John De Goes’ blog

I’ve been a Product Manager for over 10 years, leading teams in fintech and AI startups, and AI tools have transformed how I handle coding tasks without being a developer myself. By following a structured process with Perplexity and Claude Code in VS Code, I’ve cut development cycles dramatically while ensuring high-quality outputs—here’s the exact workflow that’s boosted my team’s productivity.

My Core Stack
I rely on Perplexity for research, market analysis, and PM tasks like drafting PRDs or competitive intel—its cited, real-time answers save hours compared to traditional search, making it ideal for quick, reliable insights. For coding, I use Claude Code embedded in Visual Studio Code, which offers seamless context-aware assistance, inline diffs, and keyboard shortcuts for faster iteration.
This combo lets me prototype features in days, not weeks, by handling research upfront and code generation precisely.

Step 1: Build a Cross-Functional “Team” of Personas
I never feed raw coding instructions to Claude Code—instead, I maintain personas in a Perplexity Space file for Product, Marketing, UX Research, Legal, and more. I present the problem to this virtual team, and they output polished requirement descriptions plus tailored prompts for Claude Code.
This “team review” catches edge cases early, aligns on business goals, and produces specs that are concrete and feasible, much like data-driven PRDs recommended for AI products.

Step 2: Hand Off to the Lead Engineer Persona
In Claude Code, I have a detailed Lead Engineer persona file outlining their approach—experience level, coding standards, testing habits, and focus on scalability. Every task goes through this persona first for a plan review before implementation.
It mimics real engineering rigor, reducing bugs by embedding best practices like breaking tasks into small units and providing rich context from the start.

Step 3: Enforce Standards with Shared Best Practices
My repo includes a “Best Practices” and “Coding Process” file visible to the whole team in Claude Code—defining the exact flow for instructions, my Definition of Done, security checks, and testing requirements. This ensures consistency every time.
Team members see the same guardrails, fostering collaboration; I’ve seen it prevent rework and align outputs with production needs.

Step 4: Always Use PRs for Changes
No direct commits—every code change goes through a Pull Request for review, discussion, and approval. This final gate catches issues and builds team buy-in.

The Impact I’ve Seen
Following these rules has slashed feature delivery time by 70% in my projects, empowered non-technical PMs to lead tech work confidently, and improved code quality through structured prompts and reviews. Other PMs adopting similar flows report faster prototyping and better stakeholder alignment—try it to enhance your productivity and let me know your results.