June 2026 | Agile, Scrum, AI, Software Development, Engineering Culture
Author: Dejan Milosavljevic — CEO of Singleton Solutions GmbH
I’ve been an unapologetic scrum advocate for fifteen years. I learned the methodology from Jeff Sutherland — one of the people who actually wrote it — and I’ve run it inside Singleton Solutions, scaled it across two scrum teams at comparis.ch and four teams at Adcubum, and delivered hundreds of sprints alongside more than a hundred different people. Scrum is in my DNA.
So when I tell you that scrum, in the form most teams are running it today, is dying — I’m not saying it as someone who never liked it.
I’m saying it because the world around scrum has changed.
Coding is no longer the constraint
The biggest assumption baked into traditional scrum is that writing code takes time. The two-week sprint, the story-point estimate, the burndown chart — they all exist to control the pace of human typing.
That assumption is breaking. AI coding assistants — Cursor, GitHub Copilot, Claude Code, and the agents that follow — are compressing implementation time by more than half. A feature that used to fill a sprint can fit in an afternoon. Boilerplate that used to take a junior developer two days takes a senior developer fifteen minutes.
If coding is the constraint, scrum is your friend. If coding is no longer the constraint, your two-week batch loses its meaning.
What’s dying: zombie scrum
Many teams are running what I call zombie scrum — the rituals continue, but the body is gone. Daily standups where nobody actually has news. Estimation sessions where the numbers no longer match reality. Feature factories where teams get credit for closing tickets, not for shipping outcomes.
Three things, specifically, are dying:
- Velocity as a measure of value. Story points were a proxy for human effort. With AI assistance, the same engineer can ship in an afternoon what used to take a week. The number stops meaning anything.
- Manual standups. When commits, pull requests and blockers all flow through tools that AI can summarise, an engineer reading aloud what they did yesterday is a tax, not a coordination mechanism.
- The feature factory. Output is cheap. Outcomes are not. Teams that keep being measured on output are about to look very busy and very irrelevant at the same time.
What survives: human empiricism
Here’s the part I want every developer reading this to take away. Scrum is not dying because its principles are wrong. Scrum is dying in the form of its rituals; its principles have never been more important.
Three pillars carry forward:
- Product ownership. AI is excellent at solving the problem you give it. It is terrible at choosing the right problem. Long-term strategy, user empathy and architectural intent are still human work. The Product Owner becomes more important, not less.
- Definition of Done. Mass-generated AI code has subtle structural flaws, hidden vulnerabilities, and a quiet tendency to erode architecture. A rigorous DoD — peer review, tests, security scans, accessibility checks — is the only thing standing between fast delivery and fast disaster.
- The sprint review. When teams iterate this fast, drifting from real users is easier than ever. The review is the human feedback loop that keeps “we shipped it” honest about whether anyone actually wanted it.
These three are scrum’s life support — and its competitive edge.
How roles are evolving
The roles that built classic scrum aren’t disappearing. They’re moving up the value chain.
- Developers become system architects. Less typing of syntax, more managing context, designing prompts, validating AI output and protecting domain integrity. The job is judgment, not keystrokes.
- Scrum masters become AI-flow co-pilots. Less Jira admin and meeting facilitation, more reading dashboards that AI compiles, spotting bottlenecks before they happen, and protecting team psychological safety while the work itself reshapes.
- From sprints to continuous flow. Many of the best teams I see are bending scrum toward Kanban or Lean, with continuous deployment and dynamic refinement instead of marathon bi-weekly sessions.
What this means for our team at 4IT
We’ve been a scrum house for over a decade. We’re not throwing that out. We’re keeping the parts that work — the empiricism, the DoR/DoD discipline, the review as a real conversation with stakeholders — and we’re letting go of the parts that no longer match how the work actually gets done.
If you’re a developer joining our team in the next year, expect three things:
- You’ll spend less time writing every line and more time deciding what should be written.
- Your DoD will be stricter than anything you’ve experienced, because the only way to ship fast is to ship clean.
- Your judgment, not your typing speed, is what we’ll invest in.
This is the most exciting moment to be an engineer in a generation. The constraint moved. Coding is no longer the bottleneck. Judgment is. Scrum, properly evolved, is still our best framework for sharpening that judgment in a team.
That is why scrum will survive. Just not in the form most teams are running it today.
Nazad na blog