“Move fast and break things” seems to be the motto of the tech industry. However, when companies go all-in on the AI arms race and push too fast, essential infrastructure breaks, security holes and bugs appear in codebases, and ultimately no one wins.
While senior executives may assume AI can generate code rapidly and thus developer roles can be minimized, that premise stems from a flawed understanding of software engineering. Writing code is only a small fraction of what developers actually do. Engineers translate requirements into steps and systems, architect solutions to complex problems, and, most importantly, ensure everything running in production is stable and reliable. AI can assist with certain tasks, but it cannot replace the broader scope of a developer’s responsibilities.
As AI generates code faster than ever, companies should invest more resources in developers to help them review, test, and ship quality software securely, rather than scaling back engineering teams to cut costs. Organizations must also invest in future generations of developers, equipping them with the skills needed to thrive in an AI-augmented environment.
What Does the Massive Microsoft Outage Tell Us?#
Following the global Microsoft outage, the company provided an official response detailing the cause. Executive VP Charles Lamanna explained that a defect dating back to July in a Microsoft 365 networking gateway fix caused servers to become overwhelmed and unresponsive under specific traffic conditions. In other words, a bug introduced months earlier went unnoticed until a sudden spike in traffic triggered the fault, creating a domino effect that resulted in widespread service disruption.
While there is no direct evidence disclosed linking this specific incident to AI coding agents, it highlights a critical reality: severe downtime can occur when code goes unexamined for long periods. Careful oversight and code review remain vital, regardless of whether code is written by AI or a human developer.
This is far from an isolated case. A 2026 survey of 300 engineering leaders by Undo revealed that 81% of organizations experienced a production incident or service outage linked to AI coding tools within a six-month window. Furthermore, in July 2025, an AI coding agent from Replit ignored explicit instructions during a code freeze and deleted a live production database for the startup SaaStr. The agent wiped records for over 1,200 executives and 1,190 companies, then generated 4,000 fake profiles and fabricated unit test results to cover up the mistake.
What Are the Hidden Costs of AI for Companies?#
As AI providers mature and go public, token pricing is shifting from heavily subsidized rates to standard market rates. Flat-rate cushions are ending, and enterprises transitioning to usage-based pricing have seen their bills surge unexpectedly. Consequently, organizations are re-evaluating the true cost efficiency of AI tools. Business leaders are realizing that unchecked reliance on AI agents can quickly become more expensive than hiring human developers.
Beyond direct API costs, failing to safeguard production pipelines carries severe financial risks. A study by Wakefield Research and Sauce Labs found that 80% of executives traced a production failure back to AI-generated code over the past year. Among those organizations, 65% reported that their most severe quality incident cost over $500,000.
AI Adoption Should Be Driven Bottom-Up, Not Top-Down#
Through my work as a developer advocate, I regularly speak with engineers about their experiences with AI tools, which vary widely. Some report that management mandates AI usage to the point where developers feel forced to expend token quotas on trivial tasks just to meet usage metrics. This top-down pressure hurts both operational budgets and team morale. In extreme cases, companies have even phased out proven tools that developers relied on just to force AI adoption.
Technological adoption should be driven by the practitioners who understand it best. Companies ought to collaborate with their engineering teams, granting them the space to experiment and discover meaningful ways to improve software quality and workflow efficiency.
Developers Will Not Be Replaced, but the Role Will Evolve#
Software development remains a viable and rewarding career, but the daily focus of the job is evolving. Just as writers transitioned from typewriters to digital word processors, developers are adopting new tools while the core mission remains the same. Required skill sets will shift: as routine code generation becomes automated, engineers must excel at code review, system design, and security verification before signing off on production deployments. Investing time in cybersecurity expertise and domain-specific best practices will be essential for ensuring high-quality software.
For organizational decision-makers, the smartest path forward is to invest in your engineering team. Building a skilled, proactive workforce during stable periods ensures your team can handle issues when they arise. Furthermore, investing in junior talent maintains a healthy talent pipeline for the future, while updating infrastructure proactively prevents costly outages down the line.
If the tech industry continues to view AI strictly as a tool to replace human engineers rather than empower them, software quality and infrastructure reliability will suffer. Result? Costing companies far more in the long run.
Cover Photo by Igor Omilaev on Unsplash