A Product Engineer Challenge for Non-Coding Planners: From Server to App Through Preparation and Communication

A service planner at Yogiyo shares their experience working as a Product Engineer (PE) and launching a feature by directly developing everything from the server and admin panel to Android and iOS apps, starting with zero knowledge of code. This project utilized map markers on the order tracking screen as ad placements, achieving over 200% of the target for impressions and clicks in its first external campaign after its launch on June 17, 2026.

A Development Process Centered on Preparation and Communication

The key to a non-developer planner completing development was not coding skills, but 'preparation' and 'communication.' First, the development environment was set up using guide documents written by backend developers, and Claude Code was used to analyze the scope of impact from the initial planning draft. This allowed for specific verification questions to be asked to developers instead of vague inquiries.

The core tool for the project was a design plan written in Markdown. By organizing the context in a way that was easy for AI to read, this document consolidated the scope of impact, changes, and finalized policies, maintaining consistent context even when working across multiple areas. When specifications changed, the document was immediately updated and verified via Claude Code in the terminal.

Success Conditions for a Non-Developer PE

While utilizing AI to its fullest extent, communication with existing service owners was prioritized. Design reviews before development helped catch potential oversights, and AI was instructed to follow existing code conventions to ensure maintainability. The completed code always underwent code review by the responsible developer and passed full testing by the QA team to ensure stability.

The following conditions are necessary for a non-developer to succeed as a PE:

• Domain Knowledge: While AI can write code, it cannot make business policy decisions such as whether to cancel an order, making the planner's domain understanding essential.
• Pre-communication: Concerns from existing operators must be resolved by clearly explaining the reasons for changes and the post-launch management plan.
• Verification Habit: There is a need to avoid blindly trusting AI's answers and instead develop the habit of re-asking in different ways or testing directly.
• Error Interpretation Environment: One must be familiar with server log monitoring tools (such as Datadog) or have an environment where AI can interpret logs.

Ultimately, while AI cannot completely erase the boundary between planners and developers, it serves as a bridge to understand each other's languages. Individuals must prepare enough to ask questions and verify AI's answers, while organizations must provide guides that non-developers can follow and environments for interpreting errors.