor; the published examples are now Amazon Bedrock, Kiro and SageMaker AI, among others.
• Services for building GenAI applications (2.3.1) no longer cite Bedrock PartyRock or Bedrock Data Automation, replaced by Bedrock, SageMaker AI, JumpStart, Strands Agents and AgentCore.
• The agents objective (3.1.6) used to name Amazon Bedrock Agents; it now asks about the role and business applications of AI agents generally, without naming a product.
• Customization cost (3.1.5) adds model distillation next to pre-training, fine-tuning, in-context learning and RAG.
• Evaluation metrics (3.4.2) add LLM-as-a-judge alongside ROUGE, BLEU and BERTScore.
• Explainability tools (4.2.2) add SageMaker Clarify and Amazon Bedrock Model Evaluations.
• Security considerations (5.1.4) add data leakage prevention, output filtering and validation, audit trails for AI interactions, and toxicity.

How to study with older material

The dev.to author's guidance is to keep a pre-May 2026 course, since most of the exam is still the same material, and then spend extra time on three areas older courses barely touch: what an agent is and how it uses tools, memory and MCP; how tokens drive cost, and when RAG, fine-tuning or distillation is the cheaper route; and the Bedrock pieces that make answers safer, namely Guardrails, contextual grounding and Prompt Management.

Why it matters

Certification guides rarely get a second version five weeks after the first, and this one landed without a new exam code or a headline announcement. Candidates who bought a course or practice set before May 2026 can now face scored questions on agentic AI, Model Context Protocol and token economics that their material never mentions. The revision is also a signal of what AWS itself considers current: agentic patterns, context engineering and grounding have moved from conference talk into an entry-level certification, which is a reasonable proxy for what the vendor expects practitioners, and the teams hiring them, to know.