Beware the 'Vibe-Coded' App: The Hidden Cost of AI-Assisted Development

The promise of rapid software development powered by generative AI is creating an unforeseen and costly side effect: a surge in poorly constructed applications. A new cottage industry of software consultancies is now emerging to clean up what is being termed “AI slop” or “vibe-coded” applications. These are programs often built by users with limited software engineering expertise who rely on tools like Claude Code and Codex to generate code. While the resulting applications might look good and appear functional at first glance, they are often riddled with underlying flaws, creating a new and expensive form of technical debt.

Firms like Slopfix and Redwerk are finding a growing market in refactoring these vibe-coded codebases back to a maintainable state. According to Konstantin Klyagin, founder of QAwerk, a quality assurance consultancy, clients are increasingly seeking help with applications that suffer from common AI-generated flaws. These issues include code duplication, such as one app that presented different prices on its homepage and during its payment flow, and significant security problems, like permission-handling bugs that allowed users to skip profile creation and proceed directly to payment. Other frequent problems are poor accessibility for screen readers and incomplete test coverage, which together create serious operational and security risks.

This trend matters to the finance function because it directly challenges the simple productivity narrative often used to justify investment in AI tools. While AI assistants can undoubtedly make developers faster at producing code, the lack of underlying engineering discipline can mean that organisations are simply creating flawed, insecure, and unmaintainable products at a greater velocity. The initial speed gains can be quickly eroded by the subsequent costs of extensive quality assurance, bug fixing, security patching, and, in worst-case scenarios, complete rewrites by expensive specialists. This is a classic case of hidden liabilities accumulating on the technical balance sheet.

For the CFO, the key implication is the need for deeper scrutiny of the total cost of ownership (TCO) for AI-assisted software development. Business cases that champion faster feature delivery must be balanced against the potential for direct revenue leakage from faulty payment logic or catastrophic financial penalties from security breaches. Before approving budgets based on promised efficiency gains, finance leaders must question how the organisation will mitigate the risk of producing AI slop. This means ensuring that any adoption of AI coding tools is accompanied by a renewed commitment to disciplined software architecture, rigorous testing, and comprehensive code reviews, treating these not as hindrances to speed but as essential controls for managing long-term costs and risks.