Mohit Aggarwal: Building a systematic AI toolkit as a product manager in 2026
Published in March 2026 in the Product Notes publication on Medium, this article by Mohit Aggarwal makes a specific argument: the gap between PMs who get real returns from AI and those who do not comes down to whether they have built a system or are still treating AI as a collection of individual tools.
The argument
Most product managers use AI tools — ChatGPT for drafting, Claude for synthesis, Gemini for search — but switch between them without a connecting layer. Each session starts cold, with no memory of prior work, no shared context across tools, and no integration with the systems where product work actually lives. The result is that AI assists with isolated tasks but never compounds across a project or a quarter.
Aggarwal’s proposal is to build a toolkit with three structural properties: reusable prompts organized around recurring PM tasks, a persistent context layer that retains product knowledge across sessions, and integrations that connect AI outputs directly to Jira, Slack, or whatever task management system the team uses. The Model Context Protocol (MCP) is discussed as the mechanism for those integrations — a standard that lets AI tools read from and write to external systems in a structured way.
What this looks like in practice
The article describes how a PM working on a B2B SaaS product might set up: a Claude Project that holds product specs, user research summaries, and competitive notes; a prompt library for standard deliverables like PRD sections, interview guides, and sprint review summaries; and an MCP connection that lets the AI pull open Jira tickets as context when drafting planning documents.
The framing is not prescriptive about which specific tools to use. The point is the structural pattern — persistent context, reusable prompts, external integrations — which can be assembled from different combinations of tools depending on what a team already has access to.
Who it is useful for
Product managers who have been using AI for several months and feel like they are getting limited returns. The article is not an introduction to AI tools; it assumes basic familiarity and speaks to practitioners who want to get past the initial productivity plateau. The guidance is applicable to PMs working individually as well as those trying to create shared AI workflows for a team.
The article closes by noting that this kind of setup can be built entirely from free tools, which matters for PMs at smaller organizations or those experimenting before seeking broader team adoption.