Achieving real growth in the age of AI: easy-to-use guardrails Junior Engineers live by at GetYourGuide

We empower our engineers to use AI confidently with the latest tools, regular hands-on sessions, and a strong culture of experimentation. When used correctly, AI can be a real force multiplier. But for junior engineers, if the tool removes too much of the decision-making, you can end up shipping more while building fewer skills that take you from junior to mid-level. 

Two of our talented engineers, Zuzana Moskvic and Nacef Racheh, collaborated on a simple but effective framework that they live by: a set of practical guardrails that keep AI working for them without short-circuiting their progress toward their professional goals. 

These guardrails are simple to use, highly accessible, and transferable across multiple fields. So, wherever you are on your AI journey, use these as a starting point to keep you firmly in the driver’s seat.

{{divider}}

Four guardrails for controlled and effective AI use

1. Keep the ‘productive discomfort’

Let’s start with an uncomfortable truth: struggle is part of learning. When AI removes all friction, you can lose the ‘productive discomfort’ that turns “I can copy this” into “I understand this.” 

It’s easy to lean on AI and build fake confidence, especially as a junior engineer, but this can lead to shipping something that works without understanding why. To use this guardrail, just keep up two simple habits:

A) Don’t ask AI too early.
Give yourself a focused first pass to understand the problem and identify where you’re stuck. Even taking 30 minutes matters. Vitally, it allows you to form your own mental model of the system before you borrow someone else’s.

B) Plan before you prompt.
Write down what you believe is true: assumptions, edge cases, and likely trade-offs. Then use AI to validate that thinking, rather than skipping it.

An example of a bad pattern

• “Here’s the error, fix it.”
• “Build X feature.”

An example of a better pattern

• Spend some focused time with the problem first, trying to narrow it down to the sum of its parts
• Write a rough plan (steps, assumptions, trade-offs)
• Use AI to review your plan or help you debug, without taking over

Good example prompts

• “Here’s my plan (steps + assumptions). What risks/unknowns and edge cases am I missing? What should I validate first? Any better options?”
• “Before suggesting a fix, ask me questions to help me debug.”

The point isn’t to use AI less. It’s to use it later in the loop, after you’ve done the thinking required to build a strong foundation of understanding about the problem.‍

2. Use AI as a thinking partner, not a code printer

AI is at its best when it sharpens your thinking. The shift is subtle but powerful: moving from “Write this for me” to “Help me understand this better.” 

When you use AI like a senior engineer who reviews your approach, it improves the thing that matters most over time: your judgment.

Try asking AI the kinds of questions an experienced reviewer would ask:

• What are the risks and trade-offs?
• What could break in production?
• What will negatively affect maintainability?
• What are the likely failure modes?
• Where are the performance pitfalls?

This keeps you in the driver’s seat. The goal isn’t bigger commits, it’s better decision-making.

Example prompt

“I’m implementing feature X. These are the edge cases I’ve considered: [list]. Can you review this like a senior engineer and tell me what I’m missing, including performance concerns and maintainability issues?”

A useful rule of thumb from Zuzana and Nacef: if your prompt doesn’t include your reasoning, you’re probably skipping the learning. The most impactful prompts should lay out your thought process and ask the model to challenge it.

3. Move fast without outsourcing understanding

‘Vibe coding’ is powerful because it lowers the barrier to building. Sometimes it even enables non-engineers to ship real improvements, and that should be encouraged. We’ve seen moments where someone outside engineering goes from “can we tweak this?” to “I already opened the PR.”

However, as developers, we can’t prioritize speed over our understanding. Make sure you’re maintaining deep codebase knowledge so that you can:

• ship safely at scale
• spot edge cases before shipping
• make the right trade-offs
• maintain systems over time

In a nutshell, let AI accelerate your iteration loop without outsourcing your judgment. If you can’t explain what the code is doing, you’re not done yet.

One practical habit we recommend is to run a quick “ownership check” after using AI. Ask yourself:

• “Can I explain the approach to someone else?”
• “Can I predict what will break if requirements change?”
• “Can I reason about the impact on performance, observability, or maintainability?”

If the answer is “no,” the next step isn’t to ship faster, it’s to slow down and close the understanding gap before moving on. That’s how you build deeper understanding and lasting confidence.

4. Build on the skills AI can’t replace

As AI increases raw output, differentiation shifts toward skills that compound across teams. This guardrail combines two themes: fundamentals and soft skills.

Fundamentals still matter
Strong foundations help you evaluate code quality, architecture choices, and trade-offs in your business context. They also let you guide AI tools more effectively. When you understand the basics deeply, you can spot when generated solutions are wrong, risky, or over-engineered.

Soft skills scale your impact
Communication, collaboration, and mentorship matter more (not less) when tools increase your velocity. AI can help you draft, summarize, and brainstorm, but it can’t replace your core human skills: good judgment and clear communication.

This guardrail prevents common failure modes like:

• A technically successful experiment never launches because expectations weren’t aligned with management (scope, “definition of done,” non-functional requirements).
• You solve the technical problem but don’t involve stakeholders, leading to more work later down the line.

Practical ways to apply it

• Work with a mentor/manager to identify which fundamentals to prioritize, and create a simple learning plan.
• Use team feedback to filter which tools are worth adopting (so you don’t drown in the AI tool-of-the-week cycle).
• Run lightweight AI-sharing sessions, so one person’s exploration benefits the whole team.

Wrapping up

At GetYourGuide, our view is simple: AI should amplify engineering judgment, not replace it. Used well, it helps engineers at every level learn and move faster, while staying accountable for understanding the problem, making the right trade-offs, and owning the outcome in production.

These guardrails are designed to keep you in control, so that AI can sharpen your thinking and accelerate progress without skipping the fundamentals that foster real growth. Get that balance right, and you’ll be able to ship more whilst becoming a stronger engineer along the way.

Interested in using the latest AI tools with some of the top talent in the travel industry? Check out our Junior Engineer positions here and all our open roles here to start changing the way millions experience the world.