So when I think of AI and AI-First transformation, I get excited. Victory dance excited. “Yes! Finally Agile will get its moment to shine bright like a diamond and I won’t have to convince people how beneficial it is.”
Cue Amie, spinning in glee, arms raised, streamers flying, music blasting, and shouts of hallelujah when suddenly the music screeches to a halt, dancing stops and her head snaps back to the computer, in shock and disbelief.
LinkedIn feeds saturated with posts and articles stating. “AI killed Agile”; “Agile is dead”; “Scrum is obsolete”; “Sprints are over”. What?? How can this be? How can people get it so wrong?
My happy dance quickly turned into furious typing. Consider this post my letter to the editors. What are you all talking about? Agile has never been more relevant! It is the only way I can think that will bring AI into a product and engineering team quickly, safely and consistently. Without Agile, AI will create confusion and frustration within the team. Individuals will use it (or misuse it) their own way, with no guardrails for organizational constraints like yearly and quarterly goals, or budgets and security. The result is chaos at best, security breaches at worst, and company misalignment for certain.
Without Agile,
- How else can your team experiment with tools, configuration, and adoption in an ever changing landscape, and recover quickly from the mistakes along the way?
- How else can you measure the results of each experiment and report those back to the stakeholders mandating an “AI-First” organization?
- Where else can a team learn new concepts collectively, like AI evals, Retrieval Augmented Generation (RAG), Answer Engine Optimization (AEO), context engineering, agent guardrails, and token budgets, and make them a standard part of the SDLC?
Let’s not forget the core foundation of Agile, which is to expect change and create an environment that adapts to it in a fast but structured and measurable way. It encourages planned experimentation and iteration on the results of those experiments in small cycles, so a team can discover ways to improve its productivity and synergy without sacrificing time, money, and sanity.
As a servant leader, I let teams discover for themselves what needs improvement and adjust their own processes accordingly. However, as a seasoned Agile coach, I can usually predict how those experiments will go and which adaptations the team will land on. Why? Because for decades, the changes in software delivery have been better versions of things we already did. Cloud, mobile, and CI/CD, for example. All are improved tools and processes to what we already knew. None of them changed the core of how software was being built. AI is a whole new software development ballgame. It is not only a super sized upgrade to our tools but also a brand new addition to the scrum team, one with a quirky, unpredictable personality none of us have worked with before. This is exactly the kind of disruption Agile was designed for.
For a solo founder, AI can easily be adopted, iterated, and successfully incorporated quite rapidly, while also overseeing the costs to run it and the boundaries to keep around it. Teams are now at a competive disadvantage with solo creators who can produce like a team without the struggle to maintain shared understanding and cohesion with AI usage. This is a big paradigm shift. Until now, teams could succeed with poor Agile practices and still have an advantage over a solo founder doing it all on their own. For a team to stay ahead of this shift, proper Agile delivery practice will be the only lifeline to ensure successful AI adoption that maintains their advantage over leaner solo creators.
The people claiming AI killed Agile are either blindly coasting on a finely tuned Agile autopilot, without understanding the why behind any of it, or were never doing Agile right in the first place. Either way, blind coasting and improper practices are no longer an option.
The teams that learned the Agile principles and follow a real sprint cadence can easily fold AI experimentation into their sprints, and build a solid AI best practice into their process. The teams that have only been labelling their meetings as Agile ceremonies, without the mindset or the rationale behind them, are in trouble. They can relearn the core principles and scramble to catch up. Or they can declare Agile dead, throw it out, and find out the hard way what happens when everyone on the team is running their own AI, their own way, with no shared flight plan.
Agile is far from dead. It has gotten a second wind and is back in the pilot’s seat, wide awake, and ready to fly to the rescue for teams that want to win at AI-first delivery. Drop the mic and cue in the music.
P.S. I asked AI to argue with me
Before posting this, I asked Claude to read it and tell me where it disagreed. It pushed back on three points. Here is the short version of that debate.
AI: You say Agile is the only way a team can bring AI in safely, and the only lifeline for AI success. “Only” is too strong. What a team really needs is fast feedback, measurement, and retrospection. Lean, Kanban, and good DevOps cultures have those too, and many readers will assume that by Agile you mean Scrum.
Me: Fast feedback, measurement, and retrospection are the principles of Agile, so that is not an argument against it at all. And Agile is both Scrum and Kanban. Anyone who thinks it is one or the other is incorrect. And I don’t see how a good DevOps culture replaces it.
AI: Some of the “AI killed Agile” people have a point. Two-week sprints feel long when a feature takes an afternoon, and story points stop meaning much.
Me: Agile is a framework built on principles. Anyone who claims two weeks is an Agile rule is doing Agile wrong. Story points only stop meaning much if they were never assigned properly in the first place. They show whether a team is looking at the work the same way, against a shared definition of done. Including AI in the estimate does not eliminate that.
AI: Agile doesn’t supply the budget and security guardrails. Those come from governance and policy.
Me: Agile planning is based on goals within constraints. That is where the guardrails go, so the whole team is on the same page about what success looks like.
AI: One caveat. A team’s velocity will need a new baseline once AI takes on part of the work.
Me: A team doesn’t create a velocity baseline. The data tells you your velocity and you plan to that. Velocity from before AI is still valid, and shaky velocity during adoption is expected. My bet is that the work shifts, the results get bigger, and velocity stays about the same.
The result? It conceded on all four.
Try it yourself. Hand your strongest opinion to an AI and ask it to argue back. And if you disagree with me, tell me on LinkedIn. I’m up for the challenge to defend my strong viewpoint on this debate.
About the author
Amie Walton is the founder of TriBic Software and an Agile coach with nearly 30 years in software. She helps product and engineering teams get Agile right and adopt AI safely. To talk about Agile and AI adoption for your team, reach her at info@tribicsoftware.com.