11.3 90-Minute Exam Timing, Flagging, and Guessing Workflow

Timing is part of the exam skill

AIF-C01 is 90 minutes long and includes 65 total questions. That gives about 83 seconds per question on average, but the real exam will not feel evenly spaced. Some items may be quick vocabulary or service-boundary checks. Others may require reading a business scenario, noticing a constraint, and selecting the AWS-aligned response. Your timing workflow should protect both speed and judgment.

The exam guide says the exam includes 50 scored questions and 15 unscored questions, and the unscored questions are not identified. That means you cannot decide that a strange item is safe to ignore. Treat every item as worth your best disciplined effort. At the same time, do not spend five minutes trying to perfect one uncertain answer. The scoring model is compensatory, and unanswered questions count as incorrect.

The three-pass workflow

Pass one is for forward motion. Read the question stem, identify the task, notice words such as best, most appropriate, least, first, or multiple response, and answer if you can. If the answer is clear, do not mark it out of habit. Marking too many items turns final review into a second exam and wastes the advantage of your first-pass judgment.

Pass two is for flagged items with a reason. A useful flag means you can name the uncertainty: two services seemed plausible, the scenario had a security constraint, the item asked for multiple responses, or you might have misread whether AI was appropriate. A weak flag means you felt nervous. Nervous flags are usually not actionable and can consume the final minutes.

Pass three is a safety check. Confirm every question has an answer. Revisit marked items only if you have a clear way to improve them. Do not change answers just because you are tired. Change an answer when you found a concrete misread, such as choosing a custom ML path when the scenario asked for a managed foundation model application, or choosing generation when the scenario required deterministic policy enforcement.

Guessing as structured elimination

There is no penalty for guessing, so guessing is part of the workflow. The disciplined version starts by eliminating options that violate the scenario. If the user needs current private documents, a model-only answer without retrieval may be weak. If the scenario involves sensitive data, an option that ignores IAM, privacy, logging, or human review may be weak. If a question asks for practitioner-level judgment, an option focused on coding a custom algorithm may be outside the target candidate boundary.

Use domain cues to eliminate. Domain 1 questions often test use-case fit, data type, learning type, lifecycle stage, or metric alignment. Domain 2 questions often test GenAI vocabulary, prompting, context, hallucination, embeddings, or customization concepts. Domain 3 questions often test service selection among Bedrock, Amazon Q, SageMaker AI, Knowledge Bases, Agents, Guardrails, and managed AI services. Domains 4 and 5 often test whether a proposed answer handles responsible AI, privacy, access, logging, encryption, monitoring, or governance.

Item-type handling

Multiple choice usually asks for one best answer. Multiple response requires extra care because more than one option can be correct. Ordering and matching require you to reason about sequence or association, not just select a term. If the interface shows instructions for the item type, read them carefully. A candidate who knows the concept can still lose points by answering a multiple-response item as if only one option mattered.

For scenario items, start with the business requirement before looking for service names. A support team may ask for faster ticket routing, but the best answer depends on data quality, privacy, review needs, and whether the task is classification, summarization, search, or action orchestration. A legal team may ask for AI summaries, but the governance risk may require citations, human review, and restricted data access. Read the constraint before choosing the shiny service.

Timing rehearsal checklist

• Practice at least one 65-item or equivalent timed session with a 90-minute cap.
• Track how often you exceed two minutes on a single item.
• Practice answering and flagging rather than leaving blanks.
• Keep a short list of elimination cues for each domain.
• Review only flagged items with named uncertainty.
• Stop changing answers without a concrete reason.
• End every timed practice by checking that no item is unanswered.

The right timing workflow is not reckless speed. It is controlled progress. You want enough time to read scenarios carefully, but not so much time on early items that the last 15 questions become rushed guesses. AIF-C01 rewards broad, practical judgment, and pacing is how you give that judgment a chance to show up across the whole exam.