AI Productivity Gains Face the Wall of Enterprise Legacy Software
Analyst Benedict Evans argues that AI's ability to empower individuals won't automatically fix the fragmented structural chaos of large corporations.
The belief that generative AI will fundamentally democratize software creation or dissolve traditional applications is a misunderstanding of corporate reality. While AI provides powerful new capabilities for the individual, it does not inherently rewrite the structural ways large organizations operate.
Technology analyst Benedict Evans argues that the notion of AI turning every employee into a tool-builder ignores how people actually think and the true origins of software. According to Evans, the primary obstacle to systemic transformation is not a lack of tools, but the entrenched complexity of the enterprise. Large American companies typically rely on hundreds or even thousands of disparate software pieces to function. This fragmented ecosystem includes "big iron" horizontal systems of record—such as SAP and Workday—alongside various vertical SaaS products and a chaotic web of internal scripts and spreadsheets.
The Legacy Bottleneck
This structural fragmentation creates a significant gap between individual productivity and organizational efficiency. In the current landscape, an employee might use a large language model to analyze a log file or automate a personal task, but that efficiency remains isolated. The underlying corporate infrastructure remains a collection of silos that do not communicate seamlessly. For most large firms, software is not a fluid set of tools built on the fly, but a rigid architecture of legacy systems that dictate the flow of information and authority.
Why Systemic Change Matters
If AI remains merely a productivity booster for the individual, the promised transformation of the modern business will be limited. The core tension lies in whether AI can move beyond simple tool-use to address the underlying fragmentation of enterprise software. Without a way to bridge these disparate systems, AI may simply allow employees to perform fragmented tasks faster without actually streamlining the organizational workflow.
The Path to Automation
What remains to be seen is whether AI can evolve into a system capable of navigating and unifying this complex corporate infrastructure. The industry is currently shifting toward systems that can reason and take action autonomously, but the ability to operate across a thousand different legacy scripts and SAP modules is a far higher hurdle than writing a snippet of code. Until AI can interface with the "big iron" of the enterprise, the structural inertia of the large corporation will likely outweigh the productivity gains of the individual user.