AI coding tools are boosting individual developers, but engineering output remains stalled

Artificial intelligence has demonstrably accelerated what individual developers can accomplish in isolation. Yet the infrastructure and processes surrounding those developers are negating much of that efficiency, leaving overall engineering output virtually unchanged despite massive investment increases.

This disconnect emerged from DX's latest State of AI Impact in Engineering report, which benchmarked engineering teams across velocity, effectiveness, quality, and organizational impact. The finding is particularly alarming: AI costs have climbed 28-fold for companies with more than 99 engineers, yet velocity metrics have stagnated or declined.

The Tech Debt Trap#

Deputy CTO Justin Reock of DX describes the situation as concerning. When expenditure rises exponentially but output doesn't follow suit, the return on investment becomes deeply questionable. Rather than freeing engineers to pursue innovative feature work, the report indicates that AI-generated time savings are being consumed by tech debt, backlog management, and maintenance rather than new business-critical functionality. The innovation ratio—the proportion of engineering effort allocated to features versus operational overhead—remains flat even as AI adoption widens.

Confidence Crisis in Code Quality#

An even more troubling trend involves a breakdown in the relationship between code maintainability and release confidence. Historically, these two metrics moved together: engineers who could easily understand and modify code felt more secure deploying it to production.

That relationship has inverted. AI tools are improving developers' ability to comprehend and alter existing code, yet engineers report decreased confidence in the safety of what they release. As Reock notes, the paradox is striking: developers can read, maintain, and change code more effortlessly, yet trust their production deployments less. This erosion of confidence carries measurable cost—each point of decline in DX's Developer Experience Index translates to losing ten hours annually per engineer. The industry experienced its first-ever multi-point drop in this metric.

Organizational Scale Matters#

Smaller organizations extract significantly more value from AI investments than their larger counterparts, and legacy enterprises struggle most acutely. According to Martin Davidson, CTO of micro-consultancy a2bic.ai, two technologists with combined decades of experience can orchestrate AI agent teams that replicate the output of 100 junior-to-mid-level engineers. Small teams sidestep the exponential communication overhead that scales with organization size—a principle articulated decades ago in Fred Brooks' , where coordination channels multiply as .