Vibecoding Without Developers: 6 Weeks, 30 People, Working MVPs

CORPUS Consulting launched an in-house course on vibe coding: Over the course of six weeks, 30 employees with diverse backgrounds—ranging from presales analysts to HR managers—built working MVPs. Not just prototypes in Figma, but actual applications with databases, agents, and deployment. One of the products, Tender Radar, has already been tested with real data and is 80% ready for launch. Does this sound like marketing? In part, yes—but there’s an important trade-off behind it that’s worth examining. The main mistake when discussing vibe coding is reducing it all to a tool. As if giving someone Cursor or Claude means they’ll do everything on their own. The KORUS case study shows the opposite: the working products were successful not because the neural network is magical, but because a corporate framework was built around it. A master class on product concept. Selection limited to those who came with a real idea. Six three-hour sessions on Saturdays. In-house experts and mentors. A dedicated chat for questions. Deadline. Without this structure, only a handful out of 30 people would have reached MVP—that’s always the case when a tool is provided without context. Here, context was provided. What exactly was built? Three products, each addressing a specific pain point. Vitaly Mitrofanov, a presales analyst, created a “tender radar”: the app collects data on tenders from external sources, analyzes it using AI, and ranks the results by relevance. They’ve already conducted a test launch and verified the results manually—the service identified some interesting contracts. They’re currently integrating a second data source. Vladimir Gordeev built a best-practices system: experts document successful processes in the form of checklists, colleagues subscribe, and receive updates. The idea had been in his head for several years, but he lacked both the time and the manpower—the course provided a deadline and a structure. Evgeniya Miklush created an HR analytics platform that collects labor market data from multiple sources and generates a dashboard with metrics and a summary. It’s currently in test mode; the next step is to integrate the data into the company’s internal analytics. Here’s what’s important: out of ten projects, several are already ready to be refined for implementation. This isn’t a 100% conversion rate, and KORUS is upfront about that. But the value of the course lies not only in the products. Maxim Sumin, head of the Corporate University, puts it bluntly: People who complete the course become AI ambassadors within the company. Through them, the practice spreads across departments. This is a smart strategy—not to try to convince skeptics, but to create an environment where the first product speaks louder than any webinar. What’s the catch? First, six weeks of three hours on Saturdays adds up to 18 hours of classes plus independent study. Not every employee is up for that. Second, the mentors are in-house experts, which means their time also costs money. Third, the products are still in beta, and the path from MVP to production deployment is usually longer than it seems. Vibe Coding lowers the barrier to entry for development, but it doesn’t eliminate the need for technical expertise during the refinement phase. A practical takeaway for anyone considering this approach: don’t launch VibeCoding as a hackathon for fun. Give people a real task, a deadline, and a mentor. The tool is the final link in the chain, not the first. Source: Vibecoding for Everyone: How KORUS Employees Build AI Products Without Development Experience / Habr

Source: All Articles in a Row / Artificial Intelligence / Habr