Vibe‑coded solutions arrive for problems nobody has

The boss’s AI‑wrapped app: a case study in hype‑driven development

The latest episode of the BOFH column (Episode 10, Sigh) introduces a new internal app that the boss proudly calls “AI‑powered” and “for everyone”. The premise is simple: a mobile tool that tells a new employee where the printer paper is, how to check toner levels, how to adjust the air‑conditioning, and even how to craft a compliant password. On the surface it sounds useful, but the conversation between the BOFH, the PFY and the boss reveals a deeper pattern that’s spreading across many enterprises – vibe‑coding.

What “vibe‑coding” looks like in practice

• AI as a veneer – The boss repeatedly leans on the phrase “it has AI in it” as a justification, yet the described functionality is nothing more than a scripted FAQ. The app does not actually query printer telemetry, nor does it integrate with the building‑management system. It merely offers generic, static advice that any junior admin could write in a markdown file.

• Reinventing the wheel – The BOFH suspects the codebase is a 40 000‑line mess of improvised BASIC‑style logic. Even if the language were modern, the core problem – “where is the paper?” – is already solved by existing tools such as Microsoft 365’s [SharePoint lists](https://learn.microsoft.com/sharepoint/lists) or simple inventory‑tracking bots on Slack.

• Mandated deployment – Management’s decision to push the app onto every company phone is a classic top‑down rollout of a low‑value product. The cost isn’t just the development time; it’s the ongoing maintenance, device‑management overhead, and the risk of user fatigue when employees are bombarded with irrelevant prompts.

Evidence from the field

1. Feature creep without user research – The boss lists a laundry list of potential features (printer troubleshooting, AC usage tips, travel‑option recommendations). None of these have been validated with real users. A recent [Gartner survey](https://www.gartner.com/en/newsroom/press-releases/2024-02-20-gartner-survey-reveals-70-percent-of-enterprise-apps-are-underutilized) found that 70 % of internally‑developed apps see less than 10 % adoption.

2. AI hallucinations – When the boss claims the app can “explain how to use the air‑conditioning system based on the current environment”, the BOFH points out that without real‑time sensor data the AI will hallucinate. This mirrors the broader industry issue where large language models generate plausible‑but‑incorrect instructions, leading to support tickets rather than solving them.

3. Opportunity cost – The PFY, who has previously built a niche “faulty‑window safety‑catch” tool, hints that the boss’s app could have been a platform for genuine productivity gains (e.g., automated asset discovery). Instead, time is spent on a generic knowledge base that could be delivered via a [Confluence page](https://www.atlassian.com/software/confluence) or a chatbot integrated with existing ITSM tools.

Featured image – a typical corporate rollout announcement, juxtaposed with the reality of a half‑baked internal app.

Counter‑perspectives: why some teams still back such projects

• Speed of delivery – Proponents argue that a quickly built “vibe‑coded” prototype can surface hidden pain points. In fast‑moving startups, a rough‑and‑ready tool can be iterated upon based on real feedback.
• Visibility for the boss – From a leadership standpoint, delivering a shiny‑looking app demonstrates initiative, even if the underlying value is modest. This can be a political win in organizations where visible output matters more than measurable impact.
• Learning opportunity – Junior developers sometimes use these projects as a sandbox to experiment with AI APIs (e.g., OpenAI’s [ChatGPT API](https://platform.openai.com/docs/api-reference)) and get hands‑on experience with mobile deployment pipelines.

The bigger picture: shifting development culture

The BOFH’s skepticism is not a personal gripe; it reflects a cultural shift where AI‑assisted code generation is often treated as a shortcut rather than a tool that requires disciplined engineering practices. When the “AI wrapper” becomes the primary selling point, teams risk overlooking:

• Testing and observability – Auto‑generated code rarely includes unit tests or logging, making debugging a nightmare.
• Security hygiene – Embedding AI prompts without sanitization can expose the app to prompt‑injection attacks, especially when the app interacts with internal APIs.
• Maintainability – A 40 000‑line monolith written in a mix of generated snippets becomes a nightmare for future developers, increasing technical debt.

What should organizations do?

1. Validate the problem first – Conduct user interviews and prototype with low‑fidelity tools (paper sketches, spreadsheets) before committing to a full‑stack app.
2. Leverage existing platforms – Before building a custom solution, explore whether the functionality already exists in tools like [Microsoft Power Apps](https://powerapps.microsoft.com) or [Slack workflow builder](https://slack.com/help/articles/360035692513-Create-workflows-with-Workflow-Builder).
3. Treat AI as a co‑pilot, not a driver – Use AI to generate snippets, documentation, or test data, but enforce code reviews, static analysis, and comprehensive testing.
4. Measure adoption early – Deploy a minimal viable version to a small pilot group, track usage metrics, and iterate only if the data shows real value.

Closing thoughts

The boss’s “vibe‑coded” app is a vivid illustration of a broader trend: the allure of AI‑wrapped solutions can eclipse the fundamentals of good software design. While there is room for rapid experimentation, the real competitive advantage comes from solving real problems with real data, not from dressing a thin idea in the language of artificial intelligence.

---

*For more BOFH episodes and a deeper dive into the cultural quirks of enterprise IT, visit the [BOFH archive](https://www.theregister.com/BOFH).*