Harshal Patil: Sample AI-native PM stack for 2026

What the article is about

Published in May 2026 on Substack, this post by Harshal Patil, a product manager, maps AI tools to specific phases of the product management lifecycle. Rather than reviewing tools in isolation, Patil describes how each tool fits into his working process and what it replaces or improves.

Context

The post is written as a personal snapshot — a practitioner’s view of how AI has changed the mechanics of day-to-day PM work. The organizing principle is explicit: any tool in the stack must satisfy at least one of three criteria — improve decision quality, reduce cycle time, or remove repeated manual work. Tools that fail this test are excluded regardless of how widely they are discussed.

Key argument and method

Patil maps tools to four phases of PM work.

Problem discovery uses HeyMarvin and Dovetail for synthesizing user interviews and qualitative feedback, Cursor connected to BigQuery for SQL query generation and data analysis, and Perplexity Deep Research for competitive intelligence. An n8n workflow converts newsletters into short audio episodes, addressing the information volume problem without adding screen time.

PRD development relies on Notion AI for documentation and user stories, with Cursor as an alternative for markdown workflows where tighter context control is needed. Patil creates customer journey maps using Mermaid diagrams. One specific detail stands out: he uses “pushback prompts” — asking the model to challenge his assumptions rather than elaborate on them. This addresses a real failure mode in AI-assisted writing, where the model tends to reinforce whatever direction the writer is already going.

Prototyping uses Lovable and Cursor to build functional prototypes before engineering commits resources. The argument is that agentic builders have narrowed the gap between a written specification and something stakeholders can actually interact with, which changes the economics of early-stage validation.

Communication and automation covers n8n and MCP integrations for routine task handling, Loom for asynchronous storytelling, and AI-enhanced Grafana integrations for customer onboarding support.

Who it is useful for

Product managers who are already using some AI tools but have not yet organized them into a coherent workflow. The post provides a concrete example of how one practitioner has assembled tools by function rather than by category, which is a more practical framing than comparing tools head-to-head. The adoption filter — decision quality, cycle time, or manual work reduction — is also a useful heuristic for PMs evaluating new tools as they appear.