Technology

How to Build the Optimal AI-First Engineering Team

This is not about which AI tool to pick. It is about how you build a team that can consistently deliver with AI, today and six months from now when everything has changed again.

I recently hosted a webinar with our CTO Francisco Velasco on this exact topic. We have 90+ AI-first developers working across 35+ Dutch tech companies. What follows is what we actually know from practice, not theory.

One disclaimer Francisco and I both agree on: what we share today may be old news in three or four weeks. The pace of change in AI tooling is that fast. We saw it that same morning with updates from Copilot and Codex. That is precisely why structure matters more than any specific tool.

Hero image of the article: How to Build the Optimal AI-First Engineering Team

AUTHOR

Photo of Maarten Moen

Maarten Moen

Managing Director

May 19, 2026 - 6 minutes read

Start at hiring

AI-first starts before someone's first day. We have fundamentally changed our technical interview process. AI skills are now close to mandatory for every new hire.

I introduced AI coding challenges, where the developer has to solve a very complex challenge using only agentic code. Two years ago, this would have been almost impossible. — Francisco Velasco, CTO at 25Friday

The point of the challenge is not just to evaluate the outcome. It is to observe how someone directs an agent, reviews a plan, and course-corrects when things go wrong. That skill is what separates an AI-first developer from someone who occasionally uses a code suggestion tool.

Build a structural training program

You cannot become AI-first through a one-off workshop. Our approach runs on three levels.

Weekly knowledge sharing. Every developer joins a 30-60 minute session where teams share what they have learned, what works, and what does not. As things move fast, this cross-team sharing is one of the most effective ways to keep everyone up to speed.

Monthly intensive training. A 3-4 hour deep dive into specific topics. Not a lecture but a working session where we go hands-on with real problems.

Mob programming sessions. Francisco joins each team directly to tackle complex problems together using AI. This is where real knowledge transfer happens, not in a slide deck.

Give every developer an exploration budget

Every developer at 25Friday receives a monthly budget to explore tools outside of our enterprise stack. The reason is simple: no central team can keep up with everything that is changing. By letting developers experiment and share what they find, we learn as an organisation faster than any top-down decision could achieve.

They can spend it in training, also explore other tools in the market, just to try out. And not be restricted to the tool that we provide on an enterprise level. — Francisco Velasco

Appoint one person for AI innovation

Most developers are too focused on their daily work to follow every AI development in the market. That is not a criticism, it is reality. Francisco's solution: one dedicated person whose job it is to stay ahead, identify best practices, and share that knowledge across all teams.

AI

AI-Agentic-Workflow

AI-First Engineering Team