Your Team Is Using AI Wrong
AI coding tools speed up routine work, but they can also erode your team's skills and make delivery less stable, which is why quality gates and code review matter more now.
You gave your team Copilot licenses six months ago. PRs are flying and lines of code are up, so everyone feels productive.
Then you check the incident log, and bugs are up too. Code review is taking longer, and your best senior engineer just mass-reverted a junior's AI-generated PR at 2am because nobody understood what it did.
AI coding tools are making your team faster at creating problems.
What the productivity studies show
To be clear, AI does speed things up in certain cases. Documentation tasks take half the time, tests get written faster, and boilerplate that used to eat your afternoon now takes minutes.
Here's where it gets weird.
A study by METR tracked experienced developers doing real tasks in codebases they'd maintained for years. The developers using AI tools like Cursor took 19% longer to complete their work.
The kicker is that those same developers believed they were 20% faster.
That's a 39-point perception gap, with engineers who thought they were crushing it while they were falling behind. You can't trust vibes or self-reports here, because the tool that feels like a superpower might be slowing you down on the work that actually matters.
AI is a power tool for grunt work, and it turns into a liability when the work needs judgment.
The skill atrophy problem is already here
Addy Osmani, engineering lead at Google Chrome, documented testimonials from developers experiencing skill decay. One engineer after 12 years said AI made him "worse at his own craft." First he stopped reading documentation, then his debugging skills waned, and then he lost deep comprehension of his own systems. In his words, "I've become a human clipboard."
Osmani surveyed CTOs and found that 16 of 18 reported production disasters directly caused by AI-generated code that nobody understood.
Skills fade when you stop using them, and AI speeds up the forgetting.
For juniors, the skills may never form
Skill atrophy assumes you had skills to begin with. For junior developers, the problem is worse because they may never develop foundational abilities in the first place.
When a junior uses AI to write code they don't understand, they skip the struggle that builds intuition. They never learn to read stack traces or internalize why certain patterns exist, and they ship code faster while building a house on sand.
The research calls this the "trust without verification" anti-pattern. Juniors accept AI output without the experience to know when it's wrong. They can't debug what they didn't write and they can't extend what they don't understand. After all, why argue with knowledge that is the average of the knowledge of the internet when you only recently learned what a unit test was?
Be careful about who you count as a "junior" here, because the word also covers super seniors 20 years into their career who have to be productive in an unfamiliar technology.
What do we do?
What can you do in the face of powerlessness? It's not like a ban on these tools would be effective, or even possible. I'm not going to pretend I know of a perfect solution, so I'll share what I know to work as of writing this.
Your quality gates matter more than ever
A review sequence
- Produce a working changeMake the proposed behavior concrete.
- Check it against the rulesUse representative cases and known exceptions.
- Review it with the people responsibleAccept it, correct it, or name what is still uncertain.
Google's 2024 DORA report found that for every 25% increase in AI adoption, teams saw a 7.2% drop in delivery stability.
The DORA report's core finding is that AI amplifies what's already there in a team. Teams with strong testing and review practices use AI to move faster while staying stable. Teams with existing dysfunction find that AI just intensifies their problems.
This is why quality gates, the automated checks that catch issues before humans even look at the PR, are a requirement now.
What I know works:
Set CI gates for the behavior that matters to the business. Choose coverage targets based on the risk and complexity of the code, and review the gaps, because a passing percentage does not tell you whether the tests check the right outcomes.
Run security scans that fail the build on critical findings, and run SAST on every PR.
Enforce small PR requirements. AI encourages mega-PRs because generating code is cheap, but big PRs hide bugs. Fight the urge to batch everything together.
Use type checking everywhere. If your language supports it, enforce it in CI.
Treat AI output like code from an enthusiastic but inexperienced contributor and review everything.
Support a culture of learning with book clubs, valuable mentorships, and pair programming.
Keeping your engineering fundamentals
The teams winning with AI are the ones who kept their engineering fundamentals while selectively speeding up the boring parts.
Google now generates over 30% of their code with AI, and their internal guidance emphasizes "maintaining rigor" in code review, security analysis, and long-term maintenance. The humans review everything before it ships.
Original article image
