ISTQB CT-GenAI
Introducing Generative AI in the Test Organization
Easy
Shadow AI and strategy
- What is meant by shadow AI in a test organization?
- A tester copies defect reports from a travel portal into a private AI chat. Which risk is most prominent?
- What consequence can it have if a test team uses unapproved AI tools?
- Why are unclear license terms of an AI tool a problem?
- What helps a test organization most to avoid shadow AI?
- What is a sensible starting point for a GenAI strategy in testing?
- Why is the quality of input data so important for a GenAI strategy?
Model selection and adoption phases
- Which criterion is central when selecting a language model for test tasks?
- Why are general benchmarks often not enough when selecting a model for test tasks?
- Which recurring costs need to be considered when choosing a language model?
- Why do user community and support matter when choosing a model?
- Which three phases does the adoption of GenAI in a test organization go through?
- What typically happens in the first phase of adoption?
Skills and new roles
- Which skill do testers particularly need for working with GenAI?
- Why does domain and test knowledge remain important when using GenAI?
- What does it mean to sanitize data before testware is passed to an LLM?
- How can AI skills best be built up in a test team?
- What is a prompt pattern?
- How does the role of testers change through GenAI?
- What are internal communities of practice for when using GenAI in testing?
Medium
Shadow AI and strategy
- A developer uploads a customer's source code to an AI service whose terms leave open who owns the results. Which risk becomes particularly clear here?
- Which list correctly summarizes the risks of shadow AI?
- What belongs in process guidelines for using GenAI in testing?
- A bank requires every test case to note whether it was created with GenAI. What is this mainly for?
- An insurer has generated test scripts reviewed at a quality gate before they are adopted. Which part of the strategy does this belong to?
- Why do metrics belong in a GenAI strategy for testing?
- A team chooses a model with excellent benchmark scores that cannot be connected to the existing test management tool. What was neglected?
Model selection and adoption phases
- A medical device manufacturer checks whether a model can be retrained with its own domain data. Which selection criterion is it looking at?
- What do the recurring costs of a test task with GenAI mainly depend on?
- A team has 50 defect reports summarized every day: 2,000 input and 500 output tokens each. Price: €1 per million input and €4 per million output tokens. What does this cost per day?
- A team compares a commercial model priced per token with an open source model in its own data center. What is part of a fair comparison?
- A team knows the basics and is now choosing where GenAI brings the greatest benefit in testing. Which phase is it in?
- At an online retailer, GenAI is already used routinely for test data but only being tried out for load tests. How should this be assessed?
Skills and new roles
- Why should testers know a model's context window?
- How can testers reduce energy consumption when using GenAI?
- A tester wants an LLM to analyze a defect report containing an electricity provider's customer data. How should the tester proceed?
- A test team has mastered simple prompts. What is the next sensible learning step?
- What promotes continuous improvement in using GenAI within communities of practice?
- Which task is added for test managers through GenAI?
- When GenAI is introduced, testers worry about being replaced. How should the organization respond?
Hard
Shadow AI and strategy
- Which statement about shadow AI is NOT correct?
- A logistics company wants to "somehow use AI in testing" and has already ordered a model. What is most clearly missing?
- A team uses a requirements collection that has not been maintained for two years for test case suggestions. Which aspect of the strategy is mainly affected?
- Why should GenAI training for testing teach ethical as well as technical skills?
- Which combination describes a GenAI strategy for testing most completely?
- A survey shows that many testers secretly use private AI chats because there are no approved tools. Which measure fits best?
- Which rule does NOT belong in sensible guidelines for using GenAI in testing?
Model selection and adoption phases
- A team generates 400 draft test cases per month: 3,000 input and 1,000 output tokens each. Price: €2 per million input and €8 per million output tokens. What are the monthly costs?
- A team attaches the entire 80-page functional specification to every prompt although only two pages are needed. What follows for the costs?
- A public authority may not pass test data to external vendors and has a tight budget. Which choice fits best?
- How can you tell that an organization has reached the third phase of adoption?
- Does a test organization have to choose exactly one language model?
- Which activity typically belongs to the second phase of adoption?
Skills and new roles
- What lies ahead for test managers in teams where people and GenAI agents work together?
- A test manager has the team only rubber-stamp AI results; classic test techniques are no longer practiced. What is the problem with this?
- Which task does NOT belong to the new tasks of test managers when using GenAI?
- An AI-savvy developer without testing knowledge assesses generated boundary value tests for a loan calculator. Where is the greatest risk?
- Which order corresponds to the gradual development of AI skills in the test team?
- A team has a large model reword 300 regression tests on every commit, although hardly any tests change. What speaks against this?
- A tester maintains a collection of proven prompts for boundary value analyses. Which new task is the tester taking on?