Over the past three years, we have continued to observe the transformation of Scrum and Agile in our organizations. Agile methods encourage teams to embrace change, build in increments, understand your flow of creation, and adapt to create positive change when needed. We have had some consultants claim that Agile is dead because DevOps will save us. For me, I see DevOps as a collection of practices, coaching, and experience that helps my teams improve our flow from idea, to verification, to deployment to production. Devops continues to be a logical extension of the agile movement. In this post, we will explore how the disruption of GenAI will transform the craft of agile teams.
As a proud Scrum master for years, I love to remind folks that Scrum was not designed to be the perfect process for creative project management. The authors of Scrum openly discuss that the framework has just enough rules so that teams start to show their dysfunctions. I do feel we need to keep this in mind. The most vital meeting for agile teams to keep having is their sprint retrospective or regular operational review. In retrospectives, teams openly discuss the events that were life giving in their work together. In retro, teams have a safe space to talk about their challenges and friction points with process. In general, teams should have the opportunity to openly talk about their dysfunctions and find new processes that work for them.
While doing some reading this week, I became challenged by this article: The End of Agile: When the Assumptions Beneath Your Methodology Collapse.. In brief, the post implies that GenAI tools have brought about solutions to software team challenges and dysfunctions. The post strongly states that agile should not be encouraged with teams since AI brings about transformative capabilities that help us create value faster.
I want to affirm some aspects of the post. When Agile started, the industry needed strong solutions to avoid the waste and challenges of “waterfall/plan based methods.” The agile movement focused on “optimizing human performance.” The author makes a great point that team tacit knowledge will need to become documented so that AI tools can leverage the wisdom, insight, and standards of the team. As the industry starts building tools like Gemini CLI, Claude Code, SpecKit or Google AntiGravity, we can see the industry adopting the practice of adopting code and architecture standards in written form so that AI tools can comply with team norms. As our AI journey continues to evolve, teams will continue to discuss how we might evolve from vague user stories to more precise statements of requirements or intent. Requirements management and requirements development quality will continue to be positive conversations in the future.
I really like the idea of restructuring processes around verification and validation. If team flow can dramatically increase during coding because of GenAI, teams need to create ways to rapidly complete testing, enforce quality, and have frameworks for the management of AI quality. How do you know that your AI chat bot offers correct direction in every case? How do you know that your AI system is grounded in your organizational policy? As an industry, we will continue to learn how to enforce quality on the sub-systems of AI that involve creativity and non-determinism.
Why do we still need agile?
Agile promotes positive habits: As a musician or artist, I know that I can only improve my craft through the habits of practice. Teams become superior by making a commitment to get 1% better every day. Agile reminds us of the value in regular planning, regular discussion of requirements, healthy discussion around schedule expectations, regular verification/validation, and reflection. Should our processes around planning, execution, and learning evolve in face of AI? Yes. I, however, feel that we need movements like Agile to encourage a human-centered approach to the transformation.
Agile can help us size increments to keep quality high: As I have played with GitHub Copilot, Gemini CLI, or Claude, I have observed that scope size is an interesting factor. If I ask Claude code to build me an RSS reader using clean architecture standards in a one shot spec with pretty good quality, I still find that I will have a result that feels “meh.” Why? I tried to do too much at one time. There are certain melodies of Star Wars that I love because of the emotional care and detail that John Williams crafted into his music. His goal was not just to fill dead space. He worked to make me feel an emotion appropriate to the moment of the movie. As software engineers, we need to work toward making products that make our users and stakeholders feel positive emotions. Sometimes we’ll make our users feel like super heroes. We might make our users feel a sense of delight or flow because of the playfulness of our system. With user experience, the details matter and require iteration. You probably will need to build, user test, measure, learn, and iterate. Agile can help us discover the right size of increment for these processes. Agile can help us understand the time/schedule complexities of these iteration sizes.
Agile makes us think about organizational change: Through the agile movement, we’ve learned that organizational change is hard. In my most successful agile transformations, I’ve observed the transformation starting with a small pilot team. The pilot team helps the rest of the organization feel the wins and benefits. Once the pilot team has helped the organization taste the value and benefits of the new agile habits, the agile transformation starts to spread to more teams in an organic manner. I feel like that same art of transformation is happening in large organizations. Especially in highly regulated and compliance driven culture, AI transformation requires the art of “lean change.” I acknowledge that AI code copilots give you feelings of hyper productivity, but software products are used by humans. In large schools, organizations, and cultures, we still need to be mindful of organizational change management and all the lessons that lean/agile taught us. I still believe that thoughtful “baby steps” and incremental change will continue to win the day.
Agile continues to advocate for human teams: In the post, there was a theme suggesting that human/organizational communication has become less of a concern. As mentioned before, software requirements specification and the related requirements development processes will guide the quality of output “spec driven” development. Keeping positive relationships with stakeholders and their organizations remains a uniquely human art. GenAI will not improve the quality of communication in organizations without intentional leadership and guidance. I believe it’s vital for us to continue to advocate for a human centric vision for work and economic development. We need movements like Scrum that teach us that we should value commitment, focus, openness, respect and courage. Many of the values promoted by agile and scrum have a vital human quality that leaders do not want to throw out.
Leave a Reply