Insights From a Scrum Master: Is AI a Competitor or a Partner?

Especially in recent years, many movies and news articles have come out that have been focused on the potential negative outcomes of artificial intelligence (AI) usage. Understandably, many people are afraid of being adversely affected by AI models in their workplaces.

Like many other professionals, scrum masters may fear that they’re soon to be replaced by AI. These fears can seem increasingly justified, particularly as AI's capabilities continue to grow more effective in supporting such efforts as automating routine tasks, providing efficient process management, and offering data-driven insights.

However, fear often prevents people from seeing new opportunities. As a scrum master, I don’t look at AI as an expert, source of truth, authority, or something that can replace me. Instead, I look at it as an assistant or a partner that can help me generate ideas, streamline tasks, and explore different perspectives.

I use ChatGPT-4, which is a chatbot based on a large language model (LLM), in my day-to-day work for the Rally team here at Broadcom. Below, I’ve included several different examples of the ways I’ve used AI, including creating documentation, facilitating meetings and retrospectives, running icebreaking sessions, and generating metaphors.

Important note: Due to privacy and security concerns, be very careful about providing information to ChatGPT or any other public LLM model. Sensitive data like personal details, internal processes, or confidential project information can be misused if exposed. Always ensure that shared information doesn't violate data privacy policies or compromise your team's or company's security.

Help with facilitation

As a scrum master, I facilitate a lot of team ceremonies and cross-team meetings in Rally. I use ChatGPT-4 for help with creating a facilitation plan for complicated meetings. For example, this can include meetings in which we’re tasked with developing working agreements or brainstorming. ChatGPT helps by providing structured outlines tailored to the meeting's objectives. It also suggests effective techniques for engaging participants, such as icebreakers, interactive discussions, and time-bound activities. ChatGPT can also offer guidance on allocating time efficiently for each agenda item, helping ensure we have a balanced and productive session.

Here’s an example of a prompt used to create a meeting facilitation plan:

Please design a facilitation plan for a 1.5-hour hybrid meeting. The objective is to have code owners focus on backend development and establish a working agreement. The meeting will start with a five-minute warm-up exercise to foster engagement, followed by roundtable discussions on these specific topics: communication issues, ensuring quick responses, and protocols for informing others when a pull request is being reviewed. The expected outcome is a comprehensive working agreement document, collaboratively created in Google Docs. The plan should include guidance on structuring the document, facilitating productive discussions, and integrating contributions from both in-person and online participants.

Every scrum master knows that repeatedly using the same format for retrospectives can cause teams to grow bored, compromising the effectiveness of the conversation. This is why collaboration tools such as Miro provide a lot of retrospective templates. These varied approaches can help to uncover different aspects of team dynamics and challenges, provide a more comprehensive understanding, and foster continuous improvement. This variety prevents the retrospectives from becoming monotonous, helping to sustain team interest and motivation.

I’m responsible for leading retrospectives every two weeks. Consequently, I sometimes feel like I’m out of good ideas for formatting these retrospectives or creating an interesting icebreaker. I provide ChatGPT with information about the problem I would like my team to focus on during the retrospective. For example, we may be looking at cycle time or an increase in the number of defects in the last iteration. Based on this, I will ask Chat GPT to help me generate ideas. If the retrospective happens around some specific holiday or season, or you would like to use a particular theme, like a song or movie, you can include this information in your prompt and ChatGPT will generate ideas for your specific thematic retrospective.

Following is an example of a prompt that helped me with generating retrospective ideas:

Please formulate three winter-themed retrospective strategies for a 10-person, hybrid software development team. Develop actionable solutions for increased story cycle time, featuring a five-minute winter-themed icebreaker and an appreciation segment. The session will run for one hour. We will use Lucidspark and Webex, with a mix of discussion and activities, while avoiding role-play.

I submitted the above prompt to ChatGPT and requested that it limit the answer to one sentence for each idea.

Following is the output I received:

Icicle Insights: Start with a "Frosty Facts" ice-breaker, having each person share their coolest winter fact, then dive into a collaborative analysis of the entire development lifecycle to pinpoint 'frozen' inefficiencies.

Avalanche Avoidance: After opening with a "Chill Challenge" where everyone names a winter item they could do without, the team collaborates on identifying and planning to 'melt away' the top-heavy processes threatening to 'avalanche' their cycle time.

Ski Slope Strategy: Initiate with a "Snowflake Sketch" ice-breaker, drawing unique snowflakes that represent individual team strengths, then strategize on 'skiing' down the slope of tasks faster by mapping out streamlined workflows in Lucidspark.

If you don’t like the output, ask for another three ideas. Pick one idea and ask ChatGPT to help you with the detailed facilitation plan and design of the Lucidspark or Miro board. With ChatGPT as a brainstorming partner, preparing for retrospectives became much quicker and easier.

Retrospectives often include discussions about some performance issues. Using metaphors can help me to have more effective conversations. Metaphors can help simplify complex concepts and make discussions more engaging and memorable. They can create a neutral space, shifting the focus from individuals to situations. Metaphors can help promote an open dialogue. Finally, metaphors stimulate creative thinking, leading to innovative solutions.

I explain the performance issue to ChatGPT and it helps me to create metaphors. I can ask for five options and then select and refine the one I like best. ChatGPT is an inexhaustible source of inspiration!

Here’s an example of a prompt that helped with generating metaphor ideas:

A scrum master wants to craft motivational metaphors centered around the theme of horror movies. The goal is to help teams navigate through the challenge of a significant increase in defects in code from the software development team. These metaphors will underscore the importance of teamwork in overcoming obstacles, akin to characters in horror movies banding together to face their fears and triumph over their challenges. The goal is to not only highlight the current issue of rising defects but also to inspire a unified, collaborative effort to work towards a solution, capitalizing on the team’s common interest in horror movies. Create five metaphors.

Help with process documentation

As a scrum master, I help the Rally team to continuously improve their processes. Consequently, I often write many different process documents. This takes me considerable time, especially as English is not my first language. ChatGPT helps me save time and effort.

For example, recently I wrote a multi-page document about the canary deployment process in Rally. At the beginning, I had very little knowledge about what a canary deployment is. ChatGPT helped me to learn about this process. It also assisted me with creating the draft document and then clarifying content in different sections. Rally developers added some specifics, but working with ChatGPT saved me a lot of time.

Help with training materials and Agile games

As a scrum master, I need to perform Agile training for new team members. ChatGPT can provide concise summaries of Agile principles, methodologies, and practices. It can also generate easy-to-understand explanations of key concepts like iterations, backlogs, and stand-ups. Explanations can be tailored to beginners or more experienced team members. Creating training materials has never been easier!

Here’s an example of a prompt use to help create training materials:

Please develop comprehensive informational training materials focusing on scrum events and their purposes. These materials should be tailored to an audience with some scrum knowledge. These materials are intended for both new team members and experienced Agilists within a software development team. The content will provide deeper insights into Scrum events, detailing their objectives and contributions to the scrum process. Additionally, include a section on the pros and cons of point estimations to offer a balanced view on this common practice within Agile planning. Each section of the training material is to be limited to 1,000 words. Incorporate best practice examples from the industry to illustrate effective execution of scrum events and point estimation strategies.

To help their teams and organizations learn agile principles and practices, scrum masters sometimes need to try something different. Agile games are a valuable tool for scrum masters in team coaching as they transform complex concepts into interactive, experiential learning. AI can assist a scrum master in developing an Agile game by generating creative ideas and customizing them to suit the team's needs. AI can help provide clear instructions, objectives, and rules for the game, ensuring it is engaging and easy to understand. AI can also create realistic scenarios for role-playing, reflecting real-life Agile challenges. I can do a Google search for some games suggested by other Agile professionals, which can be helpful. However, AI can help me get to another level, with a game customized for my team’s specific needs.

Recommendations on writing prompts to LLM models

Be specific and clear

Clearly state your request or question to increase the likelihood of receiving a precise and relevant response.
Example: "What are the key responsibilities of a scrum master during the sprint retrospective?"

Include context

Provide relevant background information to help the model more effectively assess the situation or problem.
Example: "Our agile team is transitioning from waterfall to scrum. What challenges should we expect during this transition, and how can we mitigate them?"

Define the desired output format

Specify if you need the response in a particular format, like a list, summary, or detailed explanation.
Example: "Can you provide a detailed explanation of the differences between Kanban and scrum, in a comparative table format?"

Limit prompts to one main idea

Focus on a single topic or question to avoid confusion and mixed responses.
Example: "How does the daily standup meeting contribute to the success of an agile project?"

Mention any constraints

If there are specific limitations, such as a maximum word count, or other requirements, include them in the prompt.
Example: "Explain how to scale agile practices for a large organization, in no more than 300 words."

How will AI be able to help in future?

As referenced in my note above, I can’t provide any sensitive information to public LLM models because of significant concerns around data privacy and security. However, what if this problem is resolved in the future and I have an enterprise LLM model that enables us to use the company's data securely? Rally already has an AI chat bot trained on Rally documentation and help data. We are now experimenting with using Rally’s own data and plain-English requests to create API queries.

There are many ways AI can help the scrum master in the future. Here are just a few examples:

• Tailored recommendations. Employ AI to analyze team performance metrics and obtain customized advice for improving efficiency and productivity.
• Customized training plans. Design training programs based on team skill gaps identified through AI-driven data analysis.
• Enhanced retrospectives. Utilize iteration data and AI to facilitate more insightful retrospectives with focused discussions on areas needing improvement.
• Continuous improvement. Use AI-powered insights into data trends to guide the team in adopting Agile best practices and continuous improvement strategies.
• Enhanced meeting efficiency. Gain AI assistance in developing meeting agendas and generating meeting notes.
• Conflict resolution assistance. Use AI, the team’s data, and best practices in Agile and team management to obtain actionable insights for optimizing communication strategies and mediation techniques.

Conclusion

AI is coming and we need to learn how to use it to our benefit. The relationship between AI and scrum masters is not one of competition, but rather of partnership. With its vast capabilities, AI serves as a powerful tool that enhances the scrum master's role, streamlining processes and fostering a more efficient and collaborative team environment.

AI can provide significant support and augment the scrum master's capabilities. However, the essence of the scrum master role, which is rooted in human interaction and empathy, means this job must remain firmly in human hands.

And yes—I used ChatGPT-4 for help in writing this blog.