Andrew Ng recently gave a great talk for Y Combinator about how you can build faster with AI. There's lots of great insight in this talk, but what most resonated for me was how AI is changing how software is built; specifically, how AI is upending coding tenants that have existed essentially forever. While there are lots of aspects to this disruption, in this blog I will focus on a few that I find interesting and seem somewhat less discussed. For engineering leaders, the challenge isn't simply adopting new tools—it's rethinking how software teams operate and create value.

Rethinking Architectural Risk

The traditional engineering maxim "get it right the first time" emerged from economic reality: architectural changes were expensive and risky. AI is dismantling this assumption by making systematic refactoring economically viable, causing the relationship between speed and quality to change completely.

Consider the implications for your team's decision-making process. If AI can assist in comprehensive code restructuring, architectural choices become more reversible. This doesn't mean abandoning architectural discipline, it means developing new frameworks for evaluating risk versus velocity trade-offs.

Engineering leaders must help their teams navigate this new territory. The question shifts from "How do we avoid architectural debt?" to "How much architectural debt optimizes our velocity?"

This requires new metrics, new decision-making processes, and new ways of thinking about technical investment. The most successful teams will be the ones that develop sophisticated approaches to managing this new flexibility.

The Product Management Crisis

The most dramatic organizational shift involves product management capacity. Historical ratios of one product manager per six developers assumed development was the bottleneck, but now AI productivity gains are turning this assumption upside down.

Recent research shows AI now generates 41% of all code. Developers complete tasks 55% faster with AI assistance. But here's the catch: Forrester's data reveals developers spend only 24% of their time writing code.

This creates a fundamental mismatch. AI accelerates the smallest portion of developer work while leaving the largest portions unchanged, and the result is a product management bottleneck that many organizations haven't anticipated.

Teams that previously operated with sufficient PM coverage now find themselves constrained by product direction and user research capacity. This challenge compounds as AI-accelerated development cycles create pressure for more frequent product decisions.

Engineering leaders must work closely with product leadership to address this imbalance. This might involve hiring more product managers, developing stronger product intuition within engineering teams, or creating new hybrid roles that bridge product and engineering functions.

Knowledge Management Crisis

AI's impact on documentation and knowledge management represents a subtle but serious challenge. When AI generates code, human understanding of details naturally decreases. The traditional assumption that code authors have a full understanding of their work no longer really holds.

This creates risks for long-term maintainability and team knowledge transfer. Teams must develop new practices for maintaining context and decision rationale. The documentation focus shifts from "how" to "why"—AI excels at generating technical documentation but cannot capture business context or architectural reasoning.

Engineering leaders must establish new knowledge management practices that preserve human insights while using AI capabilities. This includes creating processes for capturing architectural decisions, maintaining business context, and making sure critical knowledge doesn't become trapped in AI-generated code that humans struggle to understand.

Teams reporting considerable productivity gains from AI also report better code quality, largely due to AI-generated documentation. But this improvement comes with the hidden cost of reduced human understanding. It makes sense that the most successful teams will be those that find ways to maintain both productivity gains and human comprehension.

Economic Strategy and Competitive Advantage

The economic implications of AI-driven development create both opportunities and competitive pressures. McKinsey research identifies $4.4 trillion in potential productivity growth from AI, with software development representing a significant portion of this value.

Small teams can now achieve enormous impact through AI multiplication effects. This changes startup economics when five developers can accomplish what previously required twenty. Enterprise development costs shift toward compute and tooling rather than pure labor costs, creating new budget priorities and resource allocation decisions.

The time-to-market acceleration becomes a competitive weapon. Companies that successfully integrate AI development practices can iterate faster, respond to market changes more quickly, and experiment with lower opportunity costs. This creates winner-take-all dynamics in many software markets.

Engineering leaders must develop strategies that capture these economic advantages while still managing the associated risks. This includes building AI capabilities that compound over time, creating competitive moats through superior AI integration, and developing organizational learning that keeps up with AI adoption.

The Hard Truth About Results

Despite the promise, challenges remain significant. And not all organizations, and thus not all studies, report positive results. For example, the 2024 DORA report indicates that speed and stability have actually decreased due to AI. Many organizations are struggling with the transition.

Recent research from METR reveals that experienced developers using AI tools take 19% longer than without AI, yet developers estimated they were sped up by 20% on average. This perception gap means engineering leaders must rely on measurement rather than intuition when evaluating AI's impact. And even this measurement is tricky—what defines quality? How do you measure it?

Success requires deliberate practice and new organizational capabilities. Teams must develop quality gates specific to AI-generated code, create review processes that focus on business logic rather than syntax, and build systems that measure productivity changes accurately.

The most successful engineering leaders will be those who approach AI adoption with not only enthusiasm, but also rigor. They will experiment extensively, measure results carefully, and adapt their approaches based on evidence rather than assumptions.

Leading the Change

The shift in software engineering through AI isn't a future possibility—it's happening now. By 2027, 50% of software engineering organizations will use software engineering intelligence platforms to measure and increase developer productivity, representing a massive increase from 5% in 2024.

Engineering leaders face immediate decisions about team structure, skill development, and technology adoption. The organizations that thrive will be those that can orchestrate human creativity with AI capability rather than simply deploying tools.

This requires investment in product management capabilities, development of new quality assurance practices, and rethinking career development programs. Teams must experiment with AI tools while maintaining engineering discipline and measuring results carefully.

The window for strategic advantage remains open, but it's closing rapidly. The question isn't whether AI will reshape software engineering, but whether your organization will lead or follow this change. The leaders who act decisively now will shape the future of software development for their organizations and their industry, and gain competitive advantage as a result.

Your team can lead the AI transition without compromising engineering discipline by trying Flux today.