AI-assisted development is prompting Agile teams to rethink Scrum practices, shifting focus toward code validation, AI governance and new uses for key Scrum events.
As AI-assisted tools become more common in software development, some Agile teams are reevaluating whether their current Scrum practices can support the speed and scale of AI-assisted development workflows.
Scrum is a product development framework that organizes work into time-boxed, iterative cycles called sprints. Scrum's short development cycles and regular inspection checkpoints provide a structured way for development teams to design and coordinate workflows while continuously evaluating progress and adapting plans as requirements change.
Scrum was designed in the early 1990s by Ken Schwaber and Jeff Sutherland to manage complex work in product development environments where requirements and solutions evolve through experimentation, feedback and incremental progress. The framework is based on empiricism, which asserts that knowledge comes from experience and decisions should be based on observation.
Scrum's creators provided structure for organizing work through defined events, artifacts and accountabilities, but purposely left decisions about workflow design to teams using the framework. Development environments can vary widely in terms of technologies, tools and engineering practices. Prescribing a fixed workflow limits a Scrum team's ability to adapt processes to the specific product that is being built and the constraints of the development environment.
In practical terms, the decision to keep the framework lightweight has made it easier for teams to implement Scrum in AI-assisted development environments. Because workflow decisions are left up to the development team, the framework can accommodate a wide range of software engineering models and best practices, including AI-native development.
AI creates new bottlenecks
While the flexibility that Scrum provides can make it easier to integrate AI into development workflows, integrating AI-assisted tools can shift where bottlenecks appear.
Many Scrum teams are discovering that throughput bottlenecks are becoming less significant, while new constraints are emerging around code validation, integration and AI governance.
When AI tools are used to generate code, documentation, unit tests and infrastructure configurations, the outputs still need to be reviewed, validated and integrated into the codebase. The challenge is ensuring the volume and speed of AI-generated output do not exceed the team's capacity to review and integrate changes safely.
The challenge is ensuring the volume and speed of AI-generated output do not exceed the team's capacity to review and integrate changes safely.
According to a recent study conducted by Clockwise, the average software engineer spends about 10.9 hours per week in meetings. That's nearly one-third of a typical workweek, leaving little time for the additional review and validation work that AI-assisted development can introduce into the workflow.
To meet this challenge, many Agile teams are going back to basics and rethinking how they manage Scrum artifacts, distribute responsibilities among team members and conduct Scrum events. Ron Jefferies, the co-creator of Extreme Programming, refers to this type of course correction as surviving dark Scrum.
Should Scrum teams still hold daily standups?
One of the first events that many teams are reevaluating is the daily standup.
According to the official Scrum Guide, the purpose of the daily Scrum -- also known as the daily standup -- is for developers to inspect progress toward the sprint goal and coordinate the work needed to move closer to it. The problem is that in many organizations, these short daily meetings have become so ritualized that they interrupt work rather than help teams coordinate it.
Now that AI-enabled tools can be used to summarize and share progress by analyzing repository activity, issue trackers and CI/CD pipelines, it makes sense to review the original purpose of daily Scrums and change their focus to reflect the realities of AI-assisted development workflows. In practice, this means shifting the focus away from individual status updates and using the time to coordinate which AI-generated changes require review, whether any automated outputs need additional testing and how/when to implement AI governance checks into the workflow.
How AI is impacting other Scrum events
The growing acceptance of AI in development workflows is also encouraging organizations to reevaluate how they use other Scrum events.
For example, teams may spend less time during sprint planning ceremonies discussing the mechanics of work and more time strategically coordinating the use of AI and redefining the Definition of Done as they experiment with new AI tools.
During sprint retrospectives, the focus might shift from general process improvements to more targeted discussions about how AI tools are affecting the development workflow.
The growing importance of sprint reviews
As teams experiment with how they use Scrum events, however, one event is getting attention in many organizations -- the sprint review.
According to the Scrum Guide, the purpose of the sprint review is to inspect a completed increment and adapt the product backlog if needed. The guide specifies that reviews should be held at the end of each sprint, but intentionally leaves the event format flexible.
In practice, however, sprint reviews have often become quite rigid in large organizations that scale Scrum across multiple teams. Typically, each review begins with developers explaining how they built the increment and demonstrating how it functions. Then the product owner shares the state of the product backlog, reports on progress toward the product goal and suggests possible next steps. After that, stakeholders are invited to provide feedback.
This ritualistic structure has worked quite well for many enterprises over the years, but since AI-assisted development has significantly increased velocity, many Agile teams are rethinking how to use AI in sprint reviews to make the event more valuable.
For example, instead of walking product owners and stakeholders through every implementation detail, developers might use generative AI tools to summarize how the increment satisfies the Definition of Done, and the project owner could share an AI-generated progress report that recommends next steps. This would allow the team to use most of the event to discuss product direction, stakeholder priorities and the most valuable next steps for the backlog.
How AI is changing the Scrum master role
As organizations refine how Scrum events are conducted to support AI-assisted development environments, the Scrum master's role in guiding and facilitating inspection events is also evolving.
Some organizations have begun to question whether a dedicated Scrum master is still necessary. After all, the Scrum Guide defines Scrum master as a core accountability, but the framework does not require the role to be a dedicated full-time employee.
In some organizations, the Scrum master's accountabilities might be distributed across the team or carried out programmatically by orchestration platforms. Even in organizations that decide to retain Scrum master as a dedicated assignment, the role is likely to increasingly blend Scrum coaching with guidance on how AI-assisted development workflows are used within the team. According to ZipRecruiter, this will require future Scrum masters to have a strong foundation in Agile methodologies, project management and a solid understanding of AI concepts and machine learning workflows
Don't abandon Scrum -- experiment
As AI becomes more deeply embedded in software development workflows, organizations will likely continue experimenting with how Scrum is applied rather than abandoning the framework altogether.
The core principles of transparency, inspection and adaptation remain valuable in environments where both humans and AI systems contribute to the product increment. What is changing is the scale, speed and types of output that teams must evaluate.
By refining how Scrum events are conducted, extending the Definition of Done to account for AI-generated artifacts and evolving the Scrum Master's role to guide AI-assisted development practices, organizations can continue to rely on Scrum as a lightweight framework for coordinating complex work while maintaining quality and delivering value in increasingly automated development environments.
