Companies are buying AI tools. That doesn't mean they know what to do with them.

The Gap Between Buying AI and Using It Well

A growing body of research is converging on an uncomfortable truth for enterprises: purchasing AI tools is the easy part. Two new reports cited in recent coverage suggest that companies pouring budget into AI adoption are largely failing to translate those purchases into measurable returns, because the harder work — strategic investment and organizational change — is being skipped or underfunded.

This finding lands squarely in the middle of one of the most active corners of enterprise AI spending: developer tools. AI coding assistants, code review platforms, and editors built around large language models have seen explosive adoption over the past two years, driven by promises of faster shipping cycles and leaner engineering teams.

Why This Matters for Coding Tools Specifically

Tools like the Cursor AI editor have become emblematic of this wave — an entire IDE redesigned around AI-assisted coding, attracting significant enterprise interest and valuation. AI code review tools have followed a similar trajectory, marketed as ways to catch bugs, enforce standards, and reduce reviewer fatigue automatically.

But the reports' core argument — that tool purchases without organizational change rarely produce returns — applies with particular force here. Handing engineers a license to Cursor or an AI code review bot does not, by itself, change how a team plans sprints, structures code review, measures productivity, or trains junior developers to work alongside AI output. Without deliberate changes to workflows — for instance, redefining what code review means when a bot pre-screens pull requests, or rethinking onboarding when new hires lean heavily on AI-generated code — organizations risk simply layering a new tool on top of old processes and wondering why velocity hasn't meaningfully improved.

The Organizational Change Problem

This is a familiar pattern from past waves of enterprise technology adoption, from cloud migration to earlier automation tools: the software is rarely the bottleneck. The bottleneck is management practice — whether leaders retrain teams, redesign incentives, and rebuild metrics around the new capability rather than bolting it onto legacy structures. For coding assistants, that could mean rethinking how code quality is measured when much of it is machine-suggested, or how technical debt accumulates when AI-generated code passes review faster but isn't always well understood by the humans who approved it.

What to Watch

Expect increasing scrutiny of ROI claims from AI coding vendors as CFOs demand evidence beyond adoption metrics like seat counts or query volume. Companies that pair tools like Cursor or AI review systems with genuine process redesign — changed review standards, updated engineering KPIs, dedicated training — are likely to be the ones reporting real productivity gains, while those that merely swap tools without adjusting practices may find themselves repeating the reports' central warning: buying AI isn't the same as knowing what to do with it.