Skip to main content

AI is no longer just a concept for the future; it’s here and it’s transforming software development right now. AI tools are helping teams generate user stories, estimate tasks, and automate many of the boring admin jobs. This makes the whole process faster and more fun. As AI continues to improve, it will likely change traditional roles and practices, allowing teams to focus more on creativity and innovation.

To stay ahead, Agile teams need to embrace these AI tools. The future of software development is here, and it’s powered by AI—get ready for an exciting ride!

AI-Enhanced Software Development

Modern foundation models can create generative AI, which can produce new images and original text. These models do more than just summarizing articles or writing poems; they can also write software.

GitHub Copilot and Its Origins

One of the most popular tools for AI-assisted software development is Microsoft’s GitHub Copilot. Microsoft acquired GitHub, a major open-source repository, in 2018. Developers use repositories to host and share their code, which is crucial for open-source projects. Although most software is protected by copyright, developers can often use open-source software in their projects.

Microsoft utilized GPT-3 (Generative Pre-trained Transformer version three), a foundation model trained on the vast amount of code available on GitHub—about 54 million repositories, translating to hundreds of gigabytes of software. This is the same technology behind OpenAI’s ChatGPT, but in this case, it’s used for creating software rather than summarizing reports.

These foundation models are particularly well-suited for software development. Most developers don’t write code from scratch; they rely on established practices or adapt ideas from other developers. The generative AI systems can write software based on vast amounts of experience, far beyond what any single developer could achieve in multiple lifetimes.

How GitHub Copilot Works

To use GitHub Copilot, a developer simply provides a short prompt, such as “create a string to connect my database using Python 3 programming language.” Copilot then generates a block of usable code. While the code may need some editing to fit the specific product, it significantly boosts productivity, often by more than 50%. Many developers now see their role as editing software rather than writing it from scratch.

The skill set required for software development is evolving. LinkedIn’s researchers analyzed over 500 skills and found that AI could potentially replace 96% of software development tasks, especially for translating projects into different programming languages.

Implications for Agile Teams

This shift will profoundly impact Agile teams. Agile methodologies, particularly Scrum, advocate for small, collaborative teams. With AI-enhanced tools, these teams can be even smaller and more productive. In the future, a few developers, or even a single developer working closely with AI, could manage large software projects.

AI is poised to dramatically reduce the number of developers needed in organizations. As a result, embracing AI-enhanced development tools will be crucial for maintaining competitive and efficient software development teams.

Intelligent Software Repositories

Imagine a well-run development team using Scrum to embrace an Agile mindset. Typically, this team might consist of five developers, one ScrumMaster, and a product owner. They often use tools like Jira to track their software development progress. Communication happens through Jira notes, emails, or Slack messages. The team collaborates closely, sharing knowledge and cross-training newer developers. Each member might specialize in different parts of the software, working together on tasks. They rely on the product owner for business value questions and the ScrumMaster for guidance on upcoming events. The whole framework is designed for self-managed, closely collaborative teams.

Now, imagine changing the tools a bit. The software repository now has its own intelligence, like Microsoft Visual Studio or Cody. This AI-enhanced repository contains all the information about the product. New developers can directly ask the repository about the product’s structure. If they encounter issues, they can query the repository for explanations or solutions. Previously, developers would have to seek answers through emails or Slack, but now the AI repository provides real-time responses, making collaboration less necessary.

Benefits for Distributed Teams

This is particularly beneficial for organizations with distributed teams, where members work from different locations. Scrum traditionally discourages distributed teams due to collaboration challenges, but many teams still use them for cost and expertise reasons. Intelligent repositories can resolve many issues in distributed collaboration, serving as a central knowledge point. Essentially, an intelligent repository becomes the smartest and most up-to-date team member.

This shift will fundamentally change the rhythm of your Scrum team. Instead of closely collaborating, each team member will have a direct relationship with the AI repository, similar to the central AI in the movie “Alien,” where each crew member asks the AI specific questions. For instance, I once worked on a Scrum team where developers in Australia faced delays because they had to wait for U.S. developers to answer their questions. An AI software repository can provide real-time answers, making the process more efficient.

With an AI repository, the need for traditional Scrum practices diminishes. Daily stand-up meetings may become unnecessary when all information is accessible from the repository. Small teams might not need a ScrumMaster. This new AI tool, combined with the trend towards distributed teams, could significantly alter decades of established Agile practices.

Self-Generating User Stories

If you’re part of an Agile team, you’re probably familiar with user stories. These are brief notes meant to spark team discussions, providing just enough detail to convey business value without delving too deeply into functionality. User stories set the pace for the team, similar to how a bass player sets the rhythm in a jazz band. They help identify, organize, estimate, and prioritize work.

Writing user stories isn’t particularly difficult because they follow a standard format: describing the user, the goal, and the value. However, it can still be time-consuming. The Scrum product owner is crucial in creating and organizing these stories.

AI-Assisted User Story Creation

Given the standard format of user stories, it’s easy for a large language model like ChatGPT to generate them. For example, if you were developing the video player for this article, you could prompt ChatGPT with something like, “Can you write five Agile user stories for a system of educational videos with an accompanying real-time transcript?” The AI might respond with stories like, “As a student, I want to watch educational videos with real-time transcripts so that I can better understand the content.” This follows the typical user story format: the student is the user, watching videos is the goal, and better understanding the content is the value. The AI could even create acceptance criteria, such as, “When I open the video player, I should see an option to enable or disable real-time transcripts.”

You can ask the AI to generate a large number of user stories—50, 500, or more—with just a simple prompt. Providing more detailed information or creating personas can help the AI customize the stories further.

Automation and Integration with AI

In the future, AI systems might generate user stories automatically without prompts. Today, AI can summarize Zoom meetings, and this technology could be used to create user stories from those meetings. AI repositories might allow plugins for this purpose, converting any meeting into actionable user stories that are automatically prioritized.

A product owner could use a smartphone to record a meeting and have the AI create user stories based on the conversation. These stories don’t need to be perfect, just close enough to be easily edited, saving significant time. As AI systems improve, they’ll learn from feedback and become even more effective.

Large language models and generative AI are poised to revolutionize how Agile teams write and prioritize their user stories, making the process faster and more efficient.

AI-Powered Estimation

In 2023, two Australian academics explored the use of AI for estimating user stories. They employed a large language model (LLM) to assign story points to over 20,000 user stories—a task that would take humans years to complete. They then compared the AI-generated estimates to those made by Agile teams.

The study revealed that the AI’s estimates were 34% to 57% more accurate than human estimates. Additionally, when asked if they would use AI for estimation, 69% of the Agile team members indicated they were willing to adopt such a system.

Using AI for Estimations

Today, you can ask systems like ChatGPT to estimate story points. For example, you could prompt, “Can you estimate the user stories you just created using story points?” The AI typically starts with a disclaimer about the subjective nature of story point estimation and the importance of team collaboration during sprint planning. Despite this, the AI proceeds to assign estimates and provides justifications based on the complexity of the task.

For instance, the AI might assign a three-point estimate to a user story about implementing real-time synchronization of transcripts with videos. It justifies this by noting that while the task involves both front-end and back-end development, it is a common feature in video platforms, making the estimate reasonable.

Story point estimation is crucial for team collaboration. It’s not just about assigning points but about achieving a shared understanding of the effort required to deliver the product. Although the AI’s estimates might not be perfect for every team, they provide a helpful starting point and save significant time.

The AI’s ability to understand that story point estimates are relative to other stories is particularly valuable. Relative estimation is a core concept in Agile that often requires extensive coaching with human teams.

Using AI for Scrum stories highlights both the system’s strengths and limitations. One major strength is its ability to overcome the “blank page problem,” where creators struggle with starting a new task. AI systems can quickly generate initial content, which is often easier to refine than starting from scratch. However, to get the most out of the AI, you need a good understanding of what you want, as the AI’s suggestions still require human oversight and refinement.

AI-powered estimation can significantly enhance the efficiency and accuracy of assigning story points in Agile teams. While not perfect, these systems provide a valuable starting point and help teams save time, allowing them to focus more on collaboration and refinement.

Intelligent Tools for Administration

In 2019, the Gardner group predicted that AI would take over about 80% of project management within 10 years. While Scrum is different from traditional project management, it still involves considerable administrative tasks.

Scrum teams often deal with extensive communication, summarizing decisions, and using tracking systems like Atlassian, Jira, Trello, Microsoft Project, or even story cards. They also manage logistics like booking conference rooms, scheduling Zoom or Microsoft Teams meetings, and handling user story spikes—research-focused tasks designed to answer follow-up questions.

Automation Potential with AI

Many of these tasks can be automated with AI. Most companies are already integrating AI tools into their products and services. For instance, Microsoft researchers found that 80% of people would use AI to summarize meetings and generate action items, and 77% would like AI to help plan their day. Agile teams in the AI era will benefit from these advancements.

AI in Task Management and Meetings

Task management software will follow up with developers on their progress, while online meetings will be summarized, with AI even recommending follow-up meetings. Sophisticated chat systems will conduct online research and uncover new insights within the organization.

Daily Stand-Up Meetings

AI will significantly impact daily stand-up meetings in Scrum, which update everyone on work status. An AI system can manage these updates, informing team members about specific tasks and reaching out when work is abandoned. Developers might chat directly with the AI repository for updates.

Backlog Refinement Meetings

In backlog refinement meetings, many teams will use AI to draft user stories and provide estimates. The system can then automatically prioritize tasks. This automation reduces administrative overhead, allowing teams to focus on product delivery.

Embracing Automation

While some Agile coaches may have reservations about automation, most team members will likely welcome the opportunity to offload mundane tasks. This will give them more time to focus on engineering their software. The AI system will act like an administrative version of Google Maps, providing directions that are easier to follow than devising a new route.

AI systems might not be perfect at managing work, but they should be good enough for teams to follow their guidance. This will simplify administrative tasks, allowing Agile teams to concentrate more on their core responsibilities.