The Software Abundance Problem

For decades, the expense of software development has quietly shaped how enterprises operate. Limited engineering capacity dictated prioritization, and business units competed for budget to build new applications. But with AI-powered tools rapidly lowering the barrier to entry, this paradigm is shifting.

Perplexity’s recent experience—where two engineers built a production-ready key-value store (CobbleDB) in just eight weeks using AI coding agents—highlights this change. CobbleDB now handles part of Perplexity’s search traffic and delivers 80% lower latency than their previous DynamoDB setup, with projected cost savings of at least 20%.

This isn’t an isolated incident. Payhawk replaced a $70,000 commercial performance-management product with self-built software using AI assistance. Even Starbucks is reportedly exploring building AI-developed replacements for enterprise systems from Microsoft and IBM, potentially saving hundreds of millions annually.

The implications are profound: as creating software becomes cheaper, enterprises will naturally build more themselves rather than buying off-the-shelf solutions.

The New Challenge: Managing Abundance

While this represents a tremendous opportunity—allowing organizations to create tailored solutions at scale—it also introduces new risks. When development capacity is scarce, we focus on what can’t be built; with abundance, we need to manage what will be built.

Imagine a scenario where every department independently generates tools and scripts without central oversight. The official application portfolio might still list 500 applications while thousands of unsanctioned solutions operate in the shadows—creating shadow IT on an unprecedented scale.

From a governance perspective, this presents significant challenges: most control frameworks assume software enters through established channels with built-in review processes. When tools are generated ad hoc and bypass these checkpoints, critical controls like access provisioning and security audits fall through the cracks.

The solution isn’t to restrict development but rather to evolve our approach—empowering employees while maintaining appropriate oversight. Organizations need new governance models that accommodate this shift towards distributed software creation.