AI is changing everything about how we build software. It's letting us move faster, ship more, and compete in ways that weren't possible even a few years ago. On the development side alone, we're now surrounded by tools: GitLab Duo, GitHub Copilot, Kiro, Amazon Q Developer, and Rovo for backlog and story refinement. There's a tool infused into nearly every layer of the process. We've even coined a new term for it: the AI-DLC, a spinoff of the traditional software development life cycle.

‍

All of this is genuinely exciting. However, as a software engineering leader, I'm beginning to see some tradeoffs that aren't being discussed honestly enough.

‍

The Moment I Got Concerned

I review many pull requests every week. Some contain hundreds of lines of AI-generated or AI-assisted code. And if I'm being honest, I've started to wonder whether many developers are reading every line they're submitting anymore. Too often, the code compiles, passes the tests, looks reasonable, and gets approved.

‍

A while ago, I sat in a code review where someone submitted a pull request that solved the problem in front of them but had no idea what the code was actually doing to the rest of the system. When I asked why it worked, the honest answer was, "I'm not sure. It just did."

That's the moment I get concerned. Not because the code was wrong, but because nobody in the room could confidently explain it if it failed at 2 a.m. in production.

‍

That experience reinforced something I'd already been noticing: AI is making it easier than ever to produce code. But it's also making it easier to submit code we don't fully understand.

‍

The Tradeoffs Nobody's Talking About

The more I sit with this, the more I see the same handful of risks showing up again and again across teams. Mastering a prompt or accepting a suggestion isn't the same as understanding what you shipped. These are the tradeoffs I think every engineer and engineering leader needs to reckon with:

1. The Economics Haven't Caught Up Yet

Right now, many of these tools are inexpensive or heavily subsidized to drive adoption. Eventually, as pricing models mature and organizations scrutinize AI spending more closely, they'll start asking harder questions about return on investment.

Tip: Developers who continue sharpening their ability to design systems, debug complex issues, and reason through software, with or without AI, will be far more resilient if pricing models change or access becomes more limited.

2. The Biggest Long-Term Risk: Skill Atrophy

My biggest concern isn't that AI writes code. It's that developers may slowly lose the ability to think through problems on their own. If AI writes the first draft of everything, what happens to a developer's ability to reason about systems, evaluate tradeoffs, and understand why a solution works? Are we raising engineers who can write exceptional prompts but struggle to explain the code they're shipping?

Tip: AI should accelerate your learning journey, not replace it. The goal isn't to become dependent on AI—it's to become more capable because of it.

3. Speed Without Understanding

One of the biggest risks I see is the illusion of understanding. Developers are shipping code they didn't fully write and don't fully understand. That's manageable until something breaks in production and no one can explain why the original implementation worked.

Tip: Passing tests and successful builds aren't enough. Make sure you understand how your code affects the larger system, how it performs under edge cases, and what happens when assumptions change.

4. Job Transformation, Not Just Job Displacement

The nature of software engineering is changing. We're spending less time writing boilerplate and more time reviewing, refining, prompting, and making architectural decisions.

Tip: Don't aim to be the best prompt engineer. Aim to be someone who keeps investing in their craft, understands the systems AI is helping build, and knows when the generated solution isn't the right one.

5. Speed Without Wisdom

Agentic tools can move incredibly fast and touch dozens of systems at once. That's powerful. But production software isn't just about speed. It's about judgment.

Tip: When you're building systems that handle financial transactions, healthcare records, customer data, or critical infrastructure, let engineering judgment matter just as much as development velocity.

6. Who's Accountable?

As AI becomes more autonomous, accountability becomes less obvious. When an agent writes code that causes a production incident, or makes a design decision with unintended consequences, who owns that outcome?

Tip: Remember that responsibility still belongs to the engineer and the organization. AI may assist with the work, but it shouldn't replace ownership.

7. The Human Cost Nobody's Pricing In

Beyond the technology itself, there's something more personal happening. Many engineers are dealing with imposter syndrome, anxiety, and burnout as they try to keep pace with a rate of change that feels relentless. At the same time, many are quietly wondering whether they're becoming less capable as they rely on AI more each day.

‍

Tip: These are conversations worth having openly with your team, not pressures to carry alone.

Bringing It All Together

To be clear, I love AI, and I use these tools constantly. But I've realized I have to be intentional about ensuring they don't replace the parts of software engineering that make me better.

‍

For me, that means reviewing AI-generated code line by line before I trust it. I still pursue certifications and courses even when I don't need them for a project. I intentionally take on hackathons, spikes, proof of concepts, and side projects to keep solving problems myself. When AI gives me an answer, I try to understand why it's correct before I accept it.

‍

It's not about avoiding AI. It's about making sure AI amplifies my expertise instead of quietly replacing it. The tools aren't the problem. Coasting on them is.

‍

So here's what I'd ask you: what's one skill you're intentionally making sure you don't lose as AI gets better? I'd genuinely love to hear how other engineers are thinking about it.

‍

Before I go, one quick note. Over the next few weeks, I'll be sharing more content like this on my new page, Tech with Errol. My goal is to create a place where we can have honest conversations about AI, software engineering, cloud, architecture, leadership, and what it means to grow as technologists in an industry that's changing faster than ever.

‍

I'll have a few announcements coming soon, and I'm excited to share what's next. If these are the kinds of conversations you enjoy, I'd love for you to follow along for more.

‍

We're just getting started.

‍

‍