Making the AI's code open source

Being clear with stakeholders and the team about where and how AI is being used in the product or process

Using a transparent computer monitor

Explaining exactly how the LLM's neural network works

As a mandatory commitment for the upcoming Sprint

As one data point to inform the team's discussion about their capacity

To replace the team's collective intuition

To assign individual tasks to developers

AI takes over the management of the team

AI provides insights that help the team make better-informed decisions for themselves

AI makes self-management unnecessary

AI requires a central manager to oversee all prompts

Persona Prompting

Zero-shot Prompting

Implicit Bias

Iterative Debugging

It can only classify existing data

It can create new content like text, images, or code

It is always 100% correct

It is a form of physical robot

An AI-managed list of requirements

An ordered, living document of everything that is known to be needed in the product

A static document created by Generative AI at the start of the project

A list of prompts the team needs to run

It guarantees the final answer will be 100% accurate

It asks the AI to show its reasoning steps, making it easier to verify the logic

It allows the AI to bypass its training data

It prevents the AI from using technical jargon

Summarizing long text documents

Generating code from a natural language description

Feeling genuine empathy for a Scrum Team's challenges

Translating technical requirements into user stories

By summarizing the feedback received from stakeholders during the session

By acting as the primary presenter so the Developers can keep coding

By telling stakeholders they are wrong if they criticize the increment

By predicting what the stakeholders will say before they say it

Accept the first version to save time

Use iterative prompting to refine the tone, target audience, and key value propositions

Switch to a different AI model immediately

Rely solely on the AI's internal knowledge of market trends

A Logic Gap

A Hallucination

A Training Error

Personality Prompting

The AI model's training data

The prompt engineer

The Scrum Team

The organization's legal department

AI tools cannot perform mathematical calculations

AI lacks the specific context of the team's skills, technical debt, and environment

Estimation is a strictly forbidden activity for AI in Scrum

AI will always overestimate to be safe

Writing boilerplate code and unit test stubs

Deciding the technical architecture without human review

Choosing which team members should work on which tasks

Determining the Definition of Done for the organization

The AI might omit subtle but critical blockers mentioned by the team

The AI will make the Daily Scrum longer

Using AI for summarization is a violation of Scrum

The AI will try to facilitate the meeting itself

Asking the same question to ten different AI tools

Refining the AI's response by providing feedback and follow-up prompts

Waiting for the AI to iterate on its own

Prompting the AI only at the end of every iteration (Sprint)

The amount of text the model can remember or consider at one time during a conversation

The window on your computer where the AI runs

The time frame in which the AI was trained

The physical location where the AI is hosted

Asking the AI a question without providing any prior examples of the desired output

Using an AI tool that has had zero training data

A prompt that results in a zero-word response

Asking the AI to ignore all previous instructions

Accept the proposal as it supports the Scrum Value of Focus

Reject the proposal as it violates the Scrum Values of Respect and Trust

Trial it for one Sprint to see if velocity increases

Ask the Product Owner if the data is valuable

The AI's ability to speak in a human voice

Ensuring the team feels safe and that their data is being handled according to privacy standards

The speed at which the AI can generate action items

The AI's knowledge of the Scrum Guide

Are black boxes with no information on how they handle data or bias

Require a subscription fee

Don't have a voice feature

Were released more than six months ago

A technique used to manipulate an LLM into ignoring its original instructions or safety guards

A way to make the AI respond faster

Injecting code into the LLM's training set

A method of charging the AI's battery

Add Feature X to the top of the Product Backlog immediately

Verify the AI's findings against the raw data and other feedback channels before making a decision

Tell the team the AI has commanded the development of Feature X

Ask the AI to write the code for Feature X

It uses fewer words

It provides a pattern for the AI to follow, reducing the chance of irrelevant output

It is cheaper to run

It doesn't require any training data

By providing a checklist of industry-standard quality practices for the team to consider

By enforcing the Definition of Done automatically via the source code

By deciding when a Sprint is finished

By replacing the need for unit testing

Shrinking the physical size of hard drives

Providing only the essential data needed for the AI to complete the task, reducing privacy risks

Using the smallest possible LLM for every task

Deleting all data after the AI has processed it

Encourage it if it leads to better Product Backlog items

Stop the action and explain the data privacy risks and organizational policy

Ask the AI if it is safe to paste the data first

Only allow it if the AI promises to keep it secret

A parameter that controls the randomness or creativity of the output

The physical heat of the server rack

How fast the AI responds

The difficulty level of the prompt

The Product Owner and the AI

The Scrum Team

The Scrum Master only

The Developers and the Prompt Engineer

An AI that acts as a secret agent to monitor competitors

An AI that can only perform one specific task, like spell-checking

An AI system that can autonomously break down a goal, use tools, and iterate to achieve a result

An AI that has been hired as a permanent employee

Focus more on complex problem-solving and delivering high-value features

Stop worrying about quality altogether

Reduce the number of Developers on the team

Extend the length of their Sprints

Maximize the number of lines of code written

Increase the team's ability to deliver value and reach the Product Goal

Replace human developers with more efficient scripts

Prove that AI is better than human judgment

The LLM provider

The Developers

The Scrum Master

The AI Tool Administrator

By ensuring diverse perspectives are included in the review of AI outputs

By only using AI tools that claim to be bias-free

By assuming the AI is always objective

By never using AI for tasks involving human data

Make the team more professional

Mask the team's true feelings and reduce authentic collaboration

Improve the team's velocity

Be a great use of Persona Prompting

They cannot process information faster than humans

They are unable to follow complex formatting instructions

They can hallucinate or provide confidently stated but incorrect information

They require physical hardware to function

Update the Definition of Done to include specific human-led security reviews for AI code

Stop using AI for code generation immediately

Trust that the AI will learn from its mistakes in the next Sprint

Blame the AI provider for the vulnerabilities

Ask the AI to suggest three potential goals based on the selected Product Backlog items, then let the team refine and choose one

Let the AI choose the goal and the team commit to it without discussion

Use the AI to check if the goal is SMART but never to draft it

Avoid using AI for goals as it reduces team commitment

The transparency of the tool's data usage and potential for algorithmic bias

The tool's ability to replace human developers

Whether the tool has a friendly persona

If the tool can work 24/7 without a break

The AI Assistant

The Scrum Master

The Scrum Team

The Stakeholders

Choosing the most colorful AI visualization

Reviewing the roadmap with stakeholders to ensure it aligns with business strategy

Ensuring the AI uses the latest 2024 market data

Letting the AI update the roadmap in real-time without intervention

A deterministic system that always provides the same output for a given input

A database of facts that can be queried using natural language

A probabilistic model that predicts the most likely next sequence of tokens

A sentient intelligence capable of independent reasoning and moral judgment

Stop using the tool and consult with their organization's legal and security departments

Only use the tool for non-sensitive code

Ask the AI if it is training on their code

Nothing, this is a standard feature of all AI

They are probably going to be correct

They calculate the statistical likelihood of the next word/token based on patterns

They are programmed with specific if-then logic

They only work during certain times of the day

Increase velocity through healthy competition

Destroy trust and psychological safety within the team

Be the most accurate way to measure value

Improve the team's adherence to Scrum

Augmented Intelligence

Artificial General Intelligence

Scrum Distraction

Hallucination

A concrete stepping stone toward the Product Goal

A collection of AI-generated ideas

The total number of prompts used during the Sprint

The percentage of code written by AI

Focus

Commitment

Openness

Respect

Courage

Openness

Respect

Commitment

By deciding which items have the highest ROI without human input

By suggesting Acceptance Criteria and identifying potential edge cases for User Stories

By automatically deleting low-priority items from the Product Backlog

By attending the refinement session on behalf of the Product Owner

By letting the AI decide who speaks next in every meeting

By asking the AI for diverse retrospective formats based on current team dynamics

By using AI to create a script that everyone must follow during the Sprint Review

By replacing the need for a human Scrum Master

By choosing the long-term objective for the organization

By analyzing market data and customer feedback to suggest potential areas of focus

By ensuring the goal is met within the specified timeframe

By writing the goal so that it is impossible to fail

Zero-shot prompting

Few-shot prompting

Negative prompting

Chain-of-Thought

Commit the code; as long as the bug is fixed, it doesn't matter

Investigate the solution until they understand it, or seek help from a teammate, to ensure maintainability

Ask the AI to write a comment explaining the code

Mark the bug as Fixed by AI in the tracking system

Keeping humans inside the AI's processing unit

Ensuring a human reviews, verifies, and approves AI-generated content before use

A specific type of recursive algorithm

The team standing in a circle during the Daily Scrum

Ignore the data because AI cannot understand human emotions

Use the insight as a starting point for an open, human-to-human discussion

Record the frustration in the team's performance reviews

Ask the AI to write a list of rules to stop the team from being frustrated

By making it so humans don't have to talk to each other

By visualizing complex data and progress in ways that are easy for everyone to understand

By hiding the team's mistakes from stakeholders

By generating secret reports for the Scrum Master

The AI will become biased against humans

Humans will over-rely on AI suggestions and stop using their own critical judgment

The team will automate the wrong things

The Scrum Master will be replaced by a robot

Cancel the Sprint

Continue working manually, as the team remains responsible for the Sprint Goal

Wait for the tool to come back online before doing more work

Ask the Scrum Master to find a new AI tool immediately

Yes, for example, by requiring that all code be scanned by an AI-driven security tool

No, the Definition of Done must only include human activities

Yes, the AI is the one who decides if the increment is Done

No, AI is too unreliable for quality standards