AgileWoW - Agile Ways of Working

1. A Product Owner wants to use an AI model to group customers into different segments based on their purchasing behavior, without any predefined categories. Which type of machine learning is most suitable?

Supervised Learning Reinforcement Learning Unsupervised Learning Semi-Supervised Learning

2. What is the primary goal of Reinforcement Learning?

To classify data into predefined categories. To train an agent to make a sequence of decisions by rewarding it for good actions and penalizing it for bad ones. To generate new, original content based on a prompt. To find correlations in large, unstructured datasets.

3. A "Generative AI" model is one that primarily:

Analyzes existing data to make a prediction or classification. Creates new content (text, images, code, etc.) that did not exist before. Automates repetitive, rule-based tasks. Secures a system against cyber threats.

4. What does it mean to "fine-tune" an LLM?

To build a new LLM from the very beginning. To perform a small, final training run on a pre-trained model using a specialized, smaller dataset to adapt it for a specific task. To adjust the user interface of the AI application. To reduce the electrical consumption of the AI model.

5. How can a Product Owner ethically use AI during market research?

By generating fake competitor data to make their product look superior. By analyzing public sentiment from social media, reviews, and forums to identify customer pain points and trends. By creating entirely fictional user personas that are not based on any real data. By scraping private user data without consent.

6. When defining a Product Goal for an AI-powered product, it's most important that the goal focuses on:

The specific algorithm being used (e.g., "Implement a BERT-based model"). The measurable user outcome or business impact (e.g., "Reduce user research analysis time by 50% within two quarters"). The number of prompts the system can handle. The programming language used to build the feature.

7. True or False: AI can be used to convert high-level notes from a stakeholder meeting into well-formed draft user stories.

True False

8. Which of the following is the least appropriate task to delegate entirely to an AI?

Generating a first draft of acceptance criteria for a given user story. Correcting spelling and grammar in backlog items. Translating user stories into another language for distributed teams. Setting the Product Goal and defining the long-term product vision.

9. True or False: If an AI model is trained on biased data, its output will also be biased, regardless of how perfect the algorithm is.

True False

10. What is an AI "hallucination"?

A feature that makes the AI more creative. When an AI model generates information that is false, nonsensical, or untethered to its training data, but presents it as fact. A security feature that protects the model. The process of visualizing the AI's neural network.

11. What does "explainability" mean in the context of AI?

The ability of the AI to explain its jokes. The ability to understand and articulate why an AI model made a specific decision or prediction. The user manual for the AI software. The process of training an AI model.

12. A Product Owner is working on an AI feature that recommends products to users. How could they check for bias in the model?

By asking the model if it is biased. By analyzing the recommendations across different user segments (e.g., by demographics) to see if they are fair and equitable. By assuming that since the math is complex, it must be objective. By only checking if the recommendations are increasing overall sales.

13. What is the concept of "Human-in-the-Loop" (HITL) in AI systems?

A system where humans perform all the tasks and AI only watches. A design principle where the AI must get human approval for every single computation it performs. A system that requires human interaction and intervention to handle exceptions, review outputs, or make final decisions. The idea that AI will eventually become human.

14. Why is it important for a Product Owner to understand the cost structure of using a third-party LLM API (e.g., cost per token)?

It is not important; the finance department handles all costs. To ensure the AI feature is financially viable and doesn't lead to unexpectedly high operational expenses. To be able to negotiate a better salary. To manually approve every single API call.

15. What is "prompt engineering"?

A hardware engineering role focused on AI chips. The art and science of designing effective inputs (prompts) to guide an AI model toward a desired output. A marketing technique for promoting AI products. A method for building LLMs from scratch.

16. True or False: Multimodal AI refers to a model that can only process and understand text.

True False

17. A "Retrieval-Augmented Generation" (RAG) system improves LLM performance by:

Making the model's font size bigger. Allowing the model to access and retrieve information from an external, up-to-date knowledge base before generating an answer. Deleting old data from the model's training set. Asking the user for the answer.

18. What is a "foundation model"?

A small, simple AI model. A large-scale model (like GPT-4 or Gemini) that is trained on a vast quantity of broad data and can be adapted for a wide range of tasks. A model that is used exclusively for building foundations for houses. The very first AI model ever created.

19. The "black box" problem in AI refers to:

The physical casing of the computer running the model. A situation where it is difficult or impossible to understand how an AI model arrived at a particular output. A type of AI that can only produce black and white images. A secure AI model that cannot be hacked.

20. When considering a new AI feature, a Product Owner's first question should be:

"What is the coolest algorithm we can use?" "What user problem are we solving, and is AI the best way to solve it?" "How can we collect as much user data as possible?" "Can we get this done in one Sprint?"

AI for Product Owners: Knowledge Check

1. A Product Owner wants to use an AI model to group customers into different segments based on their purchasing behavior, without any predefined categories. Which type of machine learning is most suitable?

2. What is the primary goal of Reinforcement Learning?

3. A "Generative AI" model is one that primarily:

4. What does it mean to "fine-tune" an LLM?

5. How can a Product Owner ethically use AI during market research?

6. When defining a Product Goal for an AI-powered product, it's most important that the goal focuses on:

7. True or False: AI can be used to convert high-level notes from a stakeholder meeting into well-formed draft user stories.

8. Which of the following is the least appropriate task to delegate entirely to an AI?

9. True or False: If an AI model is trained on biased data, its output will also be biased, regardless of how perfect the algorithm is.

10. What is an AI "hallucination"?

11. What does "explainability" mean in the context of AI?

12. A Product Owner is working on an AI feature that recommends products to users. How could they check for bias in the model?

13. What is the concept of "Human-in-the-Loop" (HITL) in AI systems?

14. Why is it important for a Product Owner to understand the cost structure of using a third-party LLM API (e.g., cost per token)?

15. What is "prompt engineering"?

16. True or False: Multimodal AI refers to a model that can only process and understand text.

17. A "Retrieval-Augmented Generation" (RAG) system improves LLM performance by:

18. What is a "foundation model"?

19. The "black box" problem in AI refers to:

20. When considering a new AI feature, a Product Owner's first question should be:

Download Sample Questions

Success!