For anyone who has spent a career in enterprise technology, the current wave of AI-generated software development carries an unmistakable sense of déjà vu. The industry calls it "vibe coding"—a term coined by AI researcher Andrej Karpathy in early 2025 to describe the practice of building software by describing what you want in plain language and letting artificial intelligence write the code. It became Collins Dictionary's Word of the Year shortly after, and by 2026, it is reshaping how organizations think about the software they run. But for those of us who remember the enterprise IT landscape of the 1990s and early 2000s, many of the emerging trends feel less like revolution and more like a familiar road reappearing through the fog. The tools are new, but the problems that come with owning your own software, like hosting and maintenance, are thirty years old.
The enterprise software industry has always oscillated between two philosophies: build it yourself, or buy it off the shelf. In the 1990s, large organizations routinely built custom systems tailored precisely to their operations. Finance departments ran homegrown ledger applications. Sales teams used proprietary customer databases. Warehouse operations depended on bespoke inventory trackers coded by in-house teams. It was expensive, it was slow, and it was yours.
Then came the great consolidation. Platforms like SAP, Oracle, PeopleSoft, and later Salesforce and NetSuite promised a better bargain: one system, professionally maintained, continuously improved, available to everyone. Companies abandoned their custom tools in droves, opting for the efficiency and reliability of packaged software. The buy side of the equation dominated for more than two decades.
Now the pendulum is swinging back. A 2026 report from Retool found that 35 percent of enterprise teams have already replaced at least one SaaS product with a custom-built alternative, and 78 percent expect to build more custom internal tools this year. One company, Blinkist, publicly shared that it replaced roughly $60,000 per year in SaaS spending by building lightweight internal tools with AI coding platforms in a matter of days. This is not a fringe experiment. Gartner estimates that by 2028, approximately 40 percent of all new enterprise software will be assembled using vibe coding techniques.
The appeal is obvious. When a revenue operations manager can prompt an AI tool to build a pricing calculator in an afternoon—one that would have previously required either a vendor contract or a development sprint—the math changes. When a recruiter can generate a custom interview training application without filing a single IT ticket, the procurement cycle becomes irrelevant. The frustration with long B2B buying processes, which industry data suggests average 84 days and can stretch past 170 for larger deals, is giving way to a culture of instant, self-service development.
If companies are going to own more of their software again, they are eventually going to confront a question that the cloud era was supposed to have permanently retired: where does it all run?
For the last fifteen years, the dominant answer has been someone else's infrastructure. The cloud-hosted application model, pioneered by Salesforce and embraced by virtually every SaaS provider that followed, freed organizations from maintaining their own hardware. Server rooms became relics. IT budgets shifted from capital expenditure to subscription fees. Companies traded ownership for convenience, and for most of that period, the trade-off was worth it.
But custom-built tools change the equation. A vibe-coded internal dashboard or workflow automation cannot simply be hosted on the vendor's cloud—there is no vendor. Organizations that build their own software must also decide how to deploy, secure, and maintain it. For some, that means leveraging general-purpose cloud infrastructure from providers like AWS or Azure. For others, particularly those in regulated industries or with strict data governance requirements, it increasingly means bringing computing resources back in-house or into private cloud environments.
This does not mean every company is about to rebuild a data center. But it does mean that the era of treating infrastructure as someone else's problem is beginning to fracture. When you own the software, you own the responsibility for everything underneath it—and that is a reality many organizations have not had to face in over a decade.
Perhaps the most consequential shift is what vibe coding does to the architecture of enterprise technology itself. For years, the industry moved toward consolidation: the single platform that handles ERP, CRM, HR, finance, inventory, and e-commerce under one roof. The value proposition was integration. One system of record. One vendor relationship. One place where all the data lives.
Vibe coding inverts that logic. When building a custom tool becomes fast and cheap, organizations no longer need to accept the compromises that come with monolithic platforms. They can cherry-pick the functionality they actually need and build precisely what fits their workflows—nothing more, nothing less. A company might keep its core ERP for financial consolidation and compliance but replace the vendor's project management module with something purpose-built for its specific operations. It might retain the CRM for pipeline tracking but build a custom quoting tool that matches its pricing model exactly, rather than forcing its sales process into a generic template.
The result is a more fragmented technology landscape—one composed of best-of-breed components stitched together rather than a single unified system. This is, of course, exactly what enterprise IT looked like before the consolidation wave. And anyone who lived through that era knows the challenges that come with it: integration complexity, data silos, inconsistent reporting, and the constant need to maintain connections between systems that were never designed to talk to each other.
This brings us to what may be the most underappreciated consequence of the vibe coding revolution: the talent question. When companies relied on vendor-hosted, vendor-maintained SaaS platforms, the internal IT function could remain relatively lean. The vendor handled patches, security updates, infrastructure scaling, and much of the ongoing development. Internal teams focused on configuration, user training, and integration.
Custom-built software demands a fundamentally different staffing model. Someone has to monitor, maintain, update, and troubleshoot these tools. Someone has to ensure they remain secure as threats evolve. Someone has to manage the integrations between a growing number of independent systems. And someone has to understand the AI-generated code well enough to fix it when something breaks—a concern that is more than theoretical, given that research has found between 40 and 60 percent (depending on the study) of developers admit to deploying AI-generated code they do not fully understand.
Companies that embrace vibe coding at scale will need to invest in rebuilding internal IT capabilities that many spent the last two decades deliberately shedding. They will need developers who can review and refine AI output, infrastructure specialists who can manage deployment environments, and integration architects who can keep a fragmented software ecosystem functioning as a coherent whole. These are roles that were common in the early 2000s, became scarce during the SaaS era, and are now poised for a resurgence.
None of this is to suggest that vibe coding is a step backward. The tools available today are incomparably more powerful than anything that existed during the last era of in-house development. What took a team of developers months to build in 2002 can now be prototyped in hours. The barrier to entry is lower, the iteration speed is faster, and the range of what non-technical staff can accomplish is genuinely unprecedented.
But the strategic challenges that come with owning your own software—hosting responsibility, system fragmentation, integration complexity, and the need for skilled internal resources—are not new. They are the same challenges that drove organizations toward packaged software and cloud-hosted platforms in the first place. The technology has changed. The trade-offs have not.
For business leaders evaluating how vibe coding fits into their technology strategy, the lesson from history is straightforward: the ability to build software quickly does not eliminate the cost of running it well. Before replacing a vendor solution with a custom-built alternative, the full lifecycle cost deserves scrutiny—not just development, but deployment, maintenance, security, integration, and the human resources required to sustain it all.
The future of enterprise technology is undeniably being reshaped by AI. But as it takes shape, it is looking remarkably like a place we have been before. The question is whether we remember the lessons we learned the first time around.