Medium: 6 AI workflows every product manager should be using
What the article is about
Written by Blaine Joubert and published on Medium in February 2026, this article describes six AI workflows the author adopted in day-to-day product management work. It is not a survey of available tools but a direct account of what made a tangible difference: what was attempted, what the output looked like, and where real limits appeared. The central framing is that AI’s primary value is lowering the cost of iteration — making it cheaper and faster to test assumptions before committing to larger investments.
Context
Joubert works in the entertainment and technology sector, which shapes some of the examples (video metadata, influencer vetting, HeyGen-based demo production). The workflows are broadly applicable, though a few require familiarity with platforms like Google Apps Script or Databricks.
The six workflows
Prototype first, document second. Rather than writing PRDs from assumptions, Joubert uses AI to generate functional HTML prototypes, shares them with stakeholders via Google Sites, collects real reactions, and only then uses AI to produce PRDs and user stories anchored in validated feedback. The insight is that documentation written before testing describes imagined behavior, while documentation written after reflects what people actually responded to.
Build internal MVPs while waiting for resources. When engineering bandwidth is unavailable, Joubert uses AI alongside Google Sheets and Apps Script to construct rough internal tools — automated metadata scrapers, AI-powered vetting systems — that unblock teams and surface constraints before formal development begins. These sandboxes are never pushed to production unreviewed, but they make the waiting period productive.
Use AI to shape strategy decks. Treating AI as a thinking partner for narrative structure: providing business goals, user problems, and constraints, then asking questions about what angle to lead with and what concerns executives are likely to raise. The workflow helps maintain strategic altitude and avoid excessive operational detail.
Shorten the data debugging loop. Feeding SQL queries and metric definitions to AI for sanity-checking before escalating to data teams. The model reliably flags common errors — incorrect joins, double-counting, filter mistakes — which means conversations with data analysts start from a more informed baseline.
Turn demos into engaging experiences. Using HeyGen, ChatGPT, and Suno to turn routine product walkthroughs into produced demonstrations with custom voiceovers and music. The barrier to polished presentation dropped significantly without additional budget.
Use AI intentionally inside products. Rather than adding a general-purpose LLM to existing workflows, apply it to specific, bounded problems. The example is a content moderation system where AI provides context and reasoning to support human moderators, reducing false positives without removing human judgment from the loop.
Who it is useful for
Product managers who already understand the basics and are looking for concrete applications beyond writing assistance. The article assumes familiarity with PRDs, MVPs, and stakeholder alignment. It is particularly useful for PMs working in environments where engineering resources are constrained, since several workflows are specifically designed to make that constraint less blocking.