Agile matters more in the AI era

Estimated reading time: 7 minutes

Key takeaways:

• AI moved your bottleneck out of coding. Requirements and testing now set your pace.
• Agile values matter more than its practices: sprints may go but collaboration and fast feedback still win.
• Speed raises the cost of building the wrong thing so knowing what not to build matters more.

---

With AI-assisted development rapidly becoming the new standard, developers are re-examining the traditional software development life cycle (SDLC) and exploring new methods of working. This change in how we generate code allows for faster feature releases, prompting the question: “do our current practices still make sense?”

Over the past year there have been many discussions about AI-assisted development and Agile. A recurring theme is whether Agile practices like sprints, daily standups, and estimation are still useful with this new way of working. However, the Agile Manifesto mentions nothing about writing code, working in sprints, or estimation.

With this in mind, I would argue that the Agile values and principles laid out in the Manifesto are more important than ever. Companies that only optimize code generation will quickly create bottlenecks in other parts of the SDLC. Practices may have to change with this new way of working, but the values themselves are still critical to success.

What has changed in the SDLC with AI?

The SDLC stages remain the same: analysis, design, development, testing, deployment, and maintenance. However, the effort has shifted away from development and onto the stages surrounding it. As a result, further improving development processes yields minimal efficiency gains.

Across my teams, time-consuming development tasks that used to take weeks can now reach the code review stage in a day with efficient use of AI agents such as Claude Code or Cursor. Batches of simple tasks, such as minor bug fixes, can be handled by multiple AI agents working simultaneously and be reviewed by a single developer in an afternoon.

What new problems does AI-assisted development create?

The output of the analysis and design stages is now one of the main inputs to an AI agent – “garbage in, garbage out” is more applicable than ever before. Requirements need to be crystal clear. Previously, misunderstandings could be caught early in the development process, but with AI-assisted development, these misunderstandings quickly turn into finished features.

The testing stage is critical to projects being successful. Developers would usually test their work while developing. Now, an entire feature can be built from one or two prompts and so there is a risk that nobody interacts with the feature in its entirety until late into the project. This makes testing the main way bugs or differences between the spec and the finished product are caught.

Customer management also matters more with AI-assisted development. Releasing features that don’t deliver customer value, no matter how fast, is a waste of time and can make customers feel like you’re not listening to their needs. From a customer’s perspective, releasing a feature that’s almost good enough but still does not solve their problem can be just as bad, or in some cases worse, than not releasing the feature at all.

How can Agile help?

The first line of the Agile Manifesto supports the move to AI-assisted development: 

“We are uncovering better ways of developing software by doing it and helping others do it.” 

To me, this captures what AI-assisted development is all about – building the right features more efficiently, not finding better ways to write code. If teams lean into the four values and the 12 principles of Agile, they will have the right mindset to address the new challenges.

Looking at the four values, each one can help with the new challenges a team may face when switching to AI-assisted development.

Individuals and interactions over processes and tools

Teams can produce code at a rapid pace, but without close collaboration, the work piles up at the review and test stages. I’ve seen development teams continue to push out code while the review column is filling up.

People assume that assigning a reviewer to the pull request is all they need to do. Instead, Agile encourages teams to talk to each other directly and make sure a review happens in a timely manner. This mindset means the feature is reviewed quickly and progresses closer to release. 

This can also be applied to testing. In teams with a dedicated quality assurance (QA) engineer, it is easy to allow tickets to stack up at the testing stage. Now that developers can produce features faster, the team needs to help with testing features end to end so work keeps flowing to release.

Responding to change over following a plan

With AI-assisted development it’s much easier to respond to changes in priority. However, teams can still succumb to the sunk cost fallacy and keep pushing to finish ongoing work before changing direction. 

In my experience, teams that use these new tools to change priorities quickly and refocus on the most valuable work are the ones that succeed long term. There is little or no value in finishing ongoing work if the need has disappeared or priorities have shifted.

Customer collaboration over contract negotiation

Now that we can build features fast, building the right ones matters more. Working closely with customers to clarify their needs and adapt when things change is much more valuable than focusing only on delivering exactly what was asked for with zero flexibility. As change is now cheaper, there is an opportunity for teams to provide better customer service and ensure customers get exactly what they need, not just what they asked for.

Working software over comprehensive documentation

As the cost of development is now tokens rather than time, getting a working iteration of software is much more valuable than focusing on complete documentation. Instead of trying to plan the whole project at the start, teams can document just enough to build an initial iteration of the feature, then use agents to rapidly build a working release. Teams can then show this working software to users to gather feedback early in the process and use it to plan future iterations. 

The 12 principles of Agile

Along with the four values, the 12 Agile principles give valuable guidelines that directly apply to the new world of AI-assisted development. While I am not going to list all 12, these three immediately jump out to me:

“Agile processes promote sustainable development. The sponsors, developers, and users should be able to maintain a constant pace indefinitely.” 

In a world where we can go fast, ensuring teams don’t burn out trying to do too much at once is critical to team health. AI-assisted development has accelerated team progress, but we need to remember there is a limit to how much context switching a person can do.

“Simplicity – the art of maximizing the amount of work not done – is essential.” 

When we can rapidly build anything we want, knowing what not to build is just as important. Features delivered fast with no product-market fit have no value and add maintenance overhead to the product.

“At regular intervals, the team reflects on how to become more effective, then tunes and adjusts its behavior accordingly.”

AI-assisted development is new, and the models are constantly changing. Regular reflection on how these tools are used and the impact they are having on the team, the company, and the product is vital for success. It also gives the team an opportunity to share the best way to use the tools while ensuring code quality remains high, and provides a mechanism for checking on team morale.

Why Agile still matters

AI has changed the way we work, but rather than making Agile irrelevant, it has magnified the parts of the SDLC that Agile can improve. The path to success is no longer optimizing code generation. Instead, we need to focus on Agile values such as individuals and interactions, and customer collaboration.

In my experience, teams that buy into the Agile values and principles, ensuring there is emphasis on close collaboration, fast iteration, and regular releases, get far more value from this new way of working than teams that don’t.

AI-assisted development is moving the bottleneck from the coding stage to somewhere else in the SDLC. Without Agile, teams risk getting stuck at these new bottlenecks rather than realizing the gains this new technology can offer.

In my experience, many companies fall into the trap of thinking they are Agile because they don’t do Waterfall. While that mindset produced some value in the past, it won’t deliver the same benefits going forward. Now is the time to start talking about Agile – discuss the values and principles, and how they can be applied to the SDLC in your teams.

The way we work is changing. The future may not involve sprints or story point estimation, but the values we work under will still matter.

Get articles like this in your inbox

Choose your LeadDev newsletters to subscribe to.