In a recent post, Bryan Ross — Field CTO at GitLab — warns that AI doesn’t just produce more code; it also generates more bugs, broken builds, reviews and security vulnerabilities — posing a new quality bottleneck that every platform team must confront. He’s right. But if we’re familiar with the Theory of Constraints, The Goal, and The Phoenix Project, we know this is exactly how progress unfolds — each bottleneck addressed simply reveals the next. And that’s okay. In fact, that’s the nature of continuous improvement.
Seeing Bottlenecks as Signals, Not Failures
At its core, the Theory of Constraints (TOC) teaches that any system is limited — not by many things — but by one critical constraint at a time. The five focusing steps — identify, exploit, subordinate, elevate and then repeat — reflect a relentless cycle of improvement.
That’s the same lesson The Goal drives home: When you fix one bottleneck, another inevitably emerges. It’s not doom—it’s progress.
And The Phoenix Project? It’s TOC, DevOps and transformational storytelling rolled into one — platform engineering in novel form. Its core is that “the work is never done,” especially at scale.
Platform Engineering — The Perfect TOC Playground
Platform engineers — and DevOps teams — have put these ideas into practice: building golden paths, CI/CD pipelines and automated quality gates to let developers move quickly, securely and consistently. Those initiatives didn’t eliminate bottlenecks; they exposed new ones:
- Automating build pipelines exposed bottlenecks in testing infrastructure or security scanning.
- Scaling golden paths revealed challenges in modular reuse or enforceable compliance templates.
According to Ross, without robust internal developer platforms and quality gates, organizations will not just struggle — they’ll fail when AI accelerates code production. His point is well taken — and echoes TOC’s central message: Address the constraint, and the system shifts.
AI: The New Catalyst for Constraint Evolution
AI supercharges coding — but it doesn’t change the system’s rules. It magnifies both potential and pressure.
For example:
- AI-assisted test generation can elevate developer velocity — and simultaneously overwhelm test review and integration systems.
- AI-driven code suggestions help developers create secure patterns faster — but surface vulnerabilities in abstract reasoning that we haven’t fully automated yet.
- AI-generated infrastructure changes accelerate deployments — but can expose gaps in governance and safe rollout pipelines.
Each of these becomes a new bottleneck — and that’s not frustration material — it’s pure TOC process: Identify, fix and repeat.
“The Job is Never Done” — and That’s the Point
If platform engineers expect perfection, they are headed for disappointment. The point is not to eliminate constraints forever — it’s to keep improving. As The Goal puts it, ongoing improvement (POOGI) is the game.
Every layer of automation, every new AI integration reveals another opportunity:
- Could the next bottleneck be AI model drift in quality?
- Or maybe cost and latency constraints of AI inference at scale?
- Perhaps the challenge lies in AI alignment and oversight to maintain trust?
These aren’t failure signals; they’re stepping stones toward a smarter, more responsive platform.
Conclusion
Bryan Ross is spot-on about AI introducing new quality bottlenecks — and platform teams must prepare. But if we lean into the Theory of Constraints, The Goal, and The Phoenix Project, we understand that tackling one bottleneck only means the next one awaits. That’s the dynamic reality — and it’s empowering. In the evolving world of platform engineering, the job is never done — and that’s the reason we keep building.
— Alan Shimel (Techstrong)