Why Some Org Transformations Stick: Lessons from DevOps and Agile for the AI Era

Something feels different this time.

If you lead an engineering organization, you've probably felt it. The LinkedIn posts about "10x engineers" using AI to replace entire teams. The CEO memos requiring proof that AI can't do a job before approving headcount. The quiet realization that the org chart you're building for next year might look nothing like the one you have today.

There's genuine unease about what software teams will look like in five years. And unlike previous technology shifts, this one feels less like an upgrade and more like a restructuring of the work itself.

AI and agents in software development are inevitable. That much is clear. What's less clear is what that actually means for the humans involved. Will your team shrink? Will roles merge? Will the Scrum Master who facilitates your retrospectives still have a job in 2030? Will you?

These are real questions, and they deserve better answers than "AI changes everything" or "nothing to worry about."

Before we can predict where AI takes us, we need to understand why some organizational changes become permanent while others fade away.

What Makes a Transformation Permanent

The answer isn't complicated: transformations stick when they allow organizations to genuinely do more with less, without tripling workloads for individuals.

This sounds obvious, but it's important. The changes that last aren't the ones that simply cut costs or consolidate headcount. They're the ones that actually remove friction, eliminate unnecessary coordination, or automate work that nobody wanted to do in the first place.

DevOps Didn't Just Cut Release Managers

When Amazon implemented "You Build It, You Run It" in 2006, they weren't simply eliminating the operations department to save money. They were removing a handoff that created delays, blame cycles, and knowledge gaps.

Werner Vogels stated it plainly: "There is no separate operations department at Amazon. You build it, you run it."

The transformation stuck because developers who owned their code through production made better decisions. Fewer outages. Faster recovery. Better software. The Release Manager role didn't disappear because it was expensive. It disappeared because the separation between "people who write code" and "people who run code" was creating problems that integration solved.

Netflix's seven-year cloud migration followed the same pattern. They didn't just move to AWS to cut data center costs. They restructured around the principle that teams should operate what they build. The result was faster iteration, better reliability, and an organizational model that competitors spent a decade trying to copy.

Agile Didn't Just Eliminate Project Managers

When organizations adopted Agile in the 2000s and 2010s, the successful transformations weren't about removing PM headcount. They were about shortening feedback loops.

The waterfall model required extensive upfront planning because course corrections were expensive. Agile made course corrections cheap. That meant less need for detailed project plans and more need for rapid iteration. The Project Manager role evolved into Scrum Master not because PMs were expensive, but because the work changed.

Marty Cagan documented what happened to PMOs in this transition: many were "pushed aside from software efforts and relegated to orchestrating moves from one building to another." Not because companies wanted to eliminate them, but because cross-functional teams with direct customer access didn't need centralized coordination in the same way.

The Pattern: Remove Friction, Transformation Sticks

The transformations that became permanent share a common thread: they removed friction that was slowing down valuable work.

DevOps removed the friction between development and operations. Agile removed the friction between planning and building. Platform Engineering is removing the friction between developers and infrastructure.

Each transformation eliminated coordination roles that existed to manage boundaries. When the boundaries themselves dissolved, the roles that managed them became unnecessary.

This brings us to AI. The question isn't whether AI will transform software organizations. The question is: what friction does AI remove? And what roles exist primarily to manage that friction?

Why Companies Were Already Cutting Middle Management Before AI

Here's something that gets lost in the AI conversation: significant organizational consolidation was underway before ChatGPT learned to write code.

In January 2023, Capital One eliminated 1,100 positions. Not a general layoff. A targeted elimination of an entire job family: Agile Delivery Leads, Agile Coaches, Scrum Masters, and Release Train Engineers.

Their statement was revealing: "The Agile role in our Tech organization was critical to our earlier transformation phases but as our organization matured, the natural next step is to integrate agile delivery processes directly into our core engineering practices."

This happened before AI coding assistants became mainstream. The stated reason wasn't technology. It was organizational maturity. Teams had internalized Agile practices. They no longer needed dedicated facilitators.

Royal London Insurance made 90% of their Scrum Masters redundant the same year. A large UK bank cut roughly 1,000 similar positions. None of these decisions cited artificial intelligence.

The training data tells the same story. One veteran Agile trainer reported that entry-level Scrum Master certification classes represented 49% of students in 2020. By 2024, that number had collapsed to less than 5%.

Meanwhile, middle management was already flattening. Mark Zuckerberg's March 2023 "Year of Efficiency" memo reads like a manifesto for organizational consolidation:

"Flatter is faster. It's well-understood that every layer of a hierarchy adds latency and risk aversion in information flow and decision-making. I don't think you want a management structure that's just managers managing managers, managing managers, managing managers, managing the people who are doing the work."

Meta eliminated 21,000 positions and explicitly asked middle managers to become individual contributors. Google cut 12,000 jobs. Citigroup removed five management layers and eliminated 60 committees. Bayer announced plans to roughly halve their management ranks.

These weren't AI stories. They were efficiency stories. Post-pandemic correction stories. But they established a direction that AI is now accelerating.

The Stage Is Set

So if consolidation was already happening, what specifically does AI change?

That's where task-level analysis becomes essential. In Part 2, we'll examine over 100 discrete tasks that make software teams function, from writing meeting notes to debugging production issues, and assess which ones AI can reliably handle today, which it will handle soon, and which remain stubbornly human.

The pattern that emerges is clearer than the hype suggests. And more consequential.

[Part two coming soon]