Case study · 2025

PO Agent

Agentic product-planning platform

Turns a raw product idea into market research, user stories, PRDs and a phased release plan.

Ideaa product ideaResearchTavily web searchScopeMVP, user storiesDocumentPRD, flowchartsReleasephased planthe research stage I ownedorchestration I built, end to end Ideaa product ideaResearchTavily web searchScopeMVP, user storiesDocumentPRD, flowchartsReleasephased planthe researchstage I ownedorchestration Ibuilt, end toend
How PO Agent works. The coloured notes are my annotations.

What it does

PO Agent turns a product idea into a plan. The idea enters a stateful, multi-stage LangGraph workflow that researches the market on the live web, defines the product’s scope, writes the technical documentation and lays out a phased release plan.

I built the foundational orchestration and stateful workflow, and owned the Tavily-powered research stage.

How it works

  1. Idea

    A product idea starts a stateful, multi-stage LangGraph workflow.

  2. Research

    Tavily gathers current market context and evidence from live web sources, producing competitor analysis, expected product workflows and a SWOT analysis.

  3. Scope

    Research becomes an MVP feature hierarchy with user stories, sub-features, expected behavior and delivery requirements.

  4. Document

    The workflow produces PRDs, architecture flowcharts and feature mind maps.

  5. Release

    Work is sequenced into phased release plans. State is persisted in MongoDB and supports RAG retrieval.

What I built

  • Built the foundational LangGraph orchestration and stateful workflow for a multi-stage, agentic product-planning platform.
  • Converted an initial product idea into structured research, product scope, technical documentation and release-planning stages.
  • Owned a Tavily-enabled research workflow that gathers current market context and supporting evidence from live web sources.
  • Generated competitor analysis, expected product workflows and SWOT analysis to establish the opportunity and product direction.
  • Transformed research into MVP feature hierarchies, user stories, sub-features, expected behavior and delivery requirements.
  • Orchestrated PRDs, architecture flowcharts, feature mind maps and phased release plans; persisted state in MongoDB and supported RAG retrieval.