AI並未讓工程團隊更快，而是迫使他們成長

Sign up

Sign in

Sign up

Sign in

AI Didn’t Make Engineering Teams Faster. It Forced Them to Grow Up.

--

Listen

Share

I’ve been around long enough to remember when talking seriously about generative AI in a business context was considered naïve. Three or four years ago (around the GPT-2 era) it was “interesting research.” A few of us kept pushing anyway, mostly quietly.

Now the tools work. The problems are different now.

What AI is changing isn’t how fast we write code. It’s how engineering teams (and increasingly, entire organizations) actually function.

The collapse of the linear lifecycle

Yes, small teams can now build in days what used to take months. Anyone still debating that is ignoring what teams are already doing. But the impressive part isn’t the volume of code; it’s the mode of work.

Traditional engineering assumed slow execution. That assumption shaped everything: long planning phases, frozen designs, handoffs between roles, and implementation as a distinct step.

In the AI era, that timeline has collapsed.

Engineering timelines are now much shorter. Thinking and building have become inseparable. Engineering is now a state of constant architectural judgment and tight human and AI feedback loops.

This isn’t chaos, but it looks like chaos if your team is still designed for 2024.

Software as shared substrate

Let’s get this out of the way: Software still matters. Quality, reliability, and security still matter. But software has shifted roles.

It is no longer the differentiator by default. It is the shared substrate everything runs on.

When everyone has access to the same models and stacks, outcomes diverge based on how people coordinate and decide around that code. Software is the terrain; outcomes depend on how teams make decisions around it.

When execution is cheap, coordination is the cost

Traditional teams were built to minimize the risk of slow, expensive implementation. This produced silos, layered approvals, and “process” as a substitute for trust.

AI undermines those assumptions. When execution is cheap:

If your team is still waiting for a weekly “sync” to make a decision, you aren’t just slow: you’re generating technical debt every hour you wait.

Intentional collaboration

The teams that adapt to this shift don’t just “use AI”, they redesign their operating model. They move away from role-based silos and toward a model of Continuous Architecture.

Architecture is no longer a diagram or a phase. It is a daily habit encoded in constraints and tests. The team shape that thrives is:

The most fragile teams right now aren’t the slow ones. They’re the teams moving quickly without shared judgment.

Automation as shared infrastructure

As automation becomes the operating system of the organization, it can’t live in private scripts or “innovation labs.” It must be shared, versioned, and collaborative.

This doesn’t mean everyone becomes a coder. It means everyone participates in removing low-judgment work from the system. What remains is the hard part: tradeoffs, creativity, and accountability.

AI didn’t make engineering easier. It exposed weaknesses in team structure.

When things break down, the cause is usually coordination and decision-making rather than the model.

--

--

Written by Marc Puig

Always learning.

No responses yet

Help

Status

About

Careers

Press

Blog

Privacy

Rules

Terms

Text to speech