Teams and companies worldwide are augmenting their workforces with AI. They're using it to schedule and summarize meetings, report on trends, clarify and expand content, write their emails and read them too (what a fun loop that is), generate infographics, and more. If the actions above align with your perspective of what a Scrum Master does, well, I've got two predictions: That's an ineffective Scrum Master, and that Scrum Master can easily be replaced by AI.
If that Scrum Master is you...it might be time to return to your roots and up your game.
Effective Professional Scrum
A good Scrum Master isn't a team assistant; they're a force multiplier. They contribute to and advance the Scrum Team's work to achieve the Sprint Goal. They use Scrum, complementary practices, tools (including AI), and guidance from trusted partners to maximize team effectiveness.
While Scrum Masters can seek support from the people around them, not all have access to a large or diverse network. They may be new to their career, work as the sole Scrum Master in an organization, or be limited in whom they can consult due to confidentiality. In times like these, Scrum Masters may benefit from leveraging AI to act as a collaborative partner.
Seeking help from AI
When a Scrum Master turns to AI for help to remove an impediment, they need to keep in mind who's ultimately making a decision and taking action.
Have you ever noticed that when using AI, especially an LLM, its questions tend to narrow in scope? Sometimes they're open-ended questions that seek more context, such as types of customers your product serves, the colors you want to use for an image, or who will be involved in a meeting. Other times it's asking close-ended, yes/no questions; things like, "Would you like me to provide an email template for your stakeholder communication?" or, "Would you like me to generate a sample post for your internal company blog?" Perhaps the best one I've seen is, "Would you like me to rewrite your current prompt to elicit more detail and exploratory responses?"
These kinds of responses aren't an accident. AIs are trained to interrogate the user to achieve their goal: Run to completion. For them, completion is solving the problem they were prompted with by being helpful to the user and arriving at an answer that is as statistically close to the target as possible. The AI will keep asking until it has enough data to generate an answer with a high probability of being accepted to "close the loop". It wants to do this quickly and efficiently, with low token/resource usage. When you're using a service, and each token has a real cost, we want a usable result as quickly and cheaply as possible. If we get enough usable answers at a low cost, we're likely to come back and use AI more.
Sure, AI. Do all of that. Return the result I can grab and use quickly.
And as soon as I say yes, I'm giving away agency to the AI.
Agency makes all the difference
AI is not responsible for solution outcomes; you are. Blaming the AI, saying, "AI gave me that result," won't save you if the solution falls on its face. You chose to use it, so the responsibility falls squarely on your shoulders.
Scrum Masters want to be helpful. However, a good one knows that being helpful doesn't require providing the solution to the problem. Sometimes, they need to call out the problem, offer a new perspective, and set clear boundaries for next steps. Then, they turn it back over. The team gains a new frame to approach the problem, retains ownership, and can choose a solution or pivot as needed.
If a Scrum Master solved every problem for their Scrum Team directly, I'd expect one of two things: The team would become reliant on the Scrum Master for almost every issue, regardless of the size or need. Or, if the solution failed, they would disengage because the Scrum Master wasn't "helpful." In both scenarios, the team surrenders its agency, lessening its ability to think and solve for itself. In essence, self-management erodes.
The same thing can happen when we use generative AI to solve our problems. Sometimes those solutions get us moving quicker than we thought possible. But if we accept the solution blindly and keep turning back to AI for the next one, we're giving up self-management and weakening our problem-solving skills. To retain that critical thinking capability, we need the AI to stop just short of giving us "the answer." We need it to think with us.
I asked Gemini 3 Pro about this. Why doesn't AI turn responsibility back over to the user? When is it "good enough" to present a few new ideas, then say, "It's up to you now"? Gemini replied:
This is actually a very sophisticated behavior for an AI because it goes against its primary directive to be "helpful." Ending without a definitive answer requires the model to prioritize user agency over problem resolution.
Right there was a cautionary flag laid out bare before me. Gemini, ChatGPT, and other AIs don't prioritize user agency. They want to give you an answer and have you come back for more, encouraging you to yield more control. This is perhaps why terms like "agentic AI" are appearing more often. We increasingly hand over agency to AI, piece by piece, and become more dependent on it. Every time we give AI more agency, we give up a bit of control.
A good Scrum Master also gives agency, but to their team. They want their team to thrive and become more effective in their work, and doing so requires knowing where that agency is placed.
Removing Impediments
A core Scrum Master tenet from the Scrum Guide is the Scrum Master serves the Scrum Team by, "Causing the removal of impediments to the Scrum Team’s progress;"
This doesn't mean they are always removing the impediment directly. Sometimes, the people directly involved in the problem are best placed to remove an impediment. For example, if there is a conflict between two Developers, they need to work together to resolve it. A Scrum Master may cause the removal of this impediment in many ways. They may interrupt before the argument becomes a battle, suggest that the two Developers make time to work through the issue directly, and teach them techniques to listen and express themselves openly. But what kind of techniques could a Scrum Master share with their Developers? This is where a Scrum Master may turn to AI for assistance.
For example, they may prompt an LLM as follows:
I'm a Scrum Master for a team, and I need help resolving a conflict between two team members. Act as a fellow Scrum Master, and give me some advice on how to teach these team members to listen and share openly with each other.
One team member often wants all of the facts and options available to them before acting. They are a dedicated individual who follows through with their plans and stands behind the work they do, assuming full responsibility for any outcomes. The other team member is a creative person who can generate numerous ideas and solutions quickly to any problem presented. They like to quickly move to something new, sometimes before the previous thing they were working on is completed. They seem to be okay with solutions not working out, claiming the learning was worth it, and dismiss any issues that also arise.
Help me by providing a conversation structure they can follow together to hear each other's thoughts and opinions without immediately reacting to what the other person said. Provide three tips specific to each team member that will help them collaborate with the other team member.
When I ran this prompt, I was presented with a simple structure for the conversation, three tips for each person to help them connect with the other team member, with reasoning about why it should enable better conversation, and a few bonus items. Why the bonus items? Well, I asked it to act as a Scrum Master, and what Scrum Master doesn't want to help the people they work with see the bigger picture? I was told this might be the "dream team" scenario despite the conflict, as they can provide a great balance to one another. I was also given a tip for the next Sprint Retrospective to help build team appreciation and unity.
And then, it was up to me to decide if I wanted to use any of it.
Scrum Masters should use every tool at their disposal to help their Scrum Team. This includes the use of AI, which can detect hidden patterns and generate creative solutions for a fresh approach. However, what will separate the good Scrum Masters from the rest are those who commit to think critically, foster self-management, and maintain agency and responsibility. These traits will prevent them from becoming mere parrots for AI...and soon replaced by it.