The Mindset Shift That Separates AI-First Teams from the Rest

Rodrigo Vargas Rodriguez didn't set out to witness a revolution. With over twenty years in software development, he's seen plenty of technological shifts, but what's happening now feels different.

As leader of Encora's Gen AI practice, Rodrigo works with enterprises as they integrate AI into development workflows. From his vantage point, he's watching the software development world split into two distinct realities. Some teams have added AI tools and work incrementally faster. Others have fundamentally reimagined what software development is. The difference isn't about which AI tool they're using. It's about whether they've been willing to challenge decades of ingrained practices.

What follows are excerpts from our conversation about what changes when AI enters the development process.

You're watching the development world split into two. What's separating these teams?

Nothing has fundamentally changed for the majority of teams. They've adopted AI tools but still run the same agile ceremonies, following the same frameworks. It's the same sequential process.

AI-first companies look entirely different. Teams that truly embrace AI no longer think linearly. They pursue non-linear paths, engaging in multiple activities simultaneously. The cognitive load is intense. Give someone managing AI workstreams 2-3 hours, and they'll be mentally exhausted. That's proof that the work has undergone a fundamental change.

What separates organizations that see fundamental transformation from those that work faster?

Organizations that bolt AI tools onto old processes see minimal change. When we work with clients who challenge the status quo, we redesign the process to be AI-first. Teams can suddenly ideate, design, execute, and test simultaneously. But it's not just about speed. It's about doing that in three months while building solutions that fundamentally reshape how clients do business and engage with customers.

Everyone talks about productivity gains. What are the actual numbers?

On average, 30-35%, but it varies dramatically. Greenfield development can see a 70-80% uplift. QA and testing are where we've seen 80-90% productivity uplift, with almost all end-to-end processes fully automated. For mature systems, it's about 25-30%.

Assistants versus agents. Why does this distinction matter?

With an assistant, you're still in the driver's seat, working faster but linearly. With agents, you hand off tasks and orchestrate a digital workforce. You can launch multiple agents that work independently. You become an orchestrator. The challenge is enabling a proper, safe environment while staying in control.

Which skills have become more critical in this new landscape?

Three things are critical. First, communication and mental clarity. You need to know your craft deeply to instruct agents on what needs to be done and how.

Second, unlearning. The rules and playbooks need complete rewriting. Not everyone is willing to unlearn mental models that shaped them for decades.

Third, problem-solving. Computer programmers are expensive translators. We translate human language into computer language. Once agents do that translation, our job fundamentally changes to problem solvers. We need to spend most of the time solving problems while agents do the coding.

AI hallucinates. Code fails. How big a problem is this really?

That's a problem that AI didn't introduce. Engineers have been taking code from the internet, pasting it into production without review. The problem of code quality is solved by applying engineering practices: static code analysis, runtime checks, security checks, code review processes. It doesn't matter if a human or an AI generated the code. You still need those gates to warrant quality. 

What do clients get wrong with AI transformation?

That you can deploy an AI tool that will reshape how your team works overnight. No tool will transform a team overnight. Engineers are humans with mental models shaped over the years. If you want a fundamental transformation, you need to change humans. Help them acquire new skills and reshape their mental models. You start with people, with processes, then you do that with AI along the way. It doesn't happen magically. 

Has AI changed how developers work together?

There hasn't been a fundamental change yet. With AI, things will speed up. I see a world where we'll be talking not about blockers and issues but ideas, problems, and how we solve them. When AI takes care of information sharing, work will become more meaningful. Engineers will be closer to real users, concerned about making customers feel better rather than lines of code and test coverage. 

What does a developer's day look like five years from now?

Engineers will focus on deeply understanding customers' pain points. The roles of product owners and business analysts will either merge with engineering or disappear because engineers themselves will be closer to humans using the products. That reduction of the gap will only be possible if engineers have more time to engage with users and less time coding solutions, where agents pick up most of that work. Engineers will evolve to become orchestrators and human problem solvers. 

The transformation isn't about adopting the latest AI tool. It's about fundamentally rethinking what developers do, how they work, and what makes them valuable. The companies that understand this—the ones willing to challenge decades of ingrained practices—are the ones building the future. The rest are just working faster in the old world. 

Rodrigo's observations about the mindset shift and non-linear workflows reveal why most teams haven't fundamentally transformed. For a practitioner's view on what iteration and experimentation look like on the ground, read Ana Laura Robles's experience in When Everyone Can Code: What Makes Developers Valuable Now.

Return to the full exploration: Unlike Anything Before: How AI Has Transformed Software Development.