Scrum.org PSM-AI: Professional Scrum Master - AI Essentials

Last checked November 10, 2026 against the official Scrum.org exam page.

1. Question 1AI Theory and Primer

   A Scrum Master's team uses a general-purpose language model to draft acceptance criteria. The model confidently cites an industry standard clause that does not exist. Which term best describes this output?

   1. ARetrieval, because the model searched a verified standards database
   2. BQuantization, because the model compressed the standard into a shorter form
   3. COverfitting, because the model memorized the exact clause from training
   4. DHallucination, meaning fluent output that is not grounded in verified facts
   Show answer and explanationHide answer

   Correct answer: D

   Hallucination describes plausible-sounding output that is not grounded in verified sources, which fits a citation to a clause that does not exist. Retrieval would need a real search step, which the scenario does not describe, overfitting means reproducing training text, and quantization is a compression technique unrelated to invented citations.

2. Question 2AI Security and Ethics

   A developer pastes unreleased source code from a client project into a public AI chat tool to get a faster bug fix. What is the most important concern?

   1. AThe code will run slower because the tool compresses its input before analyzing it
   2. BThe tool will automatically publish the full code to a public repository for anyone to find and browse
   3. CConfidential client data may leave the organization and be handled under unapproved terms
   4. DThe bug fix will always be wrong because the tool cannot read source code properly
   Show answer and explanationHide answer

   Correct answer: C

   Pasting confidential code into an external tool can move client data outside the organization's control, so data-handling rules and approved tools should decide what may be shared. The other options describe effects the scenario does not support, such as automatic publication or guaranteed wrong fixes.

3. Question 3AI for Scrum Masters

   During a Sprint Retrospective, a Scrum Master asks an AI tool to summarize the team's sticky notes. What should happen next?

   1. AThe AI summary becomes the official list of improvements the team must implement
   2. BThe Scrum Master deletes the original notes so that only the summary remains
   3. CThe team checks the summary against its notes and decides what to act on
   4. DThe Scrum Master chooses the improvements alone and reports them to the team later
   Show answer and explanationHide answer

   Correct answer: C

   The Scrum Team owns its retrospective, so AI output is a draft that the team verifies and then decides on. The other options let the AI decide, destroy evidence the team should inspect, or have one person decide for the whole team.

4. Question 4Effective AI Prompting

   Which prompting technique includes a few examples of the desired input-output pattern before the real request?

   1. AFew-shot prompting, with examples of the desired pattern in the prompt
   2. BZero-shot prompting, which gives the model no examples at all
   3. CChain-of-thought prompting, which asks for step-by-step reasoning
   4. DSystem prompting, which sets a persistent role and rules for the whole session
   Show answer and explanationHide answer

   Correct answer: A

   Few-shot prompting shows the model a small number of worked examples so it can follow the pattern for the actual request. Zero-shot uses no examples, chain-of-thought asks for step-by-step reasoning, and system prompting sets a persistent role rather than demonstrating a pattern.

5. Question 5AI Theory and Primer

   What does a large language model do at its core when it generates text?

   1. AIt searches an indexed copy of the internet for each prompt and copies the result
   2. BIt runs a fixed set of hand-written rules that its developers wrote for each topic
   3. CIt stores every past conversation and returns the closest matching stored reply
   4. DIt predicts the next token from learned statistical patterns
   Show answer and explanationHide answer

   Correct answer: D

   A language model generates output by predicting the most likely next token given the preceding context, a process learned from training data. The other options describe live internet search, hand-written rule systems, or conversation lookup, none of which is how a base language model generates text.